fix: doc store for files larger 4GB (#1856)

Fixes an issue in the skip list deserialization, which deserialized the byte start offset incorrectly as u32.
`get_doc` will fail for any docs that live in a block with start offset larger than u32::MAX (~4GB).
Causes index corruption, if a segment with a doc store larger 4GB is merged.

tantivy version 0.19 is affected

Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "tantivy"
-version = "0.19.0"
+version = "0.19.1-quickwit"
 authors = ["Paul Masurel <paul.masurel@gmail.com>"]
 license = "MIT"
 categories = ["database-implementations", "data-structures"]
@@ -23,7 +23,7 @@ regex = { version = "1.5.5", default-features = false, features = ["std", "unico
 aho-corasick = "0.7"
 tantivy-fst = "0.4.0"
 memmap2 = { version = "0.5.3", optional = true }
-lz4_flex = { version = "0.9.2", default-features = false, features = ["checked-decode"], optional = true }
+lz4_flex = { version = "0.10", default-features = false, features = ["checked-decode"], optional = true }
 brotli = { version = "3.3.4", optional = true }
 zstd = { version = "0.12", optional = true, default-features = false }
 snap = { version = "1.0.5", optional = true }
@@ -55,6 +55,7 @@ measure_time = "0.8.2"
 async-trait = "0.1.53"
 arc-swap = "1.5.0"
 
+#columnar = { version="0.1", path="./columnar", package ="tantivy-columnar" }
 sstable = { version="0.1", path="./sstable", package ="tantivy-sstable", optional = true }
 stacker = { version="0.1", path="./stacker", package ="tantivy-stacker" }
 tantivy-query-grammar = { version= "0.19.0", path="./query-grammar" }
@@ -107,7 +108,7 @@ unstable = [] # useful for benches.
 quickwit = ["sstable"]
 
 [workspace]
-members = ["query-grammar", "bitpacker", "common", "fastfield_codecs", "ownedbytes", "stacker", "sstable", "tokenizer-api"]
+members = ["query-grammar", "bitpacker", "common", "fastfield_codecs", "ownedbytes", "stacker", "sstable", "tokenizer-api", "columnar"]
 
 # Following the "fail" crate best practises, we isolate
 # tests that define specific behavior in fail check points

README.md

@@ -41,7 +41,7 @@ Your mileage WILL vary depending on the nature of queries and their load.
 - SIMD integer compression when the platform/CPU includes the SSE2 instruction set
 - Single valued and multivalued u64, i64, and f64 fast fields (equivalent of doc values in Lucene)
 - `&[u8]` fast fields
-- Text, i64, u64, f64, dates, and hierarchical facet fields
+- Text, i64, u64, f64, dates, ip, bool, and hierarchical facet fields
 - Compressed document store (LZ4, Zstd, None, Brotli, Snap)
 - Range queries
 - Faceted search
@@ -80,56 +80,21 @@ There are many ways to support this project.
 # Contributing code
 
 We use the GitHub Pull Request workflow: reference a GitHub ticket and/or include a comprehensive commit message when opening a PR.
+Feel free to update CHANGELOG.md with your contribution.
 
 ## Tokenizer
 
 When implementing a tokenizer for tantivy depend on the `tantivy-tokenizer-api` crate.
 
-## Minimum supported Rust version
-
-Tantivy currently requires at least Rust 1.62 or later to compile.
-
 ## Clone and build locally
 
 Tantivy compiles on stable Rust.
 To check out and run tests, you can simply run:
 
 ```bash
-    git clone https://github.com/quickwit-oss/tantivy.git
-    cd tantivy
-    cargo build
-```
-
-## Run tests
-
-Some tests will not run with just `cargo test` because of `fail-rs`.
-To run the tests exhaustively, run `./run-tests.sh`.
-
-## Debug
-
-You might find it useful to step through the programme with a debugger.
-
-### A failing test
-
-Make sure you haven't run `cargo clean` after the most recent `cargo test` or `cargo build` to guarantee that the `target/` directory exists. Use this bash script to find the name of the most recent debug build of Tantivy and run it under `rust-gdb`:
-
-```bash
-find target/debug/ -maxdepth 1 -executable -type f -name "tantivy*" -printf '%TY-%Tm-%Td %TT %p\n' | sort -r | cut -d " " -f 3 | xargs -I RECENT_DBG_TANTIVY rust-gdb RECENT_DBG_TANTIVY
-```
-
-Now that you are in `rust-gdb`, you can set breakpoints on lines and methods that match your source code and run the debug executable with flags that you normally pass to `cargo test` like this:
-
-```bash
-$gdb run --test-threads 1 --test $NAME_OF_TEST
-```
-
-### An example
-
-By default, `rustc` compiles everything in the `examples/` directory in debug mode. This makes it easy for you to make examples to reproduce bugs:
-
-```bash
-rust-gdb target/debug/examples/$EXAMPLE_NAME
-$ gdb run
+git clone https://github.com/quickwit-oss/tantivy.git
+cd tantivy
+cargo test
 ```
 
 # Companies Using Tantivy

benches/index-bench.rs

@@ -34,7 +34,7 @@ pub fn hdfs_index_benchmark(c: &mut Criterion) {
             let index = Index::create_in_ram(schema.clone());
             let index_writer = index.writer_with_num_threads(1, 100_000_000).unwrap();
             for _ in 0..NUM_REPEATS {
-                for doc_json in HDFS_LOGS.trim().split("\n") {
+                for doc_json in HDFS_LOGS.trim().split('\n') {
                     let doc = schema.parse_document(doc_json).unwrap();
                     index_writer.add_document(doc).unwrap();
                 }
@@ -46,7 +46,7 @@ pub fn hdfs_index_benchmark(c: &mut Criterion) {
             let index = Index::create_in_ram(schema.clone());
             let mut index_writer = index.writer_with_num_threads(1, 100_000_000).unwrap();
             for _ in 0..NUM_REPEATS {
-                for doc_json in HDFS_LOGS.trim().split("\n") {
+                for doc_json in HDFS_LOGS.trim().split('\n') {
                     let doc = schema.parse_document(doc_json).unwrap();
                     index_writer.add_document(doc).unwrap();
                 }
@@ -59,7 +59,7 @@ pub fn hdfs_index_benchmark(c: &mut Criterion) {
             let index = Index::create_in_ram(schema_with_store.clone());
             let index_writer = index.writer_with_num_threads(1, 100_000_000).unwrap();
             for _ in 0..NUM_REPEATS {
-                for doc_json in HDFS_LOGS.trim().split("\n") {
+                for doc_json in HDFS_LOGS.trim().split('\n') {
                     let doc = schema.parse_document(doc_json).unwrap();
                     index_writer.add_document(doc).unwrap();
                 }
@@ -71,7 +71,7 @@ pub fn hdfs_index_benchmark(c: &mut Criterion) {
             let index = Index::create_in_ram(schema_with_store.clone());
             let mut index_writer = index.writer_with_num_threads(1, 100_000_000).unwrap();
             for _ in 0..NUM_REPEATS {
-                for doc_json in HDFS_LOGS.trim().split("\n") {
+                for doc_json in HDFS_LOGS.trim().split('\n') {
                     let doc = schema.parse_document(doc_json).unwrap();
                     index_writer.add_document(doc).unwrap();
                 }
@@ -85,7 +85,7 @@ pub fn hdfs_index_benchmark(c: &mut Criterion) {
             let json_field = dynamic_schema.get_field("json").unwrap();
             let mut index_writer = index.writer_with_num_threads(1, 100_000_000).unwrap();
             for _ in 0..NUM_REPEATS {
-                for doc_json in HDFS_LOGS.trim().split("\n") {
+                for doc_json in HDFS_LOGS.trim().split('\n') {
                     let json_val: serde_json::Map<String, serde_json::Value> =
                         serde_json::from_str(doc_json).unwrap();
                     let doc = tantivy::doc!(json_field=>json_val);
@@ -101,7 +101,7 @@ pub fn hdfs_index_benchmark(c: &mut Criterion) {
             let json_field = dynamic_schema.get_field("json").unwrap();
             let mut index_writer = index.writer_with_num_threads(1, 100_000_000).unwrap();
             for _ in 0..NUM_REPEATS {
-                for doc_json in HDFS_LOGS.trim().split("\n") {
+                for doc_json in HDFS_LOGS.trim().split('\n') {
                     let json_val: serde_json::Map<String, serde_json::Value> =
                         serde_json::from_str(doc_json).unwrap();
                     let doc = tantivy::doc!(json_field=>json_val);

bitpacker/Cargo.toml

@@ -15,3 +15,7 @@ homepage = "https://github.com/quickwit-oss/tantivy"
 # See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
 
 [dependencies]
+
+[dev-dependencies]
+rand = "0.8"
+proptest = "1"

bitpacker/benches/bench.rs

@@ -4,9 +4,39 @@ extern crate test;
 
 #[cfg(test)]
 mod tests {
-    use tantivy_bitpacker::BlockedBitpacker;
+    use rand::seq::IteratorRandom;
+    use rand::thread_rng;
+    use tantivy_bitpacker::{BitPacker, BitUnpacker, BlockedBitpacker};
     use test::Bencher;
 
+    #[inline(never)]
+    fn create_bitpacked_data(bit_width: u8, num_els: u32) -> Vec<u8> {
+        let mut bitpacker = BitPacker::new();
+        let mut buffer = Vec::new();
+        for _ in 0..num_els {
+            // the values do not matter.
+            bitpacker.write(0u64, bit_width, &mut buffer).unwrap();
+            bitpacker.flush(&mut buffer).unwrap();
+        }
+        buffer
+    }
+
+    #[bench]
+    fn bench_bitpacking_read(b: &mut Bencher) {
+        let bit_width = 3;
+        let num_els = 1_000_000u32;
+        let bit_unpacker = BitUnpacker::new(bit_width);
+        let data = create_bitpacked_data(bit_width, num_els);
+        let idxs: Vec<u32> = (0..num_els).choose_multiple(&mut thread_rng(), 100_000);
+        b.iter(|| {
+            let mut out = 0u64;
+            for &idx in &idxs {
+                out = out.wrapping_add(bit_unpacker.get(idx, &data[..]));
+            }
+            out
+        });
+    }
+
     #[bench]
     fn bench_blockedbitp_read(b: &mut Bencher) {
         let mut blocked_bitpacker = BlockedBitpacker::new();
@@ -14,9 +44,9 @@ mod tests {
             blocked_bitpacker.add(val * val);
         }
         b.iter(|| {
-            let mut out = 0;
+            let mut out = 0u64;
             for val in 0..=21500 {
-                out = blocked_bitpacker.get(val);
+                out = out.wrapping_add(blocked_bitpacker.get(val));
             }
             out
         });

bitpacker/src/bitpacker.rs

@@ -56,27 +56,31 @@ impl BitPacker {
 
     pub fn close<TWrite: io::Write>(&mut self, output: &mut TWrite) -> io::Result<()> {
         self.flush(output)?;
-        // Padding the write file to simplify reads.
-        output.write_all(&[0u8; 7])?;
         Ok(())
     }
 }
 
-#[derive(Clone, Debug, Default)]
+#[derive(Clone, Debug, Default, Copy)]
 pub struct BitUnpacker {
-    num_bits: u64,
+    num_bits: u32,
     mask: u64,
 }
 
 impl BitUnpacker {
+    /// Creates a bit unpacker, that assumes the same bitwidth for all values.
+    ///
+    /// The bitunpacker works by doing an unaligned read of 8 bytes.
+    /// For this reason, values of `num_bits` between
+    /// [57..63] are forbidden.
     pub fn new(num_bits: u8) -> BitUnpacker {
+        assert!(num_bits <= 7 * 8 || num_bits == 64);
         let mask: u64 = if num_bits == 64 {
             !0u64
         } else {
             (1u64 << num_bits) - 1u64
         };
         BitUnpacker {
-            num_bits: u64::from(num_bits),
+            num_bits: u32::from(num_bits),
             mask,
         }
     }
@@ -87,28 +91,40 @@ impl BitUnpacker {
 
     #[inline]
     pub fn get(&self, idx: u32, data: &[u8]) -> u64 {
-        if self.num_bits == 0 {
-            return 0u64;
-        }
-        let addr_in_bits = idx * self.num_bits as u32;
+        let addr_in_bits = idx * self.num_bits;
         let addr = (addr_in_bits >> 3) as usize;
+        if addr + 8 > data.len() {
+            if self.num_bits == 0 {
+                return 0;
+            }
+            let bit_shift = addr_in_bits & 7;
+            return self.get_slow_path(addr, bit_shift, data);
+        }
         let bit_shift = addr_in_bits & 7;
-        debug_assert!(
-            addr + 8 <= data.len(),
-            "The fast field field should have been padded with 7 bytes."
-        );
         let bytes: [u8; 8] = (&data[addr..addr + 8]).try_into().unwrap();
         let val_unshifted_unmasked: u64 = u64::from_le_bytes(bytes);
         let val_shifted = val_unshifted_unmasked >> bit_shift;
         val_shifted & self.mask
     }
+
+    #[inline(never)]
+    fn get_slow_path(&self, addr: usize, bit_shift: u32, data: &[u8]) -> u64 {
+        let mut bytes: [u8; 8] = [0u8; 8];
+        let available_bytes = data.len() - addr;
+        // This function is meant to only be called if we did not have 8 bytes to load.
+        debug_assert!(available_bytes < 8);
+        bytes[..available_bytes].copy_from_slice(&data[addr..]);
+        let val_unshifted_unmasked: u64 = u64::from_le_bytes(bytes);
+        let val_shifted = val_unshifted_unmasked >> bit_shift;
+        val_shifted & self.mask
+    }
 }
 
 #[cfg(test)]
 mod test {
     use super::{BitPacker, BitUnpacker};
 
-    fn create_fastfield_bitpacker(len: usize, num_bits: u8) -> (BitUnpacker, Vec<u64>, Vec<u8>) {
+    fn create_bitpacker(len: usize, num_bits: u8) -> (BitUnpacker, Vec<u64>, Vec<u8>) {
         let mut data = Vec::new();
         let mut bitpacker = BitPacker::new();
         let max_val: u64 = (1u64 << num_bits as u64) - 1u64;
@@ -119,13 +135,13 @@ mod test {
             bitpacker.write(val, num_bits, &mut data).unwrap();
         }
         bitpacker.close(&mut data).unwrap();
-        assert_eq!(data.len(), ((num_bits as usize) * len + 7) / 8 + 7);
+        assert_eq!(data.len(), ((num_bits as usize) * len + 7) / 8);
         let bitunpacker = BitUnpacker::new(num_bits);
         (bitunpacker, vals, data)
     }
 
     fn test_bitpacker_util(len: usize, num_bits: u8) {
-        let (bitunpacker, vals, data) = create_fastfield_bitpacker(len, num_bits);
+        let (bitunpacker, vals, data) = create_bitpacker(len, num_bits);
         for (i, val) in vals.iter().enumerate() {
             assert_eq!(bitunpacker.get(i as u32, &data), *val);
         }
@@ -139,4 +155,49 @@ mod test {
         test_bitpacker_util(6, 14);
         test_bitpacker_util(1000, 14);
     }
+
+    use proptest::prelude::*;
+
+    fn num_bits_strategy() -> impl Strategy<Value = u8> {
+        prop_oneof!(Just(0), Just(1), 2u8..56u8, Just(56), Just(64),)
+    }
+
+    fn vals_strategy() -> impl Strategy<Value = (u8, Vec<u64>)> {
+        (num_bits_strategy(), 0usize..100usize).prop_flat_map(|(num_bits, len)| {
+            let max_val = if num_bits == 64 {
+                u64::MAX
+            } else {
+                (1u64 << num_bits as u32) - 1
+            };
+            let vals = proptest::collection::vec(0..=max_val, len);
+            vals.prop_map(move |vals| (num_bits, vals))
+        })
+    }
+
+    fn test_bitpacker_aux(num_bits: u8, vals: &[u64]) {
+        let mut buffer: Vec<u8> = Vec::new();
+        let mut bitpacker = BitPacker::new();
+        for &val in vals {
+            bitpacker.write(val, num_bits, &mut buffer).unwrap();
+        }
+        bitpacker.flush(&mut buffer).unwrap();
+        assert_eq!(buffer.len(), (vals.len() * num_bits as usize + 7) / 8);
+        let bitunpacker = BitUnpacker::new(num_bits);
+        let max_val = if num_bits == 64 {
+            u64::MAX
+        } else {
+            (1u64 << num_bits) - 1
+        };
+        for (i, val) in vals.iter().copied().enumerate() {
+            assert!(val <= max_val);
+            assert_eq!(bitunpacker.get(i as u32, &buffer), val);
+        }
+    }
+
+    proptest::proptest! {
+        #[test]
+        fn test_bitpacker_proptest((num_bits, vals) in vals_strategy()) {
+            test_bitpacker_aux(num_bits, &vals);
+        }
+    }
 }

columnar/Cargo.toml

@@ -5,28 +5,23 @@ edition = "2021"
 license = "MIT"
 
 [dependencies]
+itertools = "0.10.5"
+log = "0.4.17"
+fnv = "1.0.7"
+fastdivide = "0.4.0"
+rand = { version = "0.8.5", optional = true }
+measure_time = { version = "0.8.2", optional = true }
+prettytable-rs = { version = "0.10.0", optional = true }
+
 stacker = { path = "../stacker", package="tantivy-stacker"}
-serde_json = "1"
-thiserror = "1"
-fnv = "1"
 sstable = { path = "../sstable", package = "tantivy-sstable" }
 common = { path = "../common", package = "tantivy-common" }
-itertools = "0.10"
-log = "0.4"
 tantivy-bitpacker = { version= "0.3", path = "../bitpacker/" }
-prettytable-rs = {version="0.10.0", optional= true}
-rand = {version="0.8.3", optional= true}
-fastdivide = "0.4"
-measure_time = { version="0.8.2", optional=true}
 
 [dev-dependencies]
-proptest = "1"
-more-asserts = "0.3.0"
-rand = "0.8.3"
-
-# temporary
-[workspace]
-members = []
+proptest = "1.0.0"
+more-asserts = "0.3.1"
+rand = "0.8.5"
 
 [features]
 unstable = []

columnar/src/TODO.md

@@ -35,6 +35,7 @@ remove all  doc_id occurences -> row_id
 use the rank & select naming in unit tests branch.
 multi-linear -> blockwise
 linear codec -> simply a multiplication for the index column
+rename columnar to something more explicit, like column_dictionary or columnar_table
 
 # Other
 fix enhance column-cli

columnar/src/column/dictionary_encoded.rs

@@ -5,9 +5,16 @@ use std::sync::Arc;
 use sstable::{Dictionary, VoidSSTable};
 
 use crate::column::Column;
-use crate::column_index::ColumnIndex;
+use crate::RowId;
 
 /// Dictionary encoded column.
+///
+/// The column simply gives access to a regular u64-column that, in
+/// which the values are term-ordinals.
+///
+/// These ordinals are ids uniquely identify the bytes that are stored in
+/// the column. These ordinals are small, and sorted in the same order
+/// as the term_ord_column.
 #[derive(Clone)]
 pub struct BytesColumn {
     pub(crate) dictionary: Arc<Dictionary<VoidSSTable>>,
@@ -15,26 +22,57 @@ pub struct BytesColumn {
 }
 
 impl BytesColumn {
+    /// Fills the given `output` buffer with the term associated to the ordinal `ord`.
+    ///
     /// Returns `false` if the term does not exist (e.g. `term_ord` is greater or equal to the
     /// overll number of terms).
-    pub fn term_ord_to_str(&self, term_ord: u64, output: &mut Vec<u8>) -> io::Result<bool> {
-        self.dictionary.ord_to_term(term_ord, output)
+    pub fn ord_to_bytes(&self, ord: u64, output: &mut Vec<u8>) -> io::Result<bool> {
+        self.dictionary.ord_to_term(ord, output)
     }
 
-    pub fn term_ords(&self) -> &Column<u64> {
+    /// Returns the number of rows in the column.
+    pub fn num_rows(&self) -> RowId {
+        self.term_ord_column.num_rows()
+    }
+
+    /// Returns the column of ordinals
+    pub fn ords(&self) -> &Column<u64> {
         &self.term_ord_column
     }
 }
 
-impl Deref for BytesColumn {
-    type Target = ColumnIndex<'static>;
+#[derive(Clone)]
+pub struct StrColumn(BytesColumn);
 
-    fn deref(&self) -> &Self::Target {
-        &**self.term_ords()
+impl From<BytesColumn> for StrColumn {
+    fn from(bytes_col: BytesColumn) -> Self {
+        StrColumn(bytes_col)
     }
 }
 
-#[cfg(test)]
-mod tests {
-    use crate::{ColumnarReader, ColumnarWriter};
+impl StrColumn {
+    /// Fills the buffer
+    pub fn ord_to_str(&self, term_ord: u64, output: &mut String) -> io::Result<bool> {
+        unsafe {
+            let buf = output.as_mut_vec();
+            self.0.dictionary.ord_to_term(term_ord, buf)?;
+            // TODO consider remove checks if it hurts performance.
+            if std::str::from_utf8(buf.as_slice()).is_err() {
+                buf.clear();
+                return Err(io::Error::new(
+                    io::ErrorKind::InvalidData,
+                    "Not valid utf-8",
+                ));
+            }
+        }
+        Ok(true)
+    }
+}
+
+impl Deref for StrColumn {
+    type Target = BytesColumn;
+
+    fn deref(&self) -> &Self::Target {
+        &self.0
+    }
 }

columnar/src/column/mod.rs

@@ -5,8 +5,11 @@ use std::ops::Deref;
 use std::sync::Arc;
 
 use common::BinarySerializable;
-pub use dictionary_encoded::BytesColumn;
-pub use serialize::{open_column_bytes, open_column_u64, serialize_column_u64};
+pub use dictionary_encoded::{BytesColumn, StrColumn};
+pub use serialize::{
+    open_column_bytes, open_column_u128, open_column_u64, serialize_column_mappable_to_u128,
+    serialize_column_mappable_to_u64,
+};
 
 use crate::column_index::ColumnIndex;
 use crate::column_values::ColumnValues;
@@ -18,21 +21,43 @@ pub struct Column<T> {
     pub values: Arc<dyn ColumnValues<T>>,
 }
 
-use crate::column_index::Set;
-
 impl<T: PartialOrd> Column<T> {
-    pub fn first(&self, row_id: RowId) -> Option<T> {
+    pub fn num_rows(&self) -> RowId {
         match &self.idx {
-            ColumnIndex::Full => Some(self.values.get_val(row_id)),
-            ColumnIndex::Optional(opt_idx) => {
-                let value_row_idx = opt_idx.rank_if_exists(row_id)?;
-                Some(self.values.get_val(value_row_idx))
-            }
-            ColumnIndex::Multivalued(_multivalued_index) => {
-                todo!();
+            ColumnIndex::Full => self.values.num_vals() as u32,
+            ColumnIndex::Optional(optional_index) => optional_index.num_rows(),
+            ColumnIndex::Multivalued(col_index) => {
+                // The multivalued index contains all value start row_id,
+                // and one extra value at the end with the overall number of rows.
+                col_index.num_vals() - 1
             }
         }
     }
+
+    pub fn min_value(&self) -> T {
+        self.values.min_value()
+    }
+    pub fn max_value(&self) -> T {
+        self.values.max_value()
+    }
+}
+
+impl<T: PartialOrd + Copy + Send + Sync + 'static> Column<T> {
+    pub fn first(&self, row_id: RowId) -> Option<T> {
+        self.values(row_id).next()
+    }
+
+    pub fn values(&self, row_id: RowId) -> impl Iterator<Item = T> + '_ {
+        self.value_row_ids(row_id)
+            .map(|value_row_id: RowId| self.values.get_val(value_row_id))
+    }
+
+    pub fn first_or_default_col(self, default_value: T) -> Arc<dyn ColumnValues<T>> {
+        Arc::new(FirstValueWithDefault {
+            column: self,
+            default_value,
+        })
+    }
 }
 
 impl<T> Deref for Column<T> {
@@ -54,3 +79,31 @@ impl BinarySerializable for Cardinality {
         Ok(cardinality)
     }
 }
+
+// TODO simplify or optimize
+struct FirstValueWithDefault<T: Copy> {
+    column: Column<T>,
+    default_value: T,
+}
+
+impl<T: PartialOrd + Send + Sync + Copy + 'static> ColumnValues<T> for FirstValueWithDefault<T> {
+    fn get_val(&self, idx: u32) -> T {
+        self.column.first(idx).unwrap_or(self.default_value)
+    }
+
+    fn min_value(&self) -> T {
+        self.column.values.min_value()
+    }
+
+    fn max_value(&self) -> T {
+        self.column.values.max_value()
+    }
+
+    fn num_vals(&self) -> u32 {
+        match &self.column.idx {
+            ColumnIndex::Full => self.column.values.num_vals(),
+            ColumnIndex::Optional(optional_idx) => optional_idx.num_rows(),
+            ColumnIndex::Multivalued(_) => todo!(),
+        }
+    }
+}

columnar/src/column/serialize.rs

@@ -2,24 +2,51 @@ use std::io;
 use std::io::Write;
 use std::sync::Arc;
 
-use common::{CountingWriter, OwnedBytes};
+use common::OwnedBytes;
 use sstable::Dictionary;
 
 use crate::column::{BytesColumn, Column};
 use crate::column_index::{serialize_column_index, SerializableColumnIndex};
+use crate::column_values::serialize::serialize_column_values_u128;
 use crate::column_values::{
-    serialize_column_values, ColumnValues, MonotonicallyMappableToU64, ALL_CODEC_TYPES,
+    serialize_column_values, ColumnValues, FastFieldCodecType, MonotonicallyMappableToU128,
+    MonotonicallyMappableToU64,
 };
-pub fn serialize_column_u64<T: MonotonicallyMappableToU64>(
+
+pub fn serialize_column_mappable_to_u128<
+    F: Fn() -> I,
+    I: Iterator<Item = T>,
+    T: MonotonicallyMappableToU128,
+>(
+    column_index: SerializableColumnIndex<'_>,
+    column_values: F,
+    num_vals: u32,
+    output: &mut impl Write,
+) -> io::Result<()> {
+    let column_index_num_bytes = serialize_column_index(column_index, output)?;
+    serialize_column_values_u128(
+        || column_values().map(|val| val.to_u128()),
+        num_vals,
+        output,
+    )?;
+    output.write_all(&column_index_num_bytes.to_le_bytes())?;
+    Ok(())
+}
+
+pub fn serialize_column_mappable_to_u64<T: MonotonicallyMappableToU64>(
     column_index: SerializableColumnIndex<'_>,
     column_values: &impl ColumnValues<T>,
     output: &mut impl Write,
 ) -> io::Result<()> {
-    let mut counting_writer = CountingWriter::wrap(output);
-    serialize_column_index(column_index, &mut counting_writer)?;
-    let column_index_num_bytes = counting_writer.written_bytes() as u32;
-    let output = counting_writer.finish();
-    serialize_column_values(column_values, &ALL_CODEC_TYPES[..], output)?;
+    let column_index_num_bytes = serialize_column_index(column_index, output)?;
+    serialize_column_values(
+        column_values,
+        &[
+            FastFieldCodecType::Bitpacked,
+            FastFieldCodecType::BlockwiseLinear,
+        ],
+        output,
+    )?;
     output.write_all(&column_index_num_bytes.to_le_bytes())?;
     Ok(())
 }
@@ -41,14 +68,34 @@ pub fn open_column_u64<T: MonotonicallyMappableToU64>(bytes: OwnedBytes) -> io::
     })
 }
 
-pub fn open_column_bytes(data: OwnedBytes) -> io::Result<BytesColumn> {
+pub fn open_column_u128<T: MonotonicallyMappableToU128>(
+    bytes: OwnedBytes,
+) -> io::Result<Column<T>> {
+    let (body, column_index_num_bytes_payload) = bytes.rsplit(4);
+    let column_index_num_bytes = u32::from_le_bytes(
+        column_index_num_bytes_payload
+            .as_slice()
+            .try_into()
+            .unwrap(),
+    );
+    let (column_index_data, column_values_data) = body.split(column_index_num_bytes as usize);
+    let column_index = crate::column_index::open_column_index(column_index_data)?;
+    let column_values = crate::column_values::open_u128_mapped(column_values_data)?;
+    Ok(Column {
+        idx: column_index,
+        values: column_values,
+    })
+}
+
+pub fn open_column_bytes<T: From<BytesColumn>>(data: OwnedBytes) -> io::Result<T> {
     let (body, dictionary_len_bytes) = data.rsplit(4);
     let dictionary_len = u32::from_le_bytes(dictionary_len_bytes.as_slice().try_into().unwrap());
     let (dictionary_bytes, column_bytes) = body.split(dictionary_len as usize);
     let dictionary = Arc::new(Dictionary::from_bytes(dictionary_bytes)?);
     let term_ord_column = crate::column::open_column_u64::<u64>(column_bytes)?;
-    Ok(BytesColumn {
+    let bytes_column = BytesColumn {
         dictionary,
         term_ord_column,
-    })
+    };
+    Ok(bytes_column.into())
 }

columnar/src/column_index/mod.rs

@@ -2,6 +2,7 @@ mod multivalued_index;
 mod optional_index;
 mod serialize;
 
+use std::ops::Range;
 use std::sync::Arc;
 
 pub use optional_index::{OptionalIndex, SerializableOptionalIndex, Set};
@@ -14,8 +15,12 @@ use crate::{Cardinality, RowId};
 pub enum ColumnIndex<'a> {
     Full,
     Optional(OptionalIndex),
-    // TODO remove the Arc<dyn> apart from serialization this is not
-    // dynamic at all.
+    // TODO Remove the static by fixing the codec if possible.
+    /// The column values enclosed contains for all row_id,
+    /// the value start_index.
+    ///
+    /// In addition, at index num_rows, an extra value is added
+    /// containing the overal number of values.
     Multivalued(Arc<dyn ColumnValues<RowId> + 'a>),
 }
 
@@ -28,13 +33,22 @@ impl<'a> ColumnIndex<'a> {
         }
     }
 
-    pub fn num_rows(&self) -> RowId {
+    pub fn value_row_ids(&self, row_id: RowId) -> Range<RowId> {
         match self {
-            ColumnIndex::Full => {
-                todo!()
+            ColumnIndex::Full => row_id..row_id + 1,
+            ColumnIndex::Optional(optional_index) => {
+                if let Some(val) = optional_index.rank_if_exists(row_id) {
+                    val..val + 1
+                } else {
+                    0..0
+                }
+            }
+            ColumnIndex::Multivalued(multivalued_index) => {
+                let multivalued_index_ref = &**multivalued_index;
+                let start: u32 = multivalued_index_ref.get_val(row_id);
+                let end: u32 = multivalued_index_ref.get_val(row_id + 1);
+                start..end
             }
-            ColumnIndex::Optional(optional_index) => optional_index.num_rows(),
-            ColumnIndex::Multivalued(multivalued_index) => multivalued_index.num_vals() - 1,
         }
     }
 }

columnar/src/column_index/multivalued_index.rs

@@ -11,11 +11,11 @@ use crate::RowId;
 pub struct MultivaluedIndex(Arc<dyn ColumnValues<RowId>>);
 
 pub fn serialize_multivalued_index(
-    multivalued_index: MultivaluedIndex,
+    multivalued_index: &dyn ColumnValues<RowId>,
     output: &mut impl Write,
 ) -> io::Result<()> {
     crate::column_values::serialize_column_values(
-        &*multivalued_index.0,
+        &*multivalued_index,
         &[FastFieldCodecType::Bitpacked, FastFieldCodecType::Linear],
         output,
     )?;
@@ -23,5 +23,7 @@ pub fn serialize_multivalued_index(
 }
 
 pub fn open_multivalued_index(bytes: OwnedBytes) -> io::Result<Arc<dyn ColumnValues<RowId>>> {
-    todo!();
+    let start_index_column: Arc<dyn ColumnValues<RowId>> =
+        crate::column_values::open_u64_mapped(bytes)?;
+    Ok(start_index_column)
 }

columnar/src/column_index/optional_index/set_block/mod.rs

@@ -1,7 +1,7 @@
-mod set_block;
+mod dense;
 mod sparse;
 
-pub use set_block::{DenseBlock, DenseBlockCodec, DENSE_BLOCK_NUM_BYTES};
+pub use dense::{DenseBlock, DenseBlockCodec, DENSE_BLOCK_NUM_BYTES};
 pub use sparse::{SparseBlock, SparseBlockCodec};
 
 #[cfg(test)]

columnar/src/column_index/optional_index/set_block/tests.rs

@@ -1,7 +1,8 @@
 use std::collections::HashMap;
 
-use crate::column_index::optional_index::set_block::set_block::DENSE_BLOCK_NUM_BYTES;
-use crate::column_index::optional_index::set_block::{DenseBlockCodec, SparseBlockCodec};
+use crate::column_index::optional_index::set_block::{
+    DenseBlockCodec, SparseBlockCodec, DENSE_BLOCK_NUM_BYTES,
+};
 use crate::column_index::optional_index::{Set, SetCodec};
 
 fn test_set_helper<C: SetCodec<Item = u16>>(vals: &[u16]) -> usize {
@@ -51,6 +52,7 @@ fn test_sparse_block_set_u16_max() {
 use proptest::prelude::*;
 
 proptest! {
+    #![proptest_config(ProptestConfig::with_cases(1))]
     #[test]
     fn test_prop_test_dense(els in proptest::collection::btree_set(0..=u16::MAX, 0..=u16::MAX as usize)) {
         let vals: Vec<u16> = els.into_iter().collect();

columnar/src/column_index/serialize.rs

@@ -1,19 +1,20 @@
 use std::io;
 use std::io::Write;
 
-use common::OwnedBytes;
+use common::{CountingWriter, OwnedBytes};
 
-use crate::column_index::multivalued_index::{serialize_multivalued_index, MultivaluedIndex};
+use crate::column_index::multivalued_index::serialize_multivalued_index;
 use crate::column_index::optional_index::serialize_optional_index;
 use crate::column_index::{ColumnIndex, SerializableOptionalIndex};
-use crate::Cardinality;
+use crate::column_values::ColumnValues;
+use crate::{Cardinality, RowId};
 
 pub enum SerializableColumnIndex<'a> {
     Full,
     Optional(Box<dyn SerializableOptionalIndex<'a> + 'a>),
     // TODO remove the Arc<dyn> apart from serialization this is not
     // dynamic at all.
-    Multivalued(MultivaluedIndex),
+    Multivalued(Box<dyn ColumnValues<RowId> + 'a>),
 }
 
 impl<'a> SerializableColumnIndex<'a> {
@@ -29,19 +30,21 @@ impl<'a> SerializableColumnIndex<'a> {
 pub fn serialize_column_index(
     column_index: SerializableColumnIndex,
     output: &mut impl Write,
-) -> io::Result<()> {
+) -> io::Result<u32> {
+    let mut output = CountingWriter::wrap(output);
     let cardinality = column_index.get_cardinality().to_code();
     output.write_all(&[cardinality])?;
     match column_index {
         SerializableColumnIndex::Full => {}
         SerializableColumnIndex::Optional(optional_index) => {
-            serialize_optional_index(&*optional_index, output)?
+            serialize_optional_index(&*optional_index, &mut output)?
         }
         SerializableColumnIndex::Multivalued(multivalued_index) => {
-            serialize_multivalued_index(multivalued_index, output)?
+            serialize_multivalued_index(&*multivalued_index, &mut output)?
         }
     }
-    Ok(())
+    let column_index_num_bytes = output.written_bytes() as u32;
+    Ok(column_index_num_bytes)
 }
 
 pub fn open_column_index(mut bytes: OwnedBytes) -> io::Result<ColumnIndex<'static>> {

columnar/src/column_values/column.rs

@@ -78,6 +78,32 @@ pub trait ColumnValues<T: PartialOrd = u64>: Send + Sync {
     }
 }
 
+impl<T: Copy + PartialOrd> ColumnValues<T> for std::sync::Arc<dyn ColumnValues<T>> {
+    fn get_val(&self, idx: u32) -> T {
+        self.as_ref().get_val(idx)
+    }
+
+    fn min_value(&self) -> T {
+        self.as_ref().min_value()
+    }
+
+    fn max_value(&self) -> T {
+        self.as_ref().max_value()
+    }
+
+    fn num_vals(&self) -> u32 {
+        self.as_ref().num_vals()
+    }
+
+    fn iter<'b>(&'b self) -> Box<dyn Iterator<Item = T> + 'b> {
+        self.as_ref().iter()
+    }
+
+    fn get_range(&self, start: u64, output: &mut [T]) {
+        self.as_ref().get_range(start, output)
+    }
+}
+
 impl<'a, C: ColumnValues<T> + ?Sized, T: Copy + PartialOrd> ColumnValues<T> for &'a C {
     fn get_val(&self, idx: u32) -> T {
         (*self).get_val(idx)

columnar/src/column_values/mod.rs

@@ -28,7 +28,7 @@ mod compact_space;
 mod line;
 mod linear;
 pub(crate) mod monotonic_mapping;
-// mod monotonic_mapping_u128;
+pub(crate) mod monotonic_mapping_u128;
 
 mod column;
 mod column_with_cardinality;
@@ -37,8 +37,10 @@ pub mod serialize;
 
 pub use self::column::{monotonic_map_column, ColumnValues, IterColumn, VecColumn};
 pub use self::monotonic_mapping::{MonotonicallyMappableToU64, StrictlyMonotonicFn};
-// pub use self::monotonic_mapping_u128::MonotonicallyMappableToU128;
-pub use self::serialize::{serialize_and_load, serialize_column_values, NormalizedHeader};
+pub use self::monotonic_mapping_u128::MonotonicallyMappableToU128;
+#[cfg(test)]
+pub use self::serialize::tests::serialize_and_load;
+pub use self::serialize::{serialize_column_values, NormalizedHeader};
 use crate::column_values::bitpacked::BitpackedCodec;
 use crate::column_values::blockwise_linear::BlockwiseLinearCodec;
 use crate::column_values::linear::LinearCodec;
@@ -122,19 +124,17 @@ impl U128FastFieldCodecType {
 }
 
 /// Returns the correct codec reader wrapped in the `Arc` for the data.
-// pub fn open_u128<Item: MonotonicallyMappableToU128>(
-//     bytes: OwnedBytes,
-// ) -> io::Result<Arc<dyn Column<Item>>> {
-//     todo!();
-//     // let (bytes, _format_version) = read_format_version(bytes)?;
-//     // let (mut bytes, _null_index_footer) = read_null_index_footer(bytes)?;
-//     // let header = U128Header::deserialize(&mut bytes)?;
-//     // assert_eq!(header.codec_type, U128FastFieldCodecType::CompactSpace);
-//     // let reader = CompactSpaceDecompressor::open(bytes)?;
-//     // let inverted: StrictlyMonotonicMappingInverter<StrictlyMonotonicMappingToInternal<Item>> =
-//     //     StrictlyMonotonicMappingToInternal::<Item>::new().into();
-//     // Ok(Arc::new(monotonic_map_column(reader, inverted)))
-// }
+pub fn open_u128_mapped<T: MonotonicallyMappableToU128>(
+    mut bytes: OwnedBytes,
+) -> io::Result<Arc<dyn ColumnValues<T>>> {
+    let header = U128Header::deserialize(&mut bytes)?;
+    assert_eq!(header.codec_type, U128FastFieldCodecType::CompactSpace);
+    let reader = CompactSpaceDecompressor::open(bytes)?;
+
+    let inverted: StrictlyMonotonicMappingInverter<StrictlyMonotonicMappingToInternal<T>> =
+        StrictlyMonotonicMappingToInternal::<T>::new().into();
+    Ok(Arc::new(monotonic_map_column(reader, inverted)))
+}
 
 /// Returns the correct codec reader wrapped in the `Arc` for the data.
 pub fn open_u64_mapped<T: MonotonicallyMappableToU64>(
@@ -198,13 +198,6 @@ pub(crate) trait FastFieldCodec: 'static {
     fn estimate(column: &dyn ColumnValues) -> Option<f32>;
 }
 
-/// The list of all available codecs for u64 convertible data.
-pub const ALL_CODEC_TYPES: [FastFieldCodecType; 3] = [
-    FastFieldCodecType::Bitpacked,
-    FastFieldCodecType::BlockwiseLinear,
-    FastFieldCodecType::Linear,
-];
-
 #[cfg(all(test, feature = "unstable"))]
 mod bench {
     use std::sync::Arc;

columnar/src/column_values/monotonic_mapping.rs

@@ -2,6 +2,7 @@ use std::marker::PhantomData;
 
 use fastdivide::DividerU64;
 
+use super::MonotonicallyMappableToU128;
 use crate::RowId;
 
 /// Monotonic maps a value to u64 value space.
@@ -80,21 +81,20 @@ impl<T> StrictlyMonotonicMappingToInternal<T> {
     }
 }
 
-// TODO
-// impl<External: MonotonicallyMappableToU128, T: MonotonicallyMappableToU128>
-//     StrictlyMonotonicFn<External, u128> for StrictlyMonotonicMappingToInternal<T>
-// where T: MonotonicallyMappableToU128
-// {
-//     #[inline(always)]
-//     fn mapping(&self, inp: External) -> u128 {
-//         External::to_u128(inp)
-//     }
+impl<External: MonotonicallyMappableToU128, T: MonotonicallyMappableToU128>
+    StrictlyMonotonicFn<External, u128> for StrictlyMonotonicMappingToInternal<T>
+where T: MonotonicallyMappableToU128
+{
+    #[inline(always)]
+    fn mapping(&self, inp: External) -> u128 {
+        External::to_u128(inp)
+    }
 
-//     #[inline(always)]
-//     fn inverse(&self, out: u128) -> External {
-//         External::from_u128(out)
-//     }
-// }
+    #[inline(always)]
+    fn inverse(&self, out: u128) -> External {
+        External::from_u128(out)
+    }
+}
 
 impl<External: MonotonicallyMappableToU64, T: MonotonicallyMappableToU64>
     StrictlyMonotonicFn<External, u64> for StrictlyMonotonicMappingToInternal<T>
@@ -194,6 +194,20 @@ impl MonotonicallyMappableToU64 for i64 {
     }
 }
 
+impl MonotonicallyMappableToU64 for crate::DateTime {
+    #[inline(always)]
+    fn to_u64(self) -> u64 {
+        common::i64_to_u64(self.timestamp_micros)
+    }
+
+    #[inline(always)]
+    fn from_u64(val: u64) -> Self {
+        crate::DateTime {
+            timestamp_micros: common::u64_to_i64(val),
+        }
+    }
+}
+
 impl MonotonicallyMappableToU64 for bool {
     #[inline(always)]
     fn to_u64(self) -> u64 {

columnar/src/column_values/serialize.rs

@@ -19,9 +19,8 @@
 
 use std::io;
 use std::num::NonZeroU64;
-use std::sync::Arc;
 
-use common::{BinarySerializable, OwnedBytes, VInt};
+use common::{BinarySerializable, VInt};
 use log::warn;
 
 use super::bitpacked::BitpackedCodec;
@@ -33,8 +32,9 @@ use super::monotonic_mapping::{
 };
 use super::{
     monotonic_map_column, ColumnValues, FastFieldCodec, FastFieldCodecType,
-    MonotonicallyMappableToU64, U128FastFieldCodecType, VecColumn, ALL_CODEC_TYPES,
+    MonotonicallyMappableToU64, U128FastFieldCodecType,
 };
+use crate::column_values::compact_space::CompactSpaceCompressor;
 
 /// The normalized header gives some parameters after applying the following
 /// normalization of the vector:
@@ -160,54 +160,22 @@ impl BinarySerializable for Header {
     }
 }
 
-/// Return estimated compression for given codec in the value range [0.0..1.0], where 1.0 means no
-/// compression.
-pub(crate) fn estimate<T: MonotonicallyMappableToU64>(
-    typed_column: impl ColumnValues<T>,
-    codec_type: FastFieldCodecType,
-) -> Option<f32> {
-    let column = monotonic_map_column(typed_column, StrictlyMonotonicMappingToInternal::<T>::new());
-    let min_value = column.min_value();
-    let gcd = super::gcd::find_gcd(column.iter().map(|val| val - min_value))
-        .filter(|gcd| gcd.get() > 1u64);
-    let mapping = StrictlyMonotonicMappingToInternalGCDBaseval::new(
-        gcd.map(|gcd| gcd.get()).unwrap_or(1u64),
-        min_value,
-    );
-    let normalized_column = monotonic_map_column(&column, mapping);
-    match codec_type {
-        FastFieldCodecType::Bitpacked => BitpackedCodec::estimate(&normalized_column),
-        FastFieldCodecType::Linear => LinearCodec::estimate(&normalized_column),
-        FastFieldCodecType::BlockwiseLinear => BlockwiseLinearCodec::estimate(&normalized_column),
-    }
-}
-
-// TODO
 /// Serializes u128 values with the compact space codec.
-// pub fn serialize_u128_new<F: Fn() -> I, I: Iterator<Item = u128>>(
-//     value_index: ColumnIndex,
-//     iter_gen: F,
-//     num_vals: u32,
-//     output: &mut impl io::Write,
-// ) -> io::Result<()> {
-//     let header = U128Header {
-//         num_vals,
-//         codec_type: U128FastFieldCodecType::CompactSpace,
-//     };
-//     header.serialize(output)?;
-//     let compressor = CompactSpaceCompressor::train_from(iter_gen(), num_vals);
-//     compressor.compress_into(iter_gen(), output).unwrap();
+pub fn serialize_column_values_u128<F: Fn() -> I, I: Iterator<Item = u128>>(
+    iter_gen: F,
+    num_vals: u32,
+    output: &mut impl io::Write,
+) -> io::Result<()> {
+    let header = U128Header {
+        num_vals,
+        codec_type: U128FastFieldCodecType::CompactSpace,
+    };
+    header.serialize(output)?;
+    let compressor = CompactSpaceCompressor::train_from(iter_gen(), num_vals);
+    compressor.compress_into(iter_gen(), output)?;
 
-//     let null_index_footer = ColumnFooter {
-//         cardinality: value_index.get_cardinality(),
-//         null_index_codec: NullIndexCodec::Full,
-//         null_index_byte_range: 0..0,
-//     };
-//     append_null_index_footer(output, null_index_footer)?;
-//     append_format_version(output)?;
-
-//     Ok(())
-// }
+    Ok(())
+}
 
 /// Serializes the column with the codec with the best estimate on the data.
 pub fn serialize_column_values<T: MonotonicallyMappableToU64>(
@@ -279,20 +247,29 @@ pub(crate) fn serialize_given_codec(
     Ok(())
 }
 
-/// Helper function to serialize a column (autodetect from all codecs) and then open it
-pub fn serialize_and_load<T: MonotonicallyMappableToU64 + Ord + Default>(
-    column: &[T],
-) -> Arc<dyn ColumnValues<T>> {
-    let mut buffer = Vec::new();
-    super::serialize_column_values(&VecColumn::from(&column), &ALL_CODEC_TYPES, &mut buffer)
-        .unwrap();
-    super::open_u64_mapped(OwnedBytes::new(buffer)).unwrap()
-}
-
 #[cfg(test)]
-mod tests {
-    use super::*;
+pub mod tests {
+    use std::sync::Arc;
 
+    use common::OwnedBytes;
+
+    use super::*;
+    use crate::column_values::{open_u64_mapped, VecColumn};
+
+    const ALL_CODEC_TYPES: [FastFieldCodecType; 3] = [
+        FastFieldCodecType::Bitpacked,
+        FastFieldCodecType::Linear,
+        FastFieldCodecType::BlockwiseLinear,
+    ];
+
+    /// Helper function to serialize a column (autodetect from all codecs) and then open it
+    pub fn serialize_and_load<T: MonotonicallyMappableToU64 + Ord + Default>(
+        column: &[T],
+    ) -> Arc<dyn ColumnValues<T>> {
+        let mut buffer = Vec::new();
+        serialize_column_values(&VecColumn::from(&column), &ALL_CODEC_TYPES, &mut buffer).unwrap();
+        open_u64_mapped(OwnedBytes::new(buffer)).unwrap()
+    }
     #[test]
     fn test_serialize_deserialize_u128_header() {
         let original = U128Header {
@@ -319,7 +296,7 @@ mod tests {
         serialize_column_values(&col, &ALL_CODEC_TYPES, &mut buffer).unwrap();
         // TODO put the header as a footer so that it serves as a padding.
         // 5 bytes of header, 1 byte of value, 7 bytes of padding.
-        assert_eq!(buffer.len(), 5 + 1 + 7);
+        assert_eq!(buffer.len(), 5 + 1);
     }
 
     #[test]
@@ -328,7 +305,7 @@ mod tests {
         let col = VecColumn::from(&[true][..]);
         serialize_column_values(&col, &ALL_CODEC_TYPES, &mut buffer).unwrap();
         // 5 bytes of header, 0 bytes of value, 7 bytes of padding.
-        assert_eq!(buffer.len(), 5 + 7);
+        assert_eq!(buffer.len(), 5);
     }
 
     #[test]
@@ -338,6 +315,6 @@ mod tests {
         let col = VecColumn::from(&vals[..]);
         serialize_column_values(&col, &[FastFieldCodecType::Bitpacked], &mut buffer).unwrap();
         // Values are stored over 3 bits.
-        assert_eq!(buffer.len(), 7 + (3 * 80 / 8) + 7);
+        assert_eq!(buffer.len(), 7 + (3 * 80 / 8));
     }
 }

columnar/src/columnar/column_type.rs

@@ -1,115 +1,149 @@
-use crate::utils::{place_bits, select_bits};
+use std::net::Ipv6Addr;
+
 use crate::value::NumericalType;
 use crate::InvalidData;
 
-/// The column type represents the column type and can fit on 6-bits.
-///
-/// - bits[0..3]: Column category type.
-/// - bits[3..6]: Numerical type if necessary.
-#[derive(Hash, Eq, PartialEq, Debug, Clone, Copy)]
+/// The column type represents the column type.
+/// Any changes need to be propagated to `COLUMN_TYPES`.
+#[derive(Hash, Eq, PartialEq, Debug, Clone, Copy, Ord, PartialOrd)]
+#[repr(u8)]
 pub enum ColumnType {
-    Bytes,
-    Numerical(NumericalType),
-    Bool,
+    I64 = 0u8,
+    U64 = 1u8,
+    F64 = 2u8,
+    Bytes = 3u8,
+    Str = 4u8,
+    Bool = 5u8,
+    IpAddr = 6u8,
+    DateTime = 7u8,
 }
 
+// The order needs to match _exactly_ the order in the enum
+const COLUMN_TYPES: [ColumnType; 8] = [
+    ColumnType::I64,
+    ColumnType::U64,
+    ColumnType::F64,
+    ColumnType::Bytes,
+    ColumnType::Str,
+    ColumnType::Bool,
+    ColumnType::IpAddr,
+    ColumnType::DateTime,
+];
+
 impl ColumnType {
-    /// Encoded over 6 bits.
-    pub(crate) fn to_code(self) -> u8 {
-        let column_type_category;
-        let numerical_type_code: u8;
-        match self {
-            ColumnType::Bytes => {
-                column_type_category = ColumnTypeCategory::Str;
-                numerical_type_code = 0u8;
-            }
-            ColumnType::Numerical(numerical_type) => {
-                column_type_category = ColumnTypeCategory::Numerical;
-                numerical_type_code = numerical_type.to_code();
-            }
-            ColumnType::Bool => {
-                column_type_category = ColumnTypeCategory::Bool;
-                numerical_type_code = 0u8;
-            }
-        }
-        place_bits::<0, 3>(column_type_category.to_code()) | place_bits::<3, 6>(numerical_type_code)
-    }
-
-    pub(crate) fn try_from_code(code: u8) -> Result<ColumnType, InvalidData> {
-        if select_bits::<6, 8>(code) != 0u8 {
-            return Err(InvalidData);
-        }
-        let column_type_category_code = select_bits::<0, 3>(code);
-        let numerical_type_code = select_bits::<3, 6>(code);
-        let column_type_category = ColumnTypeCategory::try_from_code(column_type_category_code)?;
-        match column_type_category {
-            ColumnTypeCategory::Bool => {
-                if numerical_type_code != 0u8 {
-                    return Err(InvalidData);
-                }
-                Ok(ColumnType::Bool)
-            }
-            ColumnTypeCategory::Str => {
-                if numerical_type_code != 0u8 {
-                    return Err(InvalidData);
-                }
-                Ok(ColumnType::Bytes)
-            }
-            ColumnTypeCategory::Numerical => {
-                let numerical_type = NumericalType::try_from_code(numerical_type_code)?;
-                Ok(ColumnType::Numerical(numerical_type))
-            }
-        }
-    }
-}
-
-/// Column types are grouped into different categories that
-/// corresponds to the different types of `JsonValue` types.
-///
-/// The columnar writer will apply coercion rules to make sure that
-/// at most one column exist per `ColumnTypeCategory`.
-///
-/// See also [README.md].
-#[derive(Copy, Clone, Ord, PartialOrd, Eq, PartialEq, Debug)]
-#[repr(u8)]
-pub(crate) enum ColumnTypeCategory {
-    Bool = 0u8,
-    Str = 1u8,
-    Numerical = 2u8,
-}
-
-impl ColumnTypeCategory {
     pub fn to_code(self) -> u8 {
         self as u8
     }
 
-    pub fn try_from_code(code: u8) -> Result<Self, InvalidData> {
-        match code {
-            0u8 => Ok(Self::Bool),
-            1u8 => Ok(Self::Str),
-            2u8 => Ok(Self::Numerical),
-            _ => Err(InvalidData),
+    pub(crate) fn try_from_code(code: u8) -> Result<ColumnType, InvalidData> {
+        COLUMN_TYPES.get(code as usize).copied().ok_or(InvalidData)
+    }
+}
+
+impl From<NumericalType> for ColumnType {
+    fn from(numerical_type: NumericalType) -> Self {
+        match numerical_type {
+            NumericalType::I64 => ColumnType::I64,
+            NumericalType::U64 => ColumnType::U64,
+            NumericalType::F64 => ColumnType::F64,
         }
     }
 }
 
+impl ColumnType {
+    pub fn numerical_type(&self) -> Option<NumericalType> {
+        match self {
+            ColumnType::I64 => Some(NumericalType::I64),
+            ColumnType::U64 => Some(NumericalType::U64),
+            ColumnType::F64 => Some(NumericalType::F64),
+            ColumnType::Bytes
+            | ColumnType::Str
+            | ColumnType::Bool
+            | ColumnType::IpAddr
+            | ColumnType::DateTime => None,
+        }
+    }
+}
+
+// TODO remove if possible
+pub trait HasAssociatedColumnType: 'static + Send + Sync + Copy + PartialOrd {
+    fn column_type() -> ColumnType;
+    fn default_value() -> Self;
+}
+
+impl HasAssociatedColumnType for u64 {
+    fn column_type() -> ColumnType {
+        ColumnType::U64
+    }
+
+    fn default_value() -> Self {
+        0u64
+    }
+}
+
+impl HasAssociatedColumnType for i64 {
+    fn column_type() -> ColumnType {
+        ColumnType::I64
+    }
+
+    fn default_value() -> Self {
+        0i64
+    }
+}
+
+impl HasAssociatedColumnType for f64 {
+    fn column_type() -> ColumnType {
+        ColumnType::F64
+    }
+
+    fn default_value() -> Self {
+        Default::default()
+    }
+}
+
+impl HasAssociatedColumnType for bool {
+    fn column_type() -> ColumnType {
+        ColumnType::Bool
+    }
+    fn default_value() -> Self {
+        Default::default()
+    }
+}
+
+impl HasAssociatedColumnType for crate::DateTime {
+    fn column_type() -> ColumnType {
+        ColumnType::DateTime
+    }
+    fn default_value() -> Self {
+        Default::default()
+    }
+}
+
+impl HasAssociatedColumnType for Ipv6Addr {
+    fn column_type() -> ColumnType {
+        ColumnType::IpAddr
+    }
+
+    fn default_value() -> Self {
+        Ipv6Addr::from([0u8; 16])
+    }
+}
+
 #[cfg(test)]
 mod tests {
-    use std::collections::HashSet;
-
     use super::*;
     use crate::Cardinality;
 
     #[test]
     fn test_column_type_to_code() {
-        let mut column_type_set: HashSet<ColumnType> = HashSet::new();
-        for code in u8::MIN..=u8::MAX {
-            if let Ok(column_type) = ColumnType::try_from_code(code) {
-                assert_eq!(column_type.to_code(), code);
-                assert!(column_type_set.insert(column_type));
+        for (code, expected_column_type) in super::COLUMN_TYPES.iter().copied().enumerate() {
+            if let Ok(column_type) = ColumnType::try_from_code(code as u8) {
+                assert_eq!(column_type, expected_column_type);
             }
         }
-        assert_eq!(column_type_set.len(), 2 + 3);
+        for code in COLUMN_TYPES.len() as u8..=u8::MAX {
+            assert!(ColumnType::try_from_code(code as u8).is_err());
+        }
     }
 
     #[test]

columnar/src/columnar/merge.rs

@@ -0,0 +1,208 @@
+use std::collections::HashMap;
+use std::io;
+
+use crate::columnar::ColumnarReader;
+use crate::dynamic_column::DynamicColumn;
+use crate::ColumnType;
+
+pub enum MergeDocOrder {
+    /// Columnar tables are simply stacked one above the other.
+    /// If the i-th columnar_readers has n_rows_i rows, then
+    /// in the resulting columnar,
+    /// rows [r0..n_row_0) contains the row of columnar_readers[0], in ordder
+    /// rows [n_row_0..n_row_0 + n_row_1 contains the row of columnar_readers[1], in order.
+    /// ..
+    Stack,
+    /// Some more complex mapping, that can interleaves rows from the different readers and
+    /// possibly drop rows.
+    Complex(()),
+}
+
+pub fn merge_columnar(
+    _columnar_readers: &[ColumnarReader],
+    mapping: MergeDocOrder,
+    _output: &mut impl io::Write,
+) -> io::Result<()> {
+    match mapping {
+        MergeDocOrder::Stack => {
+            // implement me :)
+            todo!();
+        }
+        MergeDocOrder::Complex(_) => {
+            // for later
+            todo!();
+        }
+    }
+}
+
+/// Column types are grouped into different categories.
+/// After merge, all columns belonging to the same category are coerced to
+/// the same column type.
+///
+/// In practise, today, only Numerical colummns are coerced into one type today.
+///
+/// See also [README.md].
+#[derive(Copy, Clone, Eq, PartialEq, Hash, Debug)]
+#[repr(u8)]
+enum ColumnTypeCategory {
+    Bool,
+    Str,
+    Numerical,
+    DateTime,
+    Bytes,
+    IpAddr,
+}
+
+impl From<ColumnType> for ColumnTypeCategory {
+    fn from(column_type: ColumnType) -> Self {
+        match column_type {
+            ColumnType::I64 => ColumnTypeCategory::Numerical,
+            ColumnType::U64 => ColumnTypeCategory::Numerical,
+            ColumnType::F64 => ColumnTypeCategory::Numerical,
+            ColumnType::Bytes => ColumnTypeCategory::Bytes,
+            ColumnType::Str => ColumnTypeCategory::Str,
+            ColumnType::Bool => ColumnTypeCategory::Bool,
+            ColumnType::IpAddr => ColumnTypeCategory::IpAddr,
+            ColumnType::DateTime => ColumnTypeCategory::DateTime,
+        }
+    }
+}
+
+fn collect_columns(
+    columnar_readers: &[&ColumnarReader],
+) -> io::Result<HashMap<String, HashMap<ColumnTypeCategory, Vec<DynamicColumn>>>> {
+    // Each column name may have multiple types of column associated.
+    // For merging we are interested in the same column type category since they can be merged.
+    let mut field_name_to_group: HashMap<String, HashMap<ColumnTypeCategory, Vec<DynamicColumn>>> =
+        HashMap::new();
+
+    for columnar_reader in columnar_readers {
+        let column_name_and_handle = columnar_reader.list_columns()?;
+        for (column_name, handle) in column_name_and_handle {
+            let column_type_to_handles = field_name_to_group
+                .entry(column_name.to_string())
+                .or_default();
+
+            let columns = column_type_to_handles
+                .entry(handle.column_type().into())
+                .or_default();
+            columns.push(handle.open()?);
+        }
+    }
+
+    normalize_columns(&mut field_name_to_group);
+
+    Ok(field_name_to_group)
+}
+
+/// Coerce numerical type columns to the same type
+/// TODO rename to `coerce_columns`
+fn normalize_columns(map: &mut HashMap<String, HashMap<ColumnTypeCategory, Vec<DynamicColumn>>>) {
+    for (_field_name, type_category_to_columns) in map.iter_mut() {
+        for (type_category, columns) in type_category_to_columns {
+            if type_category == &ColumnTypeCategory::Numerical {
+                let casted_columns = cast_to_common_numerical_column(&columns);
+                *columns = casted_columns;
+            }
+        }
+    }
+}
+
+/// Receives a list of columns of numerical types (u64, i64, f64)
+///
+/// Returns a list of `DynamicColumn` which are all of the same numerical type
+fn cast_to_common_numerical_column(columns: &[DynamicColumn]) -> Vec<DynamicColumn> {
+    assert!(columns
+        .iter()
+        .all(|column| column.column_type().numerical_type().is_some()));
+    let coerce_to_i64: Vec<_> = columns
+        .iter()
+        .map(|column| column.clone().coerce_to_i64())
+        .collect();
+
+    if coerce_to_i64.iter().all(|column| column.is_some()) {
+        return coerce_to_i64
+            .into_iter()
+            .map(|column| column.unwrap())
+            .collect();
+    }
+
+    let coerce_to_u64: Vec<_> = columns
+        .iter()
+        .map(|column| column.clone().coerce_to_u64())
+        .collect();
+
+    if coerce_to_u64.iter().all(|column| column.is_some()) {
+        return coerce_to_u64
+            .into_iter()
+            .map(|column| column.unwrap())
+            .collect();
+    }
+
+    columns
+        .iter()
+        .map(|column| {
+            column
+                .clone()
+                .coerce_to_f64()
+                .expect("couldn't cast column to f64")
+        })
+        .collect()
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use crate::ColumnarWriter;
+
+    #[test]
+    fn test_column_coercion() {
+        // i64 type
+        let columnar1 = {
+            let mut dataframe_writer = ColumnarWriter::default();
+            dataframe_writer.record_numerical(1u32, "numbers", 1i64);
+            let mut buffer: Vec<u8> = Vec::new();
+            dataframe_writer.serialize(2, &mut buffer).unwrap();
+            ColumnarReader::open(buffer).unwrap()
+        };
+        // u64 type
+        let columnar2 = {
+            let mut dataframe_writer = ColumnarWriter::default();
+            dataframe_writer.record_numerical(1u32, "numbers", u64::MAX - 100);
+            let mut buffer: Vec<u8> = Vec::new();
+            dataframe_writer.serialize(2, &mut buffer).unwrap();
+            ColumnarReader::open(buffer).unwrap()
+        };
+
+        // f64 type
+        let columnar3 = {
+            let mut dataframe_writer = ColumnarWriter::default();
+            dataframe_writer.record_numerical(1u32, "numbers", 30.5);
+            let mut buffer: Vec<u8> = Vec::new();
+            dataframe_writer.serialize(2, &mut buffer).unwrap();
+            ColumnarReader::open(buffer).unwrap()
+        };
+
+        let column_map = collect_columns(&[&columnar1, &columnar2, &columnar3]).unwrap();
+        assert_eq!(column_map.len(), 1);
+        let cat_to_columns = column_map.get("numbers").unwrap();
+        assert_eq!(cat_to_columns.len(), 1);
+
+        let numerical = cat_to_columns.get(&ColumnTypeCategory::Numerical).unwrap();
+        assert!(numerical.iter().all(|column| column.is_f64()));
+
+        let column_map = collect_columns(&[&columnar1, &columnar1]).unwrap();
+        assert_eq!(column_map.len(), 1);
+        let cat_to_columns = column_map.get("numbers").unwrap();
+        assert_eq!(cat_to_columns.len(), 1);
+        let numerical = cat_to_columns.get(&ColumnTypeCategory::Numerical).unwrap();
+        assert!(numerical.iter().all(|column| column.is_i64()));
+
+        let column_map = collect_columns(&[&columnar2, &columnar2]).unwrap();
+        assert_eq!(column_map.len(), 1);
+        let cat_to_columns = column_map.get("numbers").unwrap();
+        assert_eq!(cat_to_columns.len(), 1);
+        let numerical = cat_to_columns.get(&ColumnTypeCategory::Numerical).unwrap();
+        assert!(numerical.iter().all(|column| column.is_u64()));
+    }
+}

columnar/src/columnar/mod.rs

@@ -1,28 +1,10 @@
-// Copyright (C) 2022 Quickwit, Inc.
-//
-// Quickwit is offered under the AGPL v3.0 and as commercial software.
-// For commercial licensing, contact us at hello@quickwit.io.
-//
-// AGPL:
-// This program is free software: you can redistribute it and/or modify
-// it under the terms of the GNU Affero General Public License as
-// published by the Free Software Foundation, either version 3 of the
-// License, or (at your option) any later version.
-//
-// This program is distributed in the hope that it will be useful,
-// but WITHOUT ANY WARRANTY; without even the implied warranty of
-// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
-// GNU Affero General Public License for more details.
-//
-// You should have received a copy of the GNU Affero General Public License
-// along with this program. If not, see <http://www.gnu.org/licenses/>.
-//
-
 mod column_type;
 mod format_version;
+mod merge;
 mod reader;
 mod writer;
 
-pub use column_type::ColumnType;
+pub use column_type::{ColumnType, HasAssociatedColumnType};
+pub use merge::{merge_columnar, MergeDocOrder};
 pub use reader::ColumnarReader;
 pub use writer::ColumnarWriter;

columnar/src/columnar/reader/mod.rs

@@ -44,7 +44,7 @@ impl ColumnarReader {
         })
     }
 
-    // TODO fix ugly API
+    // TODO Add unit tests
     pub fn list_columns(&self) -> io::Result<Vec<(String, DynamicColumnHandle)>> {
         let mut stream = self.column_dictionary.stream()?;
         let mut results = Vec::new();
@@ -55,7 +55,8 @@ impl ColumnarReader {
                 .map_err(|_| io_invalid_data(format!("Unknown column code `{column_code}`")))?;
             let range = stream.value().clone();
             let column_name =
-                String::from_utf8_lossy(&key_bytes[..key_bytes.len() - 1]).to_string();
+                // The last two bytes are respectively the 0u8 separator and the column_type.
+                String::from_utf8_lossy(&key_bytes[..key_bytes.len() - 2]).to_string();
             let file_slice = self
                 .column_data
                 .slice(range.start as usize..range.end as usize);

columnar/src/columnar/writer/column_operation.rs

@@ -1,3 +1,5 @@
+use std::net::Ipv6Addr;
+
 use crate::dictionary::UnorderedId;
 use crate::utils::{place_bits, pop_first_byte, select_bits};
 use crate::value::NumericalValue;
@@ -25,12 +27,12 @@ struct ColumnOperationMetadata {
 
 impl ColumnOperationMetadata {
     fn to_code(self) -> u8 {
-        place_bits::<0, 4>(self.len) | place_bits::<4, 8>(self.op_type.to_code())
+        place_bits::<0, 6>(self.len) | place_bits::<6, 8>(self.op_type.to_code())
     }
 
     fn try_from_code(code: u8) -> Result<Self, InvalidData> {
-        let len = select_bits::<0, 4>(code);
-        let typ_code = select_bits::<4, 8>(code);
+        let len = select_bits::<0, 6>(code);
+        let typ_code = select_bits::<6, 8>(code);
         let column_type = ColumnOperationType::try_from_code(typ_code)?;
         Ok(ColumnOperationMetadata {
             op_type: column_type,
@@ -142,9 +144,21 @@ impl SymbolValue for bool {
     }
 }
 
+impl SymbolValue for Ipv6Addr {
+    fn serialize(self, buffer: &mut [u8]) -> u8 {
+        buffer[0..16].copy_from_slice(&self.octets());
+        16
+    }
+
+    fn deserialize(bytes: &[u8]) -> Self {
+        let octets: [u8; 16] = bytes[0..16].try_into().unwrap();
+        Ipv6Addr::from(octets)
+    }
+}
+
 #[derive(Default)]
 struct MiniBuffer {
-    pub bytes: [u8; 10],
+    pub bytes: [u8; 17],
     pub len: u8,
 }
 

columnar/src/columnar/writer/column_writers.rs

@@ -102,18 +102,29 @@ pub(crate) struct NumericalColumnWriter {
     column_writer: ColumnWriter,
 }
 
+impl NumericalColumnWriter {
+    pub fn force_numerical_type(&mut self, numerical_type: NumericalType) {
+        assert!(self
+            .compatible_numerical_types
+            .is_type_accepted(numerical_type));
+        self.compatible_numerical_types = CompatibleNumericalTypes::StaticType(numerical_type);
+    }
+}
+
 /// State used to store what types are still acceptable
 /// after having seen a set of numerical values.
 #[derive(Clone, Copy)]
-struct CompatibleNumericalTypes {
-    all_values_within_i64_range: bool,
-    all_values_within_u64_range: bool,
-    // f64 is always acceptable.
+enum CompatibleNumericalTypes {
+    Dynamic {
+        all_values_within_i64_range: bool,
+        all_values_within_u64_range: bool,
+    },
+    StaticType(NumericalType),
 }
 
 impl Default for CompatibleNumericalTypes {
     fn default() -> CompatibleNumericalTypes {
-        CompatibleNumericalTypes {
+        CompatibleNumericalTypes::Dynamic {
             all_values_within_i64_range: true,
             all_values_within_u64_range: true,
         }
@@ -121,39 +132,64 @@ impl Default for CompatibleNumericalTypes {
 }
 
 impl CompatibleNumericalTypes {
+    fn is_type_accepted(&self, numerical_type: NumericalType) -> bool {
+        match self {
+            CompatibleNumericalTypes::Dynamic {
+                all_values_within_i64_range,
+                all_values_within_u64_range,
+            } => match numerical_type {
+                NumericalType::I64 => *all_values_within_i64_range,
+                NumericalType::U64 => *all_values_within_u64_range,
+                NumericalType::F64 => true,
+            },
+            CompatibleNumericalTypes::StaticType(static_numerical_type) => {
+                *static_numerical_type == numerical_type
+            }
+        }
+    }
+
     fn accept_value(&mut self, numerical_value: NumericalValue) {
-        match numerical_value {
-            NumericalValue::I64(val_i64) => {
-                let value_within_u64_range = val_i64 >= 0i64;
-                self.all_values_within_u64_range &= value_within_u64_range;
-            }
-            NumericalValue::U64(val_u64) => {
-                let value_within_i64_range = val_u64 < i64::MAX as u64;
-                self.all_values_within_i64_range &= value_within_i64_range;
-            }
-            NumericalValue::F64(_) => {
-                self.all_values_within_i64_range = false;
-                self.all_values_within_u64_range = false;
+        match self {
+            CompatibleNumericalTypes::Dynamic {
+                all_values_within_i64_range,
+                all_values_within_u64_range,
+            } => match numerical_value {
+                NumericalValue::I64(val_i64) => {
+                    let value_within_u64_range = val_i64 >= 0i64;
+                    *all_values_within_u64_range &= value_within_u64_range;
+                }
+                NumericalValue::U64(val_u64) => {
+                    let value_within_i64_range = val_u64 < i64::MAX as u64;
+                    *all_values_within_i64_range &= value_within_i64_range;
+                }
+                NumericalValue::F64(_) => {
+                    *all_values_within_i64_range = false;
+                    *all_values_within_u64_range = false;
+                }
+            },
+            CompatibleNumericalTypes::StaticType(typ) => {
+                assert_eq!(numerical_value.numerical_type(), *typ);
             }
         }
     }
 
     pub fn to_numerical_type(self) -> NumericalType {
-        if self.all_values_within_i64_range {
-            NumericalType::I64
-        } else if self.all_values_within_u64_range {
-            NumericalType::U64
-        } else {
-            NumericalType::F64
+        for numerical_type in [NumericalType::I64, NumericalType::U64] {
+            if self.is_type_accepted(numerical_type) {
+                return numerical_type;
+            }
         }
+        NumericalType::F64
     }
 }
 
 impl NumericalColumnWriter {
-    pub fn column_type_and_cardinality(&self, num_docs: RowId) -> (NumericalType, Cardinality) {
-        let numerical_type = self.compatible_numerical_types.to_numerical_type();
-        let cardinality = self.column_writer.get_cardinality(num_docs);
-        (numerical_type, cardinality)
+    pub fn numerical_type(&self) -> NumericalType {
+        self.compatible_numerical_types.to_numerical_type()
+    }
+
+    pub fn cardinality(&self, num_docs: RowId) -> Cardinality {
+        self.column_writer.get_cardinality(num_docs)
     }
 
     pub fn record_numerical_value(
@@ -175,15 +211,15 @@ impl NumericalColumnWriter {
     }
 }
 
-#[derive(Copy, Clone, Default)]
-pub(crate) struct StrColumnWriter {
+#[derive(Copy, Clone)]
+pub(crate) struct StrOrBytesColumnWriter {
     pub(crate) dictionary_id: u32,
     pub(crate) column_writer: ColumnWriter,
 }
 
-impl StrColumnWriter {
-    pub(crate) fn with_dictionary_id(dictionary_id: u32) -> StrColumnWriter {
-        StrColumnWriter {
+impl StrOrBytesColumnWriter {
+    pub(crate) fn with_dictionary_id(dictionary_id: u32) -> StrOrBytesColumnWriter {
+        StrOrBytesColumnWriter {
             dictionary_id,
             column_writer: Default::default(),
         }
@@ -262,4 +298,27 @@ mod tests {
         test_column_writer_coercion_aux(&[1i64.into(), 1u64.into()], NumericalType::I64);
         test_column_writer_coercion_aux(&[u64::MAX.into(), (-1i64).into()], NumericalType::F64);
     }
+
+    #[test]
+    #[should_panic]
+    fn test_compatible_numerical_types_static_incompatible_type() {
+        let mut compatible_numerical_types =
+            CompatibleNumericalTypes::StaticType(NumericalType::U64);
+        compatible_numerical_types.accept_value(NumericalValue::I64(1i64));
+    }
+
+    #[test]
+    fn test_compatible_numerical_types_static_different_type_forbidden() {
+        let mut compatible_numerical_types =
+            CompatibleNumericalTypes::StaticType(NumericalType::U64);
+        compatible_numerical_types.accept_value(NumericalValue::U64(u64::MAX));
+    }
+
+    #[test]
+    fn test_compatible_numerical_types_static() {
+        for typ in [NumericalType::I64, NumericalType::I64, NumericalType::F64] {
+            let compatible_numerical_types = CompatibleNumericalTypes::StaticType(typ);
+            assert_eq!(compatible_numerical_types.to_numerical_type(), typ);
+        }
+    }
 }

columnar/src/columnar/writer/mod.rs

@@ -4,6 +4,7 @@ mod serializer;
 mod value_index;
 
 use std::io;
+use std::net::Ipv6Addr;
 
 use column_operation::ColumnOperation;
 use common::CountingWriter;
@@ -11,10 +12,12 @@ use serializer::ColumnarSerializer;
 use stacker::{Addr, ArenaHashMap, MemoryArena};
 
 use crate::column_index::SerializableColumnIndex;
-use crate::column_values::{ColumnValues, MonotonicallyMappableToU64, VecColumn};
-use crate::columnar::column_type::{ColumnType, ColumnTypeCategory};
+use crate::column_values::{
+    ColumnValues, MonotonicallyMappableToU128, MonotonicallyMappableToU64, VecColumn,
+};
+use crate::columnar::column_type::ColumnType;
 use crate::columnar::writer::column_writers::{
-    ColumnWriter, NumericalColumnWriter, StrColumnWriter,
+    ColumnWriter, NumericalColumnWriter, StrOrBytesColumnWriter,
 };
 use crate::columnar::writer::value_index::{IndexBuilder, PreallocatedIndexBuilders};
 use crate::dictionary::{DictionaryBuilder, TermIdMapping, UnorderedId};
@@ -30,6 +33,7 @@ struct SpareBuffers {
     u64_values: Vec<u64>,
     f64_values: Vec<f64>,
     bool_values: Vec<bool>,
+    ip_addr_values: Vec<Ipv6Addr>,
 }
 
 /// Makes it possible to create a new columnar.
@@ -47,8 +51,11 @@ struct SpareBuffers {
 /// ```
 pub struct ColumnarWriter {
     numerical_field_hash_map: ArenaHashMap,
+    datetime_field_hash_map: ArenaHashMap,
     bool_field_hash_map: ArenaHashMap,
+    ip_addr_field_hash_map: ArenaHashMap,
     bytes_field_hash_map: ArenaHashMap,
+    str_field_hash_map: ArenaHashMap,
     arena: MemoryArena,
     // Dictionaries used to store dictionary-encoded values.
     dictionaries: Vec<DictionaryBuilder>,
@@ -60,7 +67,10 @@ impl Default for ColumnarWriter {
         ColumnarWriter {
             numerical_field_hash_map: ArenaHashMap::new(10_000),
             bool_field_hash_map: ArenaHashMap::new(10_000),
+            ip_addr_field_hash_map: ArenaHashMap::new(10_000),
             bytes_field_hash_map: ArenaHashMap::new(10_000),
+            str_field_hash_map: ArenaHashMap::new(10_000),
+            datetime_field_hash_map: ArenaHashMap::new(10_000),
             dictionaries: Vec::new(),
             arena: MemoryArena::default(),
             buffers: SpareBuffers::default(),
@@ -68,20 +78,115 @@ impl Default for ColumnarWriter {
     }
 }
 
+#[inline]
+fn mutate_or_create_column<V, TMutator>(
+    arena_hash_map: &mut ArenaHashMap,
+    column_name: &str,
+    updater: TMutator,
+) where
+    V: Copy + 'static,
+    TMutator: FnMut(Option<V>) -> V,
+{
+    assert!(
+        !column_name.as_bytes().contains(&0u8),
+        "key may not contain the 0 byte"
+    );
+    arena_hash_map.mutate_or_create(column_name.as_bytes(), updater);
+}
+
 impl ColumnarWriter {
+    pub fn mem_usage(&self) -> usize {
+        // TODO add dictionary builders.
+        self.arena.mem_usage()
+            + self.numerical_field_hash_map.mem_usage()
+            + self.bool_field_hash_map.mem_usage()
+            + self.bytes_field_hash_map.mem_usage()
+            + self.str_field_hash_map.mem_usage()
+            + self.ip_addr_field_hash_map.mem_usage()
+            + self.datetime_field_hash_map.mem_usage()
+    }
+
+    pub fn record_column_type(&mut self, column_name: &str, column_type: ColumnType) {
+        match column_type {
+            ColumnType::Str | ColumnType::Bytes => {
+                let (hash_map, dictionaries) = (
+                    if column_type == ColumnType::Str {
+                        &mut self.str_field_hash_map
+                    } else {
+                        &mut self.bytes_field_hash_map
+                    },
+                    &mut self.dictionaries,
+                );
+                mutate_or_create_column(
+                    hash_map,
+                    column_name,
+                    |column_opt: Option<StrOrBytesColumnWriter>| {
+                        if let Some(column_writer) = column_opt {
+                            column_writer
+                        } else {
+                            let dictionary_id = dictionaries.len() as u32;
+                            dictionaries.push(DictionaryBuilder::default());
+                            StrOrBytesColumnWriter::with_dictionary_id(dictionary_id)
+                        }
+                    },
+                );
+            }
+            ColumnType::Bool => {
+                mutate_or_create_column(
+                    &mut self.bool_field_hash_map,
+                    column_name,
+                    |column_opt: Option<ColumnWriter>| column_opt.unwrap_or_default(),
+                );
+            }
+            ColumnType::DateTime => {
+                mutate_or_create_column(
+                    &mut self.datetime_field_hash_map,
+                    column_name,
+                    |column_opt: Option<ColumnWriter>| column_opt.unwrap_or_default(),
+                );
+            }
+            ColumnType::I64 | ColumnType::F64 | ColumnType::U64 => {
+                let numerical_type = column_type.numerical_type().unwrap();
+                mutate_or_create_column(
+                    &mut self.numerical_field_hash_map,
+                    column_name,
+                    |column_opt: Option<NumericalColumnWriter>| {
+                        let mut column: NumericalColumnWriter = column_opt.unwrap_or_default();
+                        column.force_numerical_type(numerical_type);
+                        column
+                    },
+                );
+            }
+            ColumnType::IpAddr => mutate_or_create_column(
+                &mut self.ip_addr_field_hash_map,
+                column_name,
+                |column_opt: Option<ColumnWriter>| column_opt.unwrap_or_default(),
+            ),
+        }
+    }
+
+    pub fn force_numerical_type(&mut self, column_name: &str, numerical_type: NumericalType) {
+        mutate_or_create_column(
+            &mut self.numerical_field_hash_map,
+            column_name,
+            |column_opt: Option<NumericalColumnWriter>| {
+                let mut column: NumericalColumnWriter = column_opt.unwrap_or_default();
+                column.force_numerical_type(numerical_type);
+                column
+            },
+        );
+    }
+
     pub fn record_numerical<T: Into<NumericalValue> + Copy>(
         &mut self,
         doc: RowId,
         column_name: &str,
         numerical_value: T,
     ) {
-        assert!(
-            !column_name.as_bytes().contains(&0u8),
-            "key may not contain the 0 byte"
-        );
         let (hash_map, arena) = (&mut self.numerical_field_hash_map, &mut self.arena);
-        hash_map.mutate_or_create(
-            column_name.as_bytes(),
+        mutate_or_create_column(
+            hash_map,
+            column_name,
             |column_opt: Option<NumericalColumnWriter>| {
                 let mut column: NumericalColumnWriter = column_opt.unwrap_or_default();
                 column.record_numerical_value(doc, numerical_value.into(), arena);
@@ -90,23 +195,62 @@ impl ColumnarWriter {
         );
     }
 
-    pub fn record_bool(&mut self, doc: RowId, column_name: &str, val: bool) {
+    pub fn record_ip_addr(&mut self, doc: RowId, column_name: &str, ip_addr: Ipv6Addr) {
         assert!(
             !column_name.as_bytes().contains(&0u8),
             "key may not contain the 0 byte"
         );
-        let (hash_map, arena) = (&mut self.bool_field_hash_map, &mut self.arena);
+        let (hash_map, arena) = (&mut self.ip_addr_field_hash_map, &mut self.arena);
         hash_map.mutate_or_create(
             column_name.as_bytes(),
             |column_opt: Option<ColumnWriter>| {
                 let mut column: ColumnWriter = column_opt.unwrap_or_default();
-                column.record(doc, val, arena);
+                column.record(doc, ip_addr, arena);
                 column
             },
         );
     }
 
+    pub fn record_bool(&mut self, doc: RowId, column_name: &str, val: bool) {
+        let (hash_map, arena) = (&mut self.bool_field_hash_map, &mut self.arena);
+        mutate_or_create_column(hash_map, column_name, |column_opt: Option<ColumnWriter>| {
+            let mut column: ColumnWriter = column_opt.unwrap_or_default();
+            column.record(doc, val, arena);
+            column
+        });
+    }
+
+    pub fn record_datetime(&mut self, doc: RowId, column_name: &str, datetime: crate::DateTime) {
+        let (hash_map, arena) = (&mut self.datetime_field_hash_map, &mut self.arena);
+        mutate_or_create_column(hash_map, column_name, |column_opt: Option<ColumnWriter>| {
+            let mut column: ColumnWriter = column_opt.unwrap_or_default();
+            column.record(doc, NumericalValue::I64(datetime.timestamp_micros), arena);
+            column
+        });
+    }
+
     pub fn record_str(&mut self, doc: RowId, column_name: &str, value: &str) {
+        let (hash_map, arena, dictionaries) = (
+            &mut self.str_field_hash_map,
+            &mut self.arena,
+            &mut self.dictionaries,
+        );
+        hash_map.mutate_or_create(
+            column_name.as_bytes(),
+            |column_opt: Option<StrOrBytesColumnWriter>| {
+                let mut column: StrOrBytesColumnWriter = column_opt.unwrap_or_else(|| {
+                    // Each column has its own dictionary
+                    let dictionary_id = dictionaries.len() as u32;
+                    dictionaries.push(DictionaryBuilder::default());
+                    StrOrBytesColumnWriter::with_dictionary_id(dictionary_id)
+                });
+                column.record_bytes(doc, value.as_bytes(), dictionaries, arena);
+                column
+            },
+        );
+    }
+
+    pub fn record_bytes(&mut self, doc: RowId, column_name: &str, value: &[u8]) {
         assert!(
             !column_name.as_bytes().contains(&0u8),
             "key may not contain the 0 byte"
@@ -118,43 +262,67 @@ impl ColumnarWriter {
         );
         hash_map.mutate_or_create(
             column_name.as_bytes(),
-            |column_opt: Option<StrColumnWriter>| {
-                let mut column: StrColumnWriter = column_opt.unwrap_or_else(|| {
+            |column_opt: Option<StrOrBytesColumnWriter>| {
+                let mut column: StrOrBytesColumnWriter = column_opt.unwrap_or_else(|| {
                     // Each column has its own dictionary
                     let dictionary_id = dictionaries.len() as u32;
                     dictionaries.push(DictionaryBuilder::default());
-                    StrColumnWriter::with_dictionary_id(dictionary_id)
+                    StrOrBytesColumnWriter::with_dictionary_id(dictionary_id)
                 });
-                column.record_bytes(doc, value.as_bytes(), dictionaries, arena);
+                column.record_bytes(doc, value, dictionaries, arena);
                 column
             },
         );
     }
-
     pub fn serialize(&mut self, num_docs: RowId, wrt: &mut dyn io::Write) -> io::Result<()> {
         let mut serializer = ColumnarSerializer::new(wrt);
-        let mut field_columns: Vec<(&[u8], ColumnTypeCategory, Addr)> = self
+        let mut columns: Vec<(&[u8], ColumnType, Addr)> = self
             .numerical_field_hash_map
             .iter()
-            .map(|(term, addr, _)| (term, ColumnTypeCategory::Numerical, addr))
+            .map(|(column_name, addr, _)| {
+                let numerical_column_writer: NumericalColumnWriter =
+                    self.numerical_field_hash_map.read(addr);
+                let column_type = numerical_column_writer.numerical_type().into();
+                (column_name, column_type, addr)
+            })
             .collect();
-        field_columns.extend(
+        columns.extend(
             self.bytes_field_hash_map
                 .iter()
-                .map(|(term, addr, _)| (term, ColumnTypeCategory::Str, addr)),
+                .map(|(term, addr, _)| (term, ColumnType::Bytes, addr)),
         );
-        field_columns.extend(
+        columns.extend(
+            self.str_field_hash_map
+                .iter()
+                .map(|(column_name, addr, _)| (column_name, ColumnType::Str, addr)),
+        );
+        columns.extend(
             self.bool_field_hash_map
                 .iter()
-                .map(|(term, addr, _)| (term, ColumnTypeCategory::Bool, addr)),
+                .map(|(column_name, addr, _)| (column_name, ColumnType::Bool, addr)),
         );
-        field_columns.sort_unstable_by_key(|(column_name, col_type, _)| (*column_name, *col_type));
+        columns.extend(
+            self.ip_addr_field_hash_map
+                .iter()
+                .map(|(column_name, addr, _)| (column_name, ColumnType::IpAddr, addr)),
+        );
+        columns.extend(
+            self.datetime_field_hash_map
+                .iter()
+                .map(|(column_name, addr, _)| (column_name, ColumnType::DateTime, addr)),
+        );
+        columns.sort_unstable_by_key(|(column_name, col_type, _)| (*column_name, *col_type));
+
         let (arena, buffers, dictionaries) = (&self.arena, &mut self.buffers, &self.dictionaries);
         let mut symbol_byte_buffer: Vec<u8> = Vec::new();
-        for (column_name, bytes_or_numerical, addr) in field_columns {
-            match bytes_or_numerical {
-                ColumnTypeCategory::Bool => {
-                    let column_writer: ColumnWriter = self.bool_field_hash_map.read(addr);
+        for (column_name, column_type, addr) in columns {
+            match column_type {
+                ColumnType::Bool | ColumnType::DateTime => {
+                    let column_writer: ColumnWriter = if column_type == ColumnType::Bool {
+                        self.bool_field_hash_map.read(addr)
+                    } else {
+                        self.datetime_field_hash_map.read(addr)
+                    };
                     let cardinality = column_writer.get_cardinality(num_docs);
                     let mut column_serializer =
                         serializer.serialize_column(column_name, ColumnType::Bool);
@@ -166,29 +334,50 @@ impl ColumnarWriter {
                         &mut column_serializer,
                     )?;
                 }
-                ColumnTypeCategory::Str => {
-                    let str_column_writer: StrColumnWriter = self.bytes_field_hash_map.read(addr);
-                    let dictionary_builder =
-                        &dictionaries[str_column_writer.dictionary_id as usize];
-                    let cardinality = str_column_writer.column_writer.get_cardinality(num_docs);
+                ColumnType::IpAddr => {
+                    let column_writer: ColumnWriter = self.ip_addr_field_hash_map.read(addr);
+                    let cardinality = column_writer.get_cardinality(num_docs);
                     let mut column_serializer =
-                        serializer.serialize_column(column_name, ColumnType::Bytes);
-                    serialize_bytes_column(
+                        serializer.serialize_column(column_name, ColumnType::IpAddr);
+                    serialize_ip_addr_column(
                         cardinality,
                         num_docs,
-                        dictionary_builder,
-                        str_column_writer.operation_iterator(arena, &mut symbol_byte_buffer),
+                        column_writer.operation_iterator(arena, &mut symbol_byte_buffer),
                         buffers,
                         &mut column_serializer,
                     )?;
                 }
-                ColumnTypeCategory::Numerical => {
+                ColumnType::Bytes | ColumnType::Str => {
+                    let str_or_bytes_column_writer: StrOrBytesColumnWriter =
+                        if column_type == ColumnType::Bytes {
+                            self.bytes_field_hash_map.read(addr)
+                        } else {
+                            self.str_field_hash_map.read(addr)
+                        };
+                    let dictionary_builder =
+                        &dictionaries[str_or_bytes_column_writer.dictionary_id as usize];
+                    let cardinality = str_or_bytes_column_writer
+                        .column_writer
+                        .get_cardinality(num_docs);
+                    let mut column_serializer =
+                        serializer.serialize_column(column_name, column_type);
+                    serialize_bytes_or_str_column(
+                        cardinality,
+                        num_docs,
+                        dictionary_builder,
+                        str_or_bytes_column_writer
+                            .operation_iterator(arena, &mut symbol_byte_buffer),
+                        buffers,
+                        &mut column_serializer,
+                    )?;
+                }
+                ColumnType::I64 | ColumnType::F64 | ColumnType::U64 => {
                     let numerical_column_writer: NumericalColumnWriter =
                         self.numerical_field_hash_map.read(addr);
-                    let (numerical_type, cardinality) =
-                        numerical_column_writer.column_type_and_cardinality(num_docs);
-                    let mut column_serializer = serializer
-                        .serialize_column(column_name, ColumnType::Numerical(numerical_type));
+                    let numerical_type = column_type.numerical_type().unwrap();
+                    let cardinality = numerical_column_writer.cardinality(num_docs);
+                    let mut column_serializer =
+                        serializer.serialize_column(column_name, ColumnType::from(numerical_type));
                     serialize_numerical_column(
                         cardinality,
                         num_docs,
@@ -205,7 +394,7 @@ impl ColumnarWriter {
     }
 }
 
-fn serialize_bytes_column(
+fn serialize_bytes_or_str_column(
     cardinality: Cardinality,
     num_docs: RowId,
     dictionary_builder: &DictionaryBuilder,
@@ -232,7 +421,7 @@ fn serialize_bytes_column(
             ColumnOperation::NewDoc(doc) => ColumnOperation::NewDoc(doc),
         }
     });
-    serialize_column(
+    send_to_serialize_column_mappable_to_u64(
         operation_iterator,
         cardinality,
         num_docs,
@@ -261,7 +450,7 @@ fn serialize_numerical_column(
     } = buffers;
     match numerical_type {
         NumericalType::I64 => {
-            serialize_column(
+            send_to_serialize_column_mappable_to_u64(
                 coerce_numerical_symbol::<i64>(op_iterator),
                 cardinality,
                 num_docs,
@@ -271,7 +460,7 @@ fn serialize_numerical_column(
             )?;
         }
         NumericalType::U64 => {
-            serialize_column(
+            send_to_serialize_column_mappable_to_u64(
                 coerce_numerical_symbol::<u64>(op_iterator),
                 cardinality,
                 num_docs,
@@ -281,7 +470,7 @@ fn serialize_numerical_column(
             )?;
         }
         NumericalType::F64 => {
-            serialize_column(
+            send_to_serialize_column_mappable_to_u64(
                 coerce_numerical_symbol::<f64>(op_iterator),
                 cardinality,
                 num_docs,
@@ -306,7 +495,7 @@ fn serialize_bool_column(
         bool_values,
         ..
     } = buffers;
-    serialize_column(
+    send_to_serialize_column_mappable_to_u64(
         column_operations_it,
         cardinality,
         num_docs,
@@ -317,7 +506,76 @@ fn serialize_bool_column(
     Ok(())
 }
 
-fn serialize_column<
+fn serialize_ip_addr_column(
+    cardinality: Cardinality,
+    num_docs: RowId,
+    column_operations_it: impl Iterator<Item = ColumnOperation<Ipv6Addr>>,
+    buffers: &mut SpareBuffers,
+    wrt: &mut impl io::Write,
+) -> io::Result<()> {
+    let SpareBuffers {
+        value_index_builders,
+        ip_addr_values,
+        ..
+    } = buffers;
+    send_to_serialize_column_mappable_to_u128(
+        column_operations_it,
+        cardinality,
+        num_docs,
+        value_index_builders,
+        ip_addr_values,
+        wrt,
+    )?;
+    Ok(())
+}
+
+fn send_to_serialize_column_mappable_to_u128<
+    T: Copy + std::fmt::Debug + Send + Sync + MonotonicallyMappableToU128 + PartialOrd,
+>(
+    op_iterator: impl Iterator<Item = ColumnOperation<T>>,
+    cardinality: Cardinality,
+    num_docs: RowId,
+    value_index_builders: &mut PreallocatedIndexBuilders,
+    values: &mut Vec<T>,
+    mut wrt: impl io::Write,
+) -> io::Result<()>
+where
+    for<'a> VecColumn<'a, T>: ColumnValues<T>,
+{
+    values.clear();
+    // TODO: split index and values
+    let serializable_column_index = match cardinality {
+        Cardinality::Full => {
+            consume_operation_iterator(
+                op_iterator,
+                value_index_builders.borrow_required_index_builder(),
+                values,
+            );
+            SerializableColumnIndex::Full
+        }
+        Cardinality::Optional => {
+            let optional_index_builder = value_index_builders.borrow_optional_index_builder();
+            consume_operation_iterator(op_iterator, optional_index_builder, values);
+            let optional_index = optional_index_builder.finish(num_docs);
+            SerializableColumnIndex::Optional(Box::new(optional_index))
+        }
+        Cardinality::Multivalued => {
+            let multivalued_index_builder = value_index_builders.borrow_multivalued_index_builder();
+            consume_operation_iterator(op_iterator, multivalued_index_builder, values);
+            let multivalued_index = multivalued_index_builder.finish(num_docs);
+            SerializableColumnIndex::Multivalued(Box::new(multivalued_index))
+        }
+    };
+    crate::column::serialize_column_mappable_to_u128(
+        serializable_column_index,
+        || values.iter().cloned(),
+        values.len() as u32,
+        &mut wrt,
+    )?;
+    Ok(())
+}
+
+fn send_to_serialize_column_mappable_to_u64<
     T: Copy + Default + std::fmt::Debug + Send + Sync + MonotonicallyMappableToU64 + PartialOrd,
 >(
     op_iterator: impl Iterator<Item = ColumnOperation<T>>,
@@ -350,11 +608,10 @@ where
             let multivalued_index_builder = value_index_builders.borrow_multivalued_index_builder();
             consume_operation_iterator(op_iterator, multivalued_index_builder, values);
             let multivalued_index = multivalued_index_builder.finish(num_docs);
-            todo!();
-            // SerializableColumnIndex::Multivalued(Box::new(multivalued_index))
+            SerializableColumnIndex::Multivalued(Box::new(multivalued_index))
         }
     };
-    crate::column::serialize_column_u64(
+    crate::column::serialize_column_mappable_to_u64(
         serializable_column_index,
         &VecColumn::from(&values[..]),
         &mut wrt,
@@ -392,59 +649,12 @@ fn consume_operation_iterator<T: std::fmt::Debug, TIndexBuilder: IndexBuilder>(
     }
 }
 
-// /// Serializes the column with the codec with the best estimate on the data.
-// fn serialize_numerical<T: MonotonicallyMappableToU64>(
-//     value_index: ValueIndexInfo,
-//     typed_column: impl Column<T>,
-//     output: &mut impl io::Write,
-//     codecs: &[FastFieldCodecType],
-// ) -> io::Result<()> {
-
-//     let counting_writer = CountingWriter::wrap(output);
-//     serialize_value_index(value_index, output)?;
-//     let value_index_len = counting_writer.written_bytes();
-//     let output = counting_writer.finish();
-
-//     serialize_column(value_index, output)?;
-//     let column = monotonic_map_column(
-//         typed_column,
-//         crate::column::monotonic_mapping::StrictlyMonotonicMappingToInternal::<T>::new(),
-//     );
-//     let header = Header::compute_header(&column, codecs).ok_or_else(|| {
-//         io::Error::new(
-//             io::ErrorKind::InvalidInput,
-//             format!(
-//                 "Data cannot be serialized with this list of codec. {:?}",
-//                 codecs
-//             ),
-//         )
-//     })?;
-//     header.serialize(output)?;
-//     let normalized_column = header.normalize_column(column);
-//     assert_eq!(normalized_column.min_value(), 0u64);
-//     serialize_given_codec(normalized_column, header.codec_type, output)?;
-
-//     let column_header = ColumnFooter {
-//         value_index_len: todo!(),
-//         cardinality: todo!(),
-//     };
-
-//     let null_index_footer = NullIndexFooter {
-//         cardinality: value_index.get_cardinality(),
-//         null_index_codec: NullIndexCodec::Full,
-//         null_index_byte_range: 0..0,
-//     };
-//     append_null_index_footer(output, null_index_footer)?;
-//     Ok(())
-// }
-
 #[cfg(test)]
 mod tests {
-    use column_operation::ColumnOperation;
     use stacker::MemoryArena;
 
-    use super::*;
-    use crate::value::NumericalValue;
+    use crate::columnar::writer::column_operation::ColumnOperation;
+    use crate::{Cardinality, NumericalValue};
 
     #[test]
     fn test_column_writer_required_simple() {

columnar/src/columnar/writer/serializer.rs

@@ -97,10 +97,10 @@ mod tests {
     #[test]
     fn test_prepare_key_bytes() {
         let mut buffer: Vec<u8> = b"somegarbage".to_vec();
-        prepare_key(b"root\0child", ColumnType::Bytes, &mut buffer);
+        prepare_key(b"root\0child", ColumnType::Str, &mut buffer);
         assert_eq!(buffer.len(), 12);
         assert_eq!(&buffer[..10], b"root\0child");
         assert_eq!(buffer[10], 0u8);
-        assert_eq!(buffer[11], ColumnType::Bytes.to_code());
+        assert_eq!(buffer[11], ColumnType::Str.to_code());
     }
 }

columnar/src/columnar/writer/value_index.rs

@@ -45,16 +45,6 @@ impl<'a> SerializableOptionalIndex<'a> for SingleValueArrayIndex<'a> {
     }
 }
 
-impl OptionalIndexBuilder {
-    fn num_non_nulls(&self) -> u32 {
-        self.docs.len() as u32
-    }
-
-    fn iter(&self) -> Box<dyn Iterator<Item = u32> + '_> {
-        Box::new(self.docs.iter().copied())
-    }
-}
-
 impl OptionalIndexBuilder {
     pub fn finish<'a>(&'a mut self, num_rows: RowId) -> impl SerializableOptionalIndex + 'a {
         debug_assert!(self
@@ -96,7 +86,7 @@ pub struct MultivaluedIndexBuilder {
 impl MultivaluedIndexBuilder {
     pub fn finish(&mut self, num_docs: RowId) -> impl ColumnValues<u32> + '_ {
         self.start_offsets
-            .resize(num_docs as usize, self.total_num_vals_seen);
+            .resize(num_docs as usize + 1, self.total_num_vals_seen);
         VecColumn {
             values: &&self.start_offsets[..],
             min_value: 0,
@@ -188,7 +178,7 @@ mod tests {
                 .finish(4u32)
                 .iter()
                 .collect::<Vec<u32>>(),
-            vec![0, 0, 2, 3]
+            vec![0, 0, 2, 3, 3]
         );
         multivalued_value_index_builder.reset();
         multivalued_value_index_builder.record_row(2u32);
@@ -199,7 +189,7 @@ mod tests {
                 .finish(4u32)
                 .iter()
                 .collect::<Vec<u32>>(),
-            vec![0, 0, 0, 2]
+            vec![0, 0, 0, 2, 2]
         );
     }
 }

columnar/src/dynamic_column.rs

@@ -1,12 +1,14 @@
 use std::io;
-use std::net::IpAddr;
+use std::net::Ipv6Addr;
+use std::sync::Arc;
 
 use common::file_slice::FileSlice;
 use common::{HasLen, OwnedBytes};
 
-use crate::column::{BytesColumn, Column};
+use crate::column::{BytesColumn, Column, StrColumn};
+use crate::column_values::{monotonic_map_column, StrictlyMonotonicFn};
 use crate::columnar::ColumnType;
-use crate::DateTime;
+use crate::{DateTime, NumericalType};
 
 #[derive(Clone)]
 pub enum DynamicColumn {
@@ -14,41 +16,163 @@ pub enum DynamicColumn {
     I64(Column<i64>),
     U64(Column<u64>),
     F64(Column<f64>),
-    IpAddr(Column<IpAddr>),
+    IpAddr(Column<Ipv6Addr>),
     DateTime(Column<DateTime>),
-    Str(BytesColumn),
+    Bytes(BytesColumn),
+    Str(StrColumn),
 }
 
-impl From<Column<i64>> for DynamicColumn {
-    fn from(column_i64: Column<i64>) -> Self {
-        DynamicColumn::I64(column_i64)
+impl DynamicColumn {
+    pub fn column_type(&self) -> ColumnType {
+        match self {
+            DynamicColumn::Bool(_) => ColumnType::Bool,
+            DynamicColumn::I64(_) => ColumnType::I64,
+            DynamicColumn::U64(_) => ColumnType::U64,
+            DynamicColumn::F64(_) => ColumnType::F64,
+            DynamicColumn::IpAddr(_) => ColumnType::IpAddr,
+            DynamicColumn::DateTime(_) => ColumnType::DateTime,
+            DynamicColumn::Bytes(_) => ColumnType::Bytes,
+            DynamicColumn::Str(_) => ColumnType::Str,
+        }
+    }
+
+    pub fn is_numerical(&self) -> bool {
+        self.column_type().numerical_type().is_some()
+    }
+
+    pub fn is_f64(&self) -> bool {
+        self.column_type().numerical_type() == Some(NumericalType::F64)
+    }
+    pub fn is_i64(&self) -> bool {
+        self.column_type().numerical_type() == Some(NumericalType::I64)
+    }
+    pub fn is_u64(&self) -> bool {
+        self.column_type().numerical_type() == Some(NumericalType::U64)
+    }
+
+    pub fn coerce_to_f64(self) -> Option<DynamicColumn> {
+        match self {
+            DynamicColumn::I64(column) => Some(DynamicColumn::F64(Column {
+                idx: column.idx,
+                values: Arc::new(monotonic_map_column(column.values, MapI64ToF64)),
+            })),
+            DynamicColumn::U64(column) => Some(DynamicColumn::F64(Column {
+                idx: column.idx,
+                values: Arc::new(monotonic_map_column(column.values, MapU64ToF64)),
+            })),
+            DynamicColumn::F64(_) => Some(self),
+            _ => None,
+        }
+    }
+    pub fn coerce_to_i64(self) -> Option<DynamicColumn> {
+        match self {
+            DynamicColumn::U64(column) => {
+                if column.max_value() > i64::MAX as u64 {
+                    return None;
+                }
+                Some(DynamicColumn::I64(Column {
+                    idx: column.idx,
+                    values: Arc::new(monotonic_map_column(column.values, MapU64ToI64)),
+                }))
+            }
+            DynamicColumn::I64(_) => Some(self),
+            _ => None,
+        }
+    }
+    pub fn coerce_to_u64(self) -> Option<DynamicColumn> {
+        match self {
+            DynamicColumn::I64(column) => {
+                if column.min_value() < 0 {
+                    return None;
+                }
+                Some(DynamicColumn::U64(Column {
+                    idx: column.idx,
+                    values: Arc::new(monotonic_map_column(column.values, MapI64ToU64)),
+                }))
+            }
+            DynamicColumn::U64(_) => Some(self),
+            _ => None,
+        }
     }
 }
 
-impl From<Column<u64>> for DynamicColumn {
-    fn from(column_u64: Column<u64>) -> Self {
-        DynamicColumn::U64(column_u64)
+struct MapI64ToF64;
+impl StrictlyMonotonicFn<i64, f64> for MapI64ToF64 {
+    #[inline(always)]
+    fn mapping(&self, inp: i64) -> f64 {
+        inp as f64
+    }
+    #[inline(always)]
+    fn inverse(&self, out: f64) -> i64 {
+        out as i64
     }
 }
 
-impl From<Column<f64>> for DynamicColumn {
-    fn from(column_f64: Column<f64>) -> Self {
-        DynamicColumn::F64(column_f64)
+struct MapU64ToF64;
+impl StrictlyMonotonicFn<u64, f64> for MapU64ToF64 {
+    #[inline(always)]
+    fn mapping(&self, inp: u64) -> f64 {
+        inp as f64
+    }
+    #[inline(always)]
+    fn inverse(&self, out: f64) -> u64 {
+        out as u64
     }
 }
 
-impl From<Column<bool>> for DynamicColumn {
-    fn from(bool_column: Column<bool>) -> Self {
-        DynamicColumn::Bool(bool_column)
+struct MapU64ToI64;
+impl StrictlyMonotonicFn<u64, i64> for MapU64ToI64 {
+    #[inline(always)]
+    fn mapping(&self, inp: u64) -> i64 {
+        inp as i64
+    }
+    #[inline(always)]
+    fn inverse(&self, out: i64) -> u64 {
+        out as u64
     }
 }
 
-impl From<BytesColumn> for DynamicColumn {
-    fn from(dictionary_encoded_col: BytesColumn) -> Self {
-        DynamicColumn::Str(dictionary_encoded_col)
+struct MapI64ToU64;
+impl StrictlyMonotonicFn<i64, u64> for MapI64ToU64 {
+    #[inline(always)]
+    fn mapping(&self, inp: i64) -> u64 {
+        inp as u64
+    }
+    #[inline(always)]
+    fn inverse(&self, out: u64) -> i64 {
+        out as i64
     }
 }
 
+macro_rules! static_dynamic_conversions {
+    ($typ:ty, $enum_name:ident) => {
+        impl Into<Option<$typ>> for DynamicColumn {
+            fn into(self) -> Option<$typ> {
+                if let DynamicColumn::$enum_name(col) = self {
+                    Some(col)
+                } else {
+                    None
+                }
+            }
+        }
+
+        impl From<$typ> for DynamicColumn {
+            fn from(typed_column: $typ) -> Self {
+                DynamicColumn::$enum_name(typed_column)
+            }
+        }
+    };
+}
+
+static_dynamic_conversions!(Column<bool>, Bool);
+static_dynamic_conversions!(Column<u64>, U64);
+static_dynamic_conversions!(Column<i64>, I64);
+static_dynamic_conversions!(Column<f64>, F64);
+static_dynamic_conversions!(Column<crate::DateTime>, DateTime);
+static_dynamic_conversions!(StrColumn, Str);
+static_dynamic_conversions!(BytesColumn, Bytes);
+static_dynamic_conversions!(Column<Ipv6Addr>, IpAddr);
+
 #[derive(Clone)]
 pub struct DynamicColumnHandle {
     pub(crate) file_slice: FileSlice,
@@ -56,31 +180,53 @@ pub struct DynamicColumnHandle {
 }
 
 impl DynamicColumnHandle {
+    // TODO rename load
     pub fn open(&self) -> io::Result<DynamicColumn> {
         let column_bytes: OwnedBytes = self.file_slice.read_bytes()?;
         self.open_internal(column_bytes)
     }
 
+    // TODO rename load_async
     pub async fn open_async(&self) -> io::Result<DynamicColumn> {
         let column_bytes: OwnedBytes = self.file_slice.read_bytes_async().await?;
         self.open_internal(column_bytes)
     }
 
+    /// Returns the `u64` fast field reader reader associated with `fields` of types
+    /// Str, u64, i64, f64, or datetime.
+    ///
+    /// If not, the fastfield reader will returns the u64-value associated with the original
+    /// FastValue.
+    pub fn open_u64_lenient(&self) -> io::Result<Option<Column<u64>>> {
+        let column_bytes = self.file_slice.read_bytes()?;
+        match self.column_type {
+            ColumnType::Str | ColumnType::Bytes => {
+                let column: BytesColumn = crate::column::open_column_bytes(column_bytes)?;
+                Ok(Some(column.term_ord_column))
+            }
+            ColumnType::Bool => Ok(None),
+            ColumnType::IpAddr => Ok(None),
+            ColumnType::I64 | ColumnType::U64 | ColumnType::F64 | ColumnType::DateTime => {
+                let column = crate::column::open_column_u64::<u64>(column_bytes)?;
+                Ok(Some(column))
+            }
+        }
+    }
+
     fn open_internal(&self, column_bytes: OwnedBytes) -> io::Result<DynamicColumn> {
         let dynamic_column: DynamicColumn = match self.column_type {
-            ColumnType::Bytes => crate::column::open_column_bytes(column_bytes)?.into(),
-            ColumnType::Numerical(numerical_type) => match numerical_type {
-                crate::NumericalType::I64 => {
-                    crate::column::open_column_u64::<i64>(column_bytes)?.into()
-                }
-                crate::NumericalType::U64 => {
-                    crate::column::open_column_u64::<u64>(column_bytes)?.into()
-                }
-                crate::NumericalType::F64 => {
-                    crate::column::open_column_u64::<f64>(column_bytes)?.into()
-                }
-            },
+            ColumnType::Bytes => {
+                crate::column::open_column_bytes::<BytesColumn>(column_bytes)?.into()
+            }
+            ColumnType::Str => crate::column::open_column_bytes::<StrColumn>(column_bytes)?.into(),
+            ColumnType::I64 => crate::column::open_column_u64::<i64>(column_bytes)?.into(),
+            ColumnType::U64 => crate::column::open_column_u64::<u64>(column_bytes)?.into(),
+            ColumnType::F64 => crate::column::open_column_u64::<f64>(column_bytes)?.into(),
             ColumnType::Bool => crate::column::open_column_u64::<bool>(column_bytes)?.into(),
+            ColumnType::IpAddr => crate::column::open_column_u128::<Ipv6Addr>(column_bytes)?.into(),
+            ColumnType::DateTime => {
+                crate::column::open_column_u64::<crate::DateTime>(column_bytes)?.into()
+            }
         };
         Ok(dynamic_column)
     }

columnar/src/lib.rs

@@ -18,16 +18,21 @@ mod dynamic_column;
 pub(crate) mod utils;
 mod value;
 
-pub use columnar::{ColumnarReader, ColumnarWriter};
+pub use column::{BytesColumn, Column, StrColumn};
+pub use column_values::ColumnValues;
+pub use columnar::{
+    merge_columnar, ColumnType, ColumnarReader, ColumnarWriter, HasAssociatedColumnType,
+    MergeDocOrder,
+};
 pub use value::{NumericalType, NumericalValue};
 
-// pub use self::dynamic_column::DynamicColumnHandle;
+pub use self::dynamic_column::{DynamicColumn, DynamicColumnHandle};
 
 pub type RowId = u32;
 
-#[derive(Clone, Copy)]
+#[derive(Clone, Copy, PartialOrd, PartialEq, Default, Debug)]
 pub struct DateTime {
-    timestamp_micros: i64,
+    pub timestamp_micros: i64,
 }
 
 #[derive(Copy, Clone, Debug)]

columnar/src/tests.rs

@@ -1,10 +1,13 @@
+use std::net::Ipv6Addr;
+
+use crate::column_values::MonotonicallyMappableToU128;
 use crate::columnar::ColumnType;
 use crate::dynamic_column::{DynamicColumn, DynamicColumnHandle};
 use crate::value::NumericalValue;
 use crate::{Cardinality, ColumnarReader, ColumnarWriter};
 
 #[test]
-fn test_dataframe_writer_bytes() {
+fn test_dataframe_writer_str() {
     let mut dataframe_writer = ColumnarWriter::default();
     dataframe_writer.record_str(1u32, "my_string", "hello");
     dataframe_writer.record_str(3u32, "my_string", "helloeee");
@@ -14,7 +17,21 @@ fn test_dataframe_writer_bytes() {
     assert_eq!(columnar.num_columns(), 1);
     let cols: Vec<DynamicColumnHandle> = columnar.read_columns("my_string").unwrap();
     assert_eq!(cols.len(), 1);
-    assert_eq!(cols[0].num_bytes(), 165);
+    assert_eq!(cols[0].num_bytes(), 158);
+}
+
+#[test]
+fn test_dataframe_writer_bytes() {
+    let mut dataframe_writer = ColumnarWriter::default();
+    dataframe_writer.record_bytes(1u32, "my_string", b"hello");
+    dataframe_writer.record_bytes(3u32, "my_string", b"helloeee");
+    let mut buffer: Vec<u8> = Vec::new();
+    dataframe_writer.serialize(5, &mut buffer).unwrap();
+    let columnar = ColumnarReader::open(buffer).unwrap();
+    assert_eq!(columnar.num_columns(), 1);
+    let cols: Vec<DynamicColumnHandle> = columnar.read_columns("my_string").unwrap();
+    assert_eq!(cols.len(), 1);
+    assert_eq!(cols[0].num_bytes(), 158);
 }
 
 #[test]
@@ -28,7 +45,7 @@ fn test_dataframe_writer_bool() {
     assert_eq!(columnar.num_columns(), 1);
     let cols: Vec<DynamicColumnHandle> = columnar.read_columns("bool.value").unwrap();
     assert_eq!(cols.len(), 1);
-    assert_eq!(cols[0].num_bytes(), 29);
+    assert_eq!(cols[0].num_bytes(), 22);
     assert_eq!(cols[0].column_type(), ColumnType::Bool);
     let dyn_bool_col = cols[0].open().unwrap();
     let DynamicColumn::Bool(bool_col) = dyn_bool_col else { panic!(); };
@@ -36,6 +53,59 @@ fn test_dataframe_writer_bool() {
     assert_eq!(&vals, &[None, Some(false), None, Some(true), None,]);
 }
 
+#[test]
+fn test_dataframe_writer_u64_multivalued() {
+    let mut dataframe_writer = ColumnarWriter::default();
+    dataframe_writer.record_numerical(2u32, "divisor", 2u64);
+    dataframe_writer.record_numerical(3u32, "divisor", 3u64);
+    dataframe_writer.record_numerical(4u32, "divisor", 2u64);
+    dataframe_writer.record_numerical(5u32, "divisor", 5u64);
+    dataframe_writer.record_numerical(6u32, "divisor", 2u64);
+    dataframe_writer.record_numerical(6u32, "divisor", 3u64);
+    let mut buffer: Vec<u8> = Vec::new();
+    dataframe_writer.serialize(7, &mut buffer).unwrap();
+    let columnar = ColumnarReader::open(buffer).unwrap();
+    assert_eq!(columnar.num_columns(), 1);
+    let cols: Vec<DynamicColumnHandle> = columnar.read_columns("divisor").unwrap();
+    assert_eq!(cols.len(), 1);
+    assert_eq!(cols[0].num_bytes(), 29);
+    let dyn_i64_col = cols[0].open().unwrap();
+    let DynamicColumn::I64(divisor_col) = dyn_i64_col else { panic!(); };
+    assert_eq!(
+        divisor_col.get_cardinality(),
+        crate::Cardinality::Multivalued
+    );
+    assert_eq!(divisor_col.num_rows(), 7);
+}
+
+#[test]
+fn test_dataframe_writer_ip_addr() {
+    let mut dataframe_writer = ColumnarWriter::default();
+    dataframe_writer.record_ip_addr(1, "ip_addr", Ipv6Addr::from_u128(1001));
+    dataframe_writer.record_ip_addr(3, "ip_addr", Ipv6Addr::from_u128(1050));
+    let mut buffer: Vec<u8> = Vec::new();
+    dataframe_writer.serialize(5, &mut buffer).unwrap();
+    let columnar = ColumnarReader::open(buffer).unwrap();
+    assert_eq!(columnar.num_columns(), 1);
+    let cols: Vec<DynamicColumnHandle> = columnar.read_columns("ip_addr").unwrap();
+    assert_eq!(cols.len(), 1);
+    assert_eq!(cols[0].num_bytes(), 42);
+    assert_eq!(cols[0].column_type(), ColumnType::IpAddr);
+    let dyn_bool_col = cols[0].open().unwrap();
+    let DynamicColumn::IpAddr(ip_col) = dyn_bool_col else { panic!(); };
+    let vals: Vec<Option<Ipv6Addr>> = (0..5).map(|row_id| ip_col.first(row_id)).collect();
+    assert_eq!(
+        &vals,
+        &[
+            None,
+            Some(Ipv6Addr::from_u128(1001)),
+            None,
+            Some(Ipv6Addr::from_u128(1050)),
+            None,
+        ]
+    );
+}
+
 #[test]
 fn test_dataframe_writer_numerical() {
     let mut dataframe_writer = ColumnarWriter::default();
@@ -53,7 +123,7 @@ fn test_dataframe_writer_numerical() {
     // - header 14 bytes
     // - vals  8 //< due to padding? could have been 1byte?.
     // - null footer 6 bytes
-    assert_eq!(cols[0].num_bytes(), 40);
+    assert_eq!(cols[0].num_bytes(), 33);
     let column = cols[0].open().unwrap();
     let DynamicColumn::I64(column_i64) = column else { panic!(); };
     assert_eq!(column_i64.idx.get_cardinality(), Cardinality::Optional);
@@ -67,18 +137,76 @@ fn test_dataframe_writer_numerical() {
 }
 
 #[test]
-fn test_dictionary_encoded() {
+fn test_dictionary_encoded_str() {
     let mut buffer = Vec::new();
     let mut columnar_writer = ColumnarWriter::default();
-    columnar_writer.record_str(1, "my.column", "my.key");
-    columnar_writer.record_str(3, "my.column", "my.key2");
+    columnar_writer.record_str(1, "my.column", "a");
+    columnar_writer.record_str(3, "my.column", "c");
     columnar_writer.record_str(3, "my.column2", "different_column!");
+    columnar_writer.record_str(4, "my.column", "b");
     columnar_writer.serialize(5, &mut buffer).unwrap();
     let columnar_reader = ColumnarReader::open(buffer).unwrap();
     assert_eq!(columnar_reader.num_columns(), 2);
     let col_handles = columnar_reader.read_columns("my.column").unwrap();
     assert_eq!(col_handles.len(), 1);
     let DynamicColumn::Str(str_col) = col_handles[0].open().unwrap() else  { panic!(); };
+    let index: Vec<Option<u64>> = (0..5).map(|row_id| str_col.ords().first(row_id)).collect();
+    assert_eq!(index, &[None, Some(0), None, Some(2), Some(1)]);
     assert_eq!(str_col.num_rows(), 5);
-    // let term_ords = (0..)
+    let mut term_buffer = String::new();
+    let term_ords = str_col.ords();
+    assert_eq!(term_ords.first(0), None);
+    assert_eq!(term_ords.first(1), Some(0));
+    str_col.ord_to_str(0u64, &mut term_buffer).unwrap();
+    assert_eq!(term_buffer, "a");
+    assert_eq!(term_ords.first(2), None);
+    assert_eq!(term_ords.first(3), Some(2));
+    str_col.ord_to_str(2u64, &mut term_buffer).unwrap();
+    assert_eq!(term_buffer, "c");
+    assert_eq!(term_ords.first(4), Some(1));
+    str_col.ord_to_str(1u64, &mut term_buffer).unwrap();
+    assert_eq!(term_buffer, "b");
+}
+
+#[test]
+fn test_dictionary_encoded_bytes() {
+    let mut buffer = Vec::new();
+    let mut columnar_writer = ColumnarWriter::default();
+    columnar_writer.record_bytes(1, "my.column", b"a");
+    columnar_writer.record_bytes(3, "my.column", b"c");
+    columnar_writer.record_bytes(3, "my.column2", b"different_column!");
+    columnar_writer.record_bytes(4, "my.column", b"b");
+    columnar_writer.serialize(5, &mut buffer).unwrap();
+    let columnar_reader = ColumnarReader::open(buffer).unwrap();
+    assert_eq!(columnar_reader.num_columns(), 2);
+    let col_handles = columnar_reader.read_columns("my.column").unwrap();
+    assert_eq!(col_handles.len(), 1);
+    let DynamicColumn::Bytes(bytes_col) = col_handles[0].open().unwrap() else  { panic!(); };
+    let index: Vec<Option<u64>> = (0..5)
+        .map(|row_id| bytes_col.ords().first(row_id))
+        .collect();
+    assert_eq!(index, &[None, Some(0), None, Some(2), Some(1)]);
+    assert_eq!(bytes_col.num_rows(), 5);
+    let mut term_buffer = Vec::new();
+    let term_ords = bytes_col.ords();
+    assert_eq!(term_ords.first(0), None);
+    assert_eq!(term_ords.first(1), Some(0));
+    bytes_col
+        .dictionary
+        .ord_to_term(0u64, &mut term_buffer)
+        .unwrap();
+    assert_eq!(term_buffer, b"a");
+    assert_eq!(term_ords.first(2), None);
+    assert_eq!(term_ords.first(3), Some(2));
+    bytes_col
+        .dictionary
+        .ord_to_term(2u64, &mut term_buffer)
+        .unwrap();
+    assert_eq!(term_buffer, b"c");
+    assert_eq!(term_ords.first(4), Some(1));
+    bytes_col
+        .dictionary
+        .ord_to_term(1u64, &mut term_buffer)
+        .unwrap();
+    assert_eq!(term_buffer, b"b");
 }

columnar/src/value.rs

@@ -1,12 +1,22 @@
 use crate::InvalidData;
 
-#[derive(Copy, Clone, Debug, PartialEq)]
+#[derive(Copy, Clone, PartialEq, Debug)]
 pub enum NumericalValue {
     I64(i64),
     U64(u64),
     F64(f64),
 }
 
+impl NumericalValue {
+    pub fn numerical_type(&self) -> NumericalType {
+        match self {
+            NumericalValue::I64(_) => NumericalType::I64,
+            NumericalValue::U64(_) => NumericalType::U64,
+            NumericalValue::F64(_) => NumericalType::F64,
+        }
+    }
+}
+
 impl From<u64> for NumericalValue {
     fn from(val: u64) -> NumericalValue {
         NumericalValue::U64(val)
@@ -25,18 +35,6 @@ impl From<f64> for NumericalValue {
     }
 }
 
-impl NumericalValue {
-    pub fn numerical_type(&self) -> NumericalType {
-        match self {
-            NumericalValue::F64(_) => NumericalType::F64,
-            NumericalValue::I64(_) => NumericalType::I64,
-            NumericalValue::U64(_) => NumericalType::U64,
-        }
-    }
-}
-
-impl Eq for NumericalValue {}
-
 #[derive(Clone, Copy, Debug, Default, Hash, Eq, PartialEq)]
 #[repr(u8)]
 pub enum NumericalType {
@@ -106,6 +104,13 @@ impl Coerce for f64 {
     }
 }
 
+impl Coerce for crate::DateTime {
+    fn coerce(value: NumericalValue) -> Self {
+        let timestamp_micros = i64::coerce(value);
+        crate::DateTime { timestamp_micros }
+    }
+}
+
 #[cfg(test)]
 mod tests {
     use super::NumericalType;

examples/aggregation.rs

@@ -27,7 +27,7 @@ fn main() -> tantivy::Result<()> {
     let score_fieldtype =
         crate::schema::NumericOptions::default().set_fast(Cardinality::SingleValue);
     let highscore_field = schema_builder.add_f64_field("highscore", score_fieldtype.clone());
-    let price_field = schema_builder.add_f64_field("price", score_fieldtype.clone());
+    let price_field = schema_builder.add_f64_field("price", score_fieldtype);
 
     let schema = schema_builder.build();
 
@@ -112,7 +112,7 @@ fn main() -> tantivy::Result<()> {
                 ],
                 ..Default::default()
             }),
-            sub_aggregation: sub_agg_req_1.clone(),
+            sub_aggregation: sub_agg_req_1,
         }),
     )]
     .into_iter()
@@ -123,7 +123,7 @@ fn main() -> tantivy::Result<()> {
     let searcher = reader.searcher();
     let agg_res: AggregationResults = searcher.search(&term_query, &collector).unwrap();
 
-    let res: Value = serde_json::to_value(&agg_res)?;
+    let res: Value = serde_json::to_value(agg_res)?;
     println!("{}", serde_json::to_string_pretty(&res)?);
 
     Ok(())

examples/custom_collector.rs

@@ -14,7 +14,7 @@ use fastfield_codecs::Column;
 // Importing tantivy...
 use tantivy::collector::{Collector, SegmentCollector};
 use tantivy::query::QueryParser;
-use tantivy::schema::{Field, Schema, FAST, INDEXED, TEXT};
+use tantivy::schema::{Schema, FAST, INDEXED, TEXT};
 use tantivy::{doc, Index, Score, SegmentReader};
 
 #[derive(Default)]
@@ -52,11 +52,11 @@ impl Stats {
 }
 
 struct StatsCollector {
-    field: Field,
+    field: String,
 }
 
 impl StatsCollector {
-    fn with_field(field: Field) -> StatsCollector {
+    fn with_field(field: String) -> StatsCollector {
         StatsCollector { field }
     }
 }
@@ -73,7 +73,7 @@ impl Collector for StatsCollector {
         _segment_local_id: u32,
         segment_reader: &SegmentReader,
     ) -> tantivy::Result<StatsSegmentCollector> {
-        let fast_field_reader = segment_reader.fast_fields().u64(self.field)?;
+        let fast_field_reader = segment_reader.fast_fields().u64(&self.field)?;
         Ok(StatsSegmentCollector {
             fast_field_reader,
             stats: Stats::default(),
@@ -171,7 +171,9 @@ fn main() -> tantivy::Result<()> {
 
     // here we want to get a hit on the 'ken' in Frankenstein
     let query = query_parser.parse_query("broom")?;
-    if let Some(stats) = searcher.search(&query, &StatsCollector::with_field(price))? {
+    if let Some(stats) =
+        searcher.search(&query, &StatsCollector::with_field("price".to_string()))?
+    {
         println!("count: {}", stats.count());
         println!("mean: {}", stats.mean());
         println!("standard deviation: {}", stats.standard_deviation());

examples/integer_range_search.rs

@@ -27,7 +27,7 @@ fn main() -> Result<()> {
     reader.reload()?;
     let searcher = reader.searcher();
     // The end is excluded i.e. here we are searching up to 1969
-    let docs_in_the_sixties = RangeQuery::new_u64(year_field, 1960..1970);
+    let docs_in_the_sixties = RangeQuery::new_u64("year".to_string(), 1960..1970);
     // Uses a Count collector to sum the total number of docs in the range
     let num_60s_books = searcher.search(&docs_in_the_sixties, &Count)?;
     assert_eq!(num_60s_books, 10);

examples/warmer.rs

@@ -4,7 +4,7 @@ use std::sync::{Arc, RwLock, Weak};
 
 use tantivy::collector::TopDocs;
 use tantivy::query::QueryParser;
-use tantivy::schema::{Field, Schema, FAST, TEXT};
+use tantivy::schema::{Schema, FAST, TEXT};
 use tantivy::{
     doc, DocAddress, DocId, Index, IndexReader, Opstamp, Searcher, SearcherGeneration, SegmentId,
     SegmentReader, Warmer,
@@ -25,13 +25,13 @@ pub trait PriceFetcher: Send + Sync + 'static {
 }
 
 struct DynamicPriceColumn {
-    field: Field,
+    field: String,
     price_cache: RwLock<HashMap<(SegmentId, Option<Opstamp>), Arc<Vec<Price>>>>,
     price_fetcher: Box<dyn PriceFetcher>,
 }
 
 impl DynamicPriceColumn {
-    pub fn with_product_id_field<T: PriceFetcher>(field: Field, price_fetcher: T) -> Self {
+    pub fn with_product_id_field<T: PriceFetcher>(field: String, price_fetcher: T) -> Self {
         DynamicPriceColumn {
             field,
             price_cache: Default::default(),
@@ -48,7 +48,7 @@ impl Warmer for DynamicPriceColumn {
     fn warm(&self, searcher: &Searcher) -> tantivy::Result<()> {
         for segment in searcher.segment_readers() {
             let key = (segment.segment_id(), segment.delete_opstamp());
-            let product_id_reader = segment.fast_fields().u64(self.field)?;
+            let product_id_reader = segment.fast_fields().u64(&self.field)?;
             let product_ids: Vec<ProductId> = segment
                 .doc_ids_alive()
                 .map(|doc| product_id_reader.get_val(doc))
@@ -123,7 +123,7 @@ fn main() -> tantivy::Result<()> {
 
     let price_table = ExternalPriceTable::default();
     let price_dynamic_column = Arc::new(DynamicPriceColumn::with_product_id_field(
-        product_id,
+        "product_id".to_string(),
         price_table.clone(),
     ));
     price_table.update_price(OLIVE_OIL, 12);

fastfield_codecs/src/serialize.rs

@@ -402,8 +402,8 @@ mod tests {
         let mut buffer = Vec::new();
         let col = VecColumn::from(&[false, true][..]);
         serialize(col, &mut buffer, &ALL_CODEC_TYPES).unwrap();
-        // 5 bytes of header, 1 byte of value, 7 bytes of padding.
-        assert_eq!(buffer.len(), 3 + 5 + 8 + 4 + 2);
+        // 5 bytes of header, 1 byte of value
+        assert_eq!(buffer.len(), 3 + 5 + 1 + 4 + 2);
     }
 
     #[test]
@@ -411,8 +411,8 @@ mod tests {
         let mut buffer = Vec::new();
         let col = VecColumn::from(&[true][..]);
         serialize(col, &mut buffer, &ALL_CODEC_TYPES).unwrap();
-        // 5 bytes of header, 0 bytes of value, 7 bytes of padding.
-        assert_eq!(buffer.len(), 3 + 5 + 7 + 4 + 2);
+        // 5 bytes of header, 0 bytes of value
+        assert_eq!(buffer.len(), 3 + 5 + 4 + 2);
     }
 
     #[test]
@@ -422,6 +422,6 @@ mod tests {
         let col = VecColumn::from(&vals[..]);
         serialize(col, &mut buffer, &[FastFieldCodecType::Bitpacked]).unwrap();
         // Values are stored over 3 bits.
-        assert_eq!(buffer.len(), 3 + 7 + (3 * 80 / 8) + 7 + 4 + 2);
+        assert_eq!(buffer.len(), 3 + 7 + (3 * 80 / 8) + 4 + 2);
     }
 }

run-tests.sh

@@ -1,2 +0,0 @@
-#!/bin/bash
-cargo test

src/aggregation/agg_req_with_accessor.rs

@@ -94,10 +94,7 @@ impl BucketAggregationWithAccessor {
             BucketAggregationType::Terms(TermsAggregation {
                 field: field_name, ..
             }) => {
-                let field = reader
-                    .schema()
-                    .get_field(field_name)
-                    .ok_or_else(|| TantivyError::FieldNotFound(field_name.to_string()))?;
+                let field = reader.schema().get_field(field_name)?;
                 inverted_index = Some(reader.inverted_index(field)?);
                 get_ff_reader_and_validate(reader, field_name, Cardinality::MultiValues)?
             }
@@ -195,10 +192,7 @@ fn get_ff_reader_and_validate(
     field_name: &str,
     cardinality: Cardinality,
 ) -> crate::Result<(FastFieldAccessor, Type)> {
-    let field = reader
-        .schema()
-        .get_field(field_name)
-        .ok_or_else(|| TantivyError::FieldNotFound(field_name.to_string()))?;
+    let field = reader.schema().get_field(field_name)?;
     let field_type = reader.schema().get_field_entry(field).field_type();
 
     if let Some((_ff_type, field_cardinality)) = type_and_cardinality(field_type) {
@@ -218,10 +212,10 @@ fn get_ff_reader_and_validate(
     let ff_fields = reader.fast_fields();
     match cardinality {
         Cardinality::SingleValue => ff_fields
-            .u64_lenient(field)
+            .u64_lenient(field_name)
             .map(|field| (FastFieldAccessor::Single(field), field_type.value_type())),
         Cardinality::MultiValues => ff_fields
-            .u64s_lenient(field)
+            .u64s_lenient(field_name)
             .map(|field| (FastFieldAccessor::Multi(field), field_type.value_type())),
     }
 }

src/aggregation/bucket/histogram/histogram.rs

@@ -548,9 +548,7 @@ pub(crate) fn intermediate_histogram_buckets_to_final_buckets(
     };
 
     // If we have a date type on the histogram buckets, we add the `key_as_string` field as rfc339
-    let field = schema
-        .get_field(&histogram_req.field)
-        .ok_or_else(|| TantivyError::FieldNotFound(histogram_req.field.to_string()))?;
+    let field = schema.get_field(&histogram_req.field)?;
     if schema.get_field_entry(field).field_type().is_date() {
         for bucket in buckets.iter_mut() {
             if let crate::aggregation::Key::F64(val) = bucket.key {

src/aggregation/bucket/term_agg.rs

@@ -362,13 +362,19 @@ impl SegmentTermCollector {
         let mut entries: Vec<(u32, TermBucketEntry)> =
             self.term_buckets.entries.into_iter().collect();
 
-        let order_by_key = self.req.order.target == OrderTarget::Key;
         let order_by_sub_aggregation =
             matches!(self.req.order.target, OrderTarget::SubAggregation(_));
 
         match self.req.order.target {
             OrderTarget::Key => {
-                // defer order and cut_off after loading the texts from the dictionary
+                // We rely on the fact, that term ordinals match the order of the strings
+                // TODO: We could have a special collector, that keeps only TOP n results at any
+                // time.
+                if self.req.order.order == Order::Desc {
+                    entries.sort_unstable_by_key(|bucket| std::cmp::Reverse(bucket.0));
+                } else {
+                    entries.sort_unstable_by_key(|bucket| bucket.0);
+                }
             }
             OrderTarget::SubAggregation(_name) => {
                 // don't sort and cut off since it's hard to make assumptions on the quality of the
@@ -384,12 +390,11 @@ impl SegmentTermCollector {
             }
         }
 
-        let (term_doc_count_before_cutoff, mut sum_other_doc_count) =
-            if order_by_key || order_by_sub_aggregation {
-                (0, 0)
-            } else {
-                cut_off_buckets(&mut entries, self.req.segment_size as usize)
-            };
+        let (term_doc_count_before_cutoff, sum_other_doc_count) = if order_by_sub_aggregation {
+            (0, 0)
+        } else {
+            cut_off_buckets(&mut entries, self.req.segment_size as usize)
+        };
 
         let inverted_index = agg_with_accessor
             .inverted_index
@@ -412,6 +417,10 @@ impl SegmentTermCollector {
         if self.req.min_doc_count == 0 {
             let mut stream = term_dict.stream()?;
             while let Some((key, _ord)) = stream.next() {
+                if dict.len() >= self.req.segment_size as usize {
+                    break;
+                }
+
                 let key = std::str::from_utf8(key)
                     .map_err(|utf8_err| DataCorruption::comment_only(utf8_err.to_string()))?;
                 if !dict.contains_key(key) {
@@ -420,20 +429,6 @@ impl SegmentTermCollector {
             }
         }
 
-        if order_by_key {
-            let mut dict_entries = dict.into_iter().collect_vec();
-            if self.req.order.order == Order::Desc {
-                dict_entries.sort_unstable_by(|(key1, _), (key2, _)| key1.cmp(key2));
-            } else {
-                dict_entries.sort_unstable_by(|(key1, _), (key2, _)| key2.cmp(key1));
-            }
-            let (_, sum_other_docs) =
-                cut_off_buckets(&mut dict_entries, self.req.segment_size as usize);
-
-            sum_other_doc_count += sum_other_docs;
-            dict = dict_entries.into_iter().collect();
-        }
-
         Ok(IntermediateBucketResult::Terms(
             IntermediateTermBucketResult {
                 entries: dict,
@@ -923,14 +918,14 @@ mod tests {
         ];
         let index = get_test_index_from_values_and_terms(merge_segments, &segment_and_terms)?;
 
-        // key desc
+        // key asc
         let agg_req: Aggregations = vec![(
             "my_texts".to_string(),
             Aggregation::Bucket(BucketAggregation {
                 bucket_agg: BucketAggregationType::Terms(TermsAggregation {
                     field: "string_id".to_string(),
                     order: Some(CustomOrder {
-                        order: Order::Desc,
+                        order: Order::Asc,
                         target: OrderTarget::Key,
                     }),
                     ..Default::default()
@@ -957,7 +952,7 @@ mod tests {
                 bucket_agg: BucketAggregationType::Terms(TermsAggregation {
                     field: "string_id".to_string(),
                     order: Some(CustomOrder {
-                        order: Order::Desc,
+                        order: Order::Asc,
                         target: OrderTarget::Key,
                     }),
                     size: Some(2),
@@ -981,14 +976,14 @@ mod tests {
 
         assert_eq!(res["my_texts"]["sum_other_doc_count"], 3);
 
-        // key desc and segment_size cut_off
+        // key asc and segment_size cut_off
         let agg_req: Aggregations = vec![(
             "my_texts".to_string(),
             Aggregation::Bucket(BucketAggregation {
                 bucket_agg: BucketAggregationType::Terms(TermsAggregation {
                     field: "string_id".to_string(),
                     order: Some(CustomOrder {
-                        order: Order::Desc,
+                        order: Order::Asc,
                         target: OrderTarget::Key,
                     }),
                     size: Some(2),
@@ -1011,14 +1006,14 @@ mod tests {
             serde_json::Value::Null
         );
 
-        // key asc
+        // key desc
         let agg_req: Aggregations = vec![(
             "my_texts".to_string(),
             Aggregation::Bucket(BucketAggregation {
                 bucket_agg: BucketAggregationType::Terms(TermsAggregation {
                     field: "string_id".to_string(),
                     order: Some(CustomOrder {
-                        order: Order::Asc,
+                        order: Order::Desc,
                         target: OrderTarget::Key,
                     }),
                     ..Default::default()
@@ -1038,14 +1033,14 @@ mod tests {
         assert_eq!(res["my_texts"]["buckets"][2]["doc_count"], 5);
         assert_eq!(res["my_texts"]["sum_other_doc_count"], 0);
 
-        // key asc, size cut_off
+        // key desc, size cut_off
         let agg_req: Aggregations = vec![(
             "my_texts".to_string(),
             Aggregation::Bucket(BucketAggregation {
                 bucket_agg: BucketAggregationType::Terms(TermsAggregation {
                     field: "string_id".to_string(),
                     order: Some(CustomOrder {
-                        order: Order::Asc,
+                        order: Order::Desc,
                         target: OrderTarget::Key,
                     }),
                     size: Some(2),
@@ -1068,14 +1063,14 @@ mod tests {
         );
         assert_eq!(res["my_texts"]["sum_other_doc_count"], 5);
 
-        // key asc, segment_size cut_off
+        // key desc, segment_size cut_off
         let agg_req: Aggregations = vec![(
             "my_texts".to_string(),
             Aggregation::Bucket(BucketAggregation {
                 bucket_agg: BucketAggregationType::Terms(TermsAggregation {
                     field: "string_id".to_string(),
                     order: Some(CustomOrder {
-                        order: Order::Asc,
+                        order: Order::Desc,
                         target: OrderTarget::Key,
                     }),
                     size: Some(2),
@@ -1352,68 +1347,3 @@ mod tests {
         Ok(())
     }
 }
-
-#[cfg(all(test, feature = "unstable"))]
-mod bench {
-
-    use itertools::Itertools;
-    use rand::seq::SliceRandom;
-    use rand::thread_rng;
-
-    use super::*;
-
-    fn get_collector_with_buckets(num_docs: u64) -> TermBuckets {
-        TermBuckets::from_req_and_validate(&Default::default(), num_docs as usize).unwrap()
-    }
-
-    fn get_rand_terms(total_terms: u64, num_terms_returned: u64) -> Vec<u64> {
-        let mut rng = thread_rng();
-
-        let all_terms = (0..total_terms - 1).collect_vec();
-
-        let mut vals = vec![];
-        for _ in 0..num_terms_returned {
-            let val = all_terms.as_slice().choose(&mut rng).unwrap();
-            vals.push(*val);
-        }
-
-        vals
-    }
-
-    fn bench_term_buckets(b: &mut test::Bencher, num_terms: u64, total_terms: u64) {
-        let mut collector = get_collector_with_buckets(total_terms);
-        let vals = get_rand_terms(total_terms, num_terms);
-        let aggregations_with_accessor: AggregationsWithAccessor = Default::default();
-        let bucket_count: BucketCount = BucketCount {
-            bucket_count: Default::default(),
-            max_bucket_count: 1_000_001u32,
-        };
-        b.iter(|| {
-            for &val in &vals {
-                collector
-                    .increment_bucket(&[val], 0, &aggregations_with_accessor, &bucket_count, &None)
-                    .unwrap();
-            }
-        })
-    }
-
-    #[bench]
-    fn bench_term_buckets_500_of_1_000_000(b: &mut test::Bencher) {
-        bench_term_buckets(b, 500u64, 1_000_000u64)
-    }
-
-    #[bench]
-    fn bench_term_buckets_1_000_000_of_50_000(b: &mut test::Bencher) {
-        bench_term_buckets(b, 1_000_000u64, 50_000u64)
-    }
-
-    #[bench]
-    fn bench_term_buckets_1_000_000_of_50(b: &mut test::Bencher) {
-        bench_term_buckets(b, 1_000_000u64, 50u64)
-    }
-
-    #[bench]
-    fn bench_term_buckets_1_000_000_of_1_000_000(b: &mut test::Bencher) {
-        bench_term_buckets(b, 1_000_000u64, 1_000_000u64)
-    }
-}

src/aggregation/intermediate_agg_result.rs

@@ -26,7 +26,6 @@ use super::{format_date, Key, SerializedKey, VecWithNames};
 use crate::aggregation::agg_result::{AggregationResults, BucketEntries, BucketEntry};
 use crate::aggregation::bucket::TermsAggregationInternal;
 use crate::schema::Schema;
-use crate::TantivyError;
 
 /// Contains the intermediate aggregation result, which is optimized to be merged with other
 /// intermediate results.
@@ -500,7 +499,7 @@ impl IntermediateTermBucketResult {
         match req.order.target {
             OrderTarget::Key => {
                 buckets.sort_by(|left, right| {
-                    if req.order.order == Order::Desc {
+                    if req.order.order == Order::Asc {
                         left.key.partial_cmp(&right.key)
                     } else {
                         right.key.partial_cmp(&left.key)
@@ -658,9 +657,7 @@ impl IntermediateRangeBucketEntry {
 
         // If we have a date type on the histogram buckets, we add the `key_as_string` field as
         // rfc339
-        let field = schema
-            .get_field(&range_req.field)
-            .ok_or_else(|| TantivyError::FieldNotFound(range_req.field.to_string()))?;
+        let field = schema.get_field(&range_req.field)?;
         if schema.get_field_entry(field).field_type().is_date() {
             if let Some(val) = range_bucket_entry.to {
                 let key_as_string = format_date(val as i64)?;

src/aggregation/metric/average.rs

@@ -6,14 +6,13 @@ use super::{IntermediateStats, SegmentStatsCollector};
 
 /// A single-value metric aggregation that computes the average of numeric values that are
 /// extracted from the aggregated documents.
-/// Supported field types are u64, i64, and f64.
 /// See [super::SingleMetricResult] for return value.
 ///
 /// # JSON Format
 /// ```json
 /// {
 ///     "avg": {
-///         "field": "score",
+///         "field": "score"
 ///     }
 /// }
 /// ```

src/aggregation/metric/count.rs

@@ -6,14 +6,13 @@ use super::{IntermediateStats, SegmentStatsCollector};
 
 /// A single-value metric aggregation that counts the number of values that are
 /// extracted from the aggregated documents.
-/// Supported field types are u64, i64, and f64.
 /// See [super::SingleMetricResult] for return value.
 ///
 /// # JSON Format
 /// ```json
 /// {
 ///     "value_count": {
-///         "field": "score",
+///         "field": "score"
 ///     }
 /// }
 /// ```

src/aggregation/metric/max.rs

@@ -6,14 +6,13 @@ use super::{IntermediateStats, SegmentStatsCollector};
 
 /// A single-value metric aggregation that computes the maximum of numeric values that are
 /// extracted from the aggregated documents.
-/// Supported field types are u64, i64, and f64.
 /// See [super::SingleMetricResult] for return value.
 ///
 /// # JSON Format
 /// ```json
 /// {
 ///     "max": {
-///         "field": "score",
+///         "field": "score"
 ///     }
 /// }
 /// ```

src/aggregation/metric/min.rs

@@ -6,14 +6,13 @@ use super::{IntermediateStats, SegmentStatsCollector};
 
 /// A single-value metric aggregation that computes the minimum of numeric values that are
 /// extracted from the aggregated documents.
-/// Supported field types are u64, i64, and f64.
 /// See [super::SingleMetricResult] for return value.
 ///
 /// # JSON Format
 /// ```json
 /// {
 ///     "min": {
-///         "field": "score",
+///         "field": "score"
 ///     }
 /// }
 /// ```

src/aggregation/metric/mod.rs

@@ -80,12 +80,12 @@ mod tests {
             "price_stats": { "stats": { "field": "price" } },
             "price_sum": { "sum": { "field": "price" } }
         }"#;
-        let aggregations: Aggregations = serde_json::from_str(&aggregations_json).unwrap();
+        let aggregations: Aggregations = serde_json::from_str(aggregations_json).unwrap();
         let collector = AggregationCollector::from_aggs(aggregations, None, index.schema());
         let reader = index.reader().unwrap();
         let searcher = reader.searcher();
         let aggregations_res: AggregationResults = searcher.search(&AllQuery, &collector).unwrap();
-        let aggregations_res_json = serde_json::to_value(&aggregations_res).unwrap();
+        let aggregations_res_json = serde_json::to_value(aggregations_res).unwrap();
 
         assert_eq!(aggregations_res_json["price_avg"]["value"], 2.5);
         assert_eq!(aggregations_res_json["price_count"]["value"], 6.0);

src/aggregation/metric/stats.rs

@@ -7,14 +7,13 @@ use crate::{DocId, TantivyError};
 
 /// A multi-value metric aggregation that computes a collection of statistics on numeric values that
 /// are extracted from the aggregated documents.
-/// Supported field types are `u64`, `i64`, and `f64`.
 /// See [`Stats`] for returned statistics.
 ///
 /// # JSON Format
 /// ```json
 /// {
 ///     "stats": {
-///         "field": "score",
+///         "field": "score"
 ///     }
 ///  }
 /// ```

src/aggregation/metric/sum.rs

@@ -6,14 +6,13 @@ use super::{IntermediateStats, SegmentStatsCollector};
 
 /// A single-value metric aggregation that sums up numeric values that are
 /// extracted from the aggregated documents.
-/// Supported field types are u64, i64, and f64.
 /// See [super::SingleMetricResult] for return value.
 ///
 /// # JSON Format
 /// ```json
 /// {
 ///     "sum": {
-///         "field": "score",
+///         "field": "score"
 ///     }
 /// }
 /// ```

src/aggregation/mod.rs

@@ -1,6 +1,5 @@
 //! # Aggregations
 //!
-//!
 //! An aggregation summarizes your data as statistics on buckets or metrics.
 //!
 //! Aggregations can provide answer to questions like:
@@ -41,6 +40,10 @@
 //! - [Metric](metric)
 //!     - [Average](metric::AverageAggregation)
 //!     - [Stats](metric::StatsAggregation)
+//!     - [Min](metric::MinAggregation)
+//!     - [Max](metric::MaxAggregation)
+//!     - [Sum](metric::SumAggregation)
+//!     - [Count](metric::CountAggregation)
 //!
 //! # Example
 //! Compute the average metric, by building [`agg_req::Aggregations`], which is built from an
@@ -75,7 +78,7 @@
 //! }
 //! ```
 //! # Example JSON
-//! Requests are compatible with the elasticsearch json request format.
+//! Requests are compatible with the elasticsearch JSON request format.
 //!
 //! ```
 //! use tantivy::aggregation::agg_req::Aggregations;
@@ -1153,12 +1156,6 @@ mod tests {
             r#"FieldNotFound("not_exist_field")"#
         );
 
-        let agg_res = avg_on_field("scores_i64");
-        assert_eq!(
-            format!("{:?}", agg_res),
-            r#"InvalidArgument("Invalid field cardinality on field scores_i64 expected SingleValue, but got MultiValues")"#
-        );
-
         Ok(())
     }
 

src/collector/filter_collector_wrapper.rs

@@ -130,7 +130,7 @@ where
 
         let fast_field_reader = segment_reader
             .fast_fields()
-            .typed_fast_field_reader(self.field)?;
+            .typed_fast_field_reader(schema.get_field_name(self.field))?;
 
         let segment_collector = self
             .collector

src/collector/histogram_collector.rs

@@ -5,7 +5,7 @@ use fastfield_codecs::Column;
 
 use crate::collector::{Collector, SegmentCollector};
 use crate::fastfield::FastValue;
-use crate::schema::{Field, Type};
+use crate::schema::Type;
 use crate::{DocId, Score};
 
 /// Histogram builds an histogram of the values of a fastfield for the
@@ -28,7 +28,7 @@ pub struct HistogramCollector {
     min_value: u64,
     num_buckets: usize,
     divider: DividerU64,
-    field: Field,
+    field: String,
 }
 
 impl HistogramCollector {
@@ -46,7 +46,7 @@ impl HistogramCollector {
     /// # Disclaimer
     /// This function panics if the field given is of type f64.
     pub fn new<TFastValue: FastValue>(
-        field: Field,
+        field: String,
         min_value: TFastValue,
         bucket_width: u64,
         num_buckets: usize,
@@ -112,7 +112,7 @@ impl Collector for HistogramCollector {
         _segment_local_id: crate::SegmentOrdinal,
         segment: &crate::SegmentReader,
     ) -> crate::Result<Self::Child> {
-        let ff_reader = segment.fast_fields().u64_lenient(self.field)?;
+        let ff_reader = segment.fast_fields().u64_lenient(&self.field)?;
         Ok(SegmentHistogramCollector {
             histogram_computer: HistogramComputer {
                 counts: vec![0; self.num_buckets],
@@ -211,13 +211,13 @@ mod tests {
     #[test]
     fn test_no_segments() -> crate::Result<()> {
         let mut schema_builder = Schema::builder();
-        let val_field = schema_builder.add_u64_field("val_field", FAST);
+        schema_builder.add_u64_field("val_field", FAST);
         let schema = schema_builder.build();
         let index = Index::create_in_ram(schema);
         let reader = index.reader()?;
         let searcher = reader.searcher();
         let all_query = AllQuery;
-        let histogram_collector = HistogramCollector::new(val_field, 0u64, 2, 5);
+        let histogram_collector = HistogramCollector::new("val_field".to_string(), 0u64, 2, 5);
         let histogram = searcher.search(&all_query, &histogram_collector)?;
         assert_eq!(histogram, vec![0; 5]);
         Ok(())
@@ -238,7 +238,8 @@ mod tests {
         let reader = index.reader()?;
         let searcher = reader.searcher();
         let all_query = AllQuery;
-        let histogram_collector = HistogramCollector::new(val_field, -20i64, 10u64, 4);
+        let histogram_collector =
+            HistogramCollector::new("val_field".to_string(), -20i64, 10u64, 4);
         let histogram = searcher.search(&all_query, &histogram_collector)?;
         assert_eq!(histogram, vec![1, 1, 0, 1]);
         Ok(())
@@ -262,7 +263,8 @@ mod tests {
         let reader = index.reader()?;
         let searcher = reader.searcher();
         let all_query = AllQuery;
-        let histogram_collector = HistogramCollector::new(val_field, -20i64, 10u64, 4);
+        let histogram_collector =
+            HistogramCollector::new("val_field".to_string(), -20i64, 10u64, 4);
         let histogram = searcher.search(&all_query, &histogram_collector)?;
         assert_eq!(histogram, vec![1, 1, 0, 1]);
         Ok(())
@@ -285,7 +287,7 @@ mod tests {
         let searcher = reader.searcher();
         let all_query = AllQuery;
         let week_histogram_collector = HistogramCollector::new(
-            date_field,
+            "date_field".to_string(),
             DateTime::from_primitive(
                 Date::from_calendar_date(1980, Month::January, 1)?.with_hms(0, 0, 0)?,
             ),

src/collector/tests.rs

@@ -155,7 +155,7 @@ impl SegmentCollector for TestSegmentCollector {
 ///
 /// This collector is mainly useful for tests.
 pub struct FastFieldTestCollector {
-    field: Field,
+    field: String,
 }
 
 pub struct FastFieldSegmentCollector {
@@ -164,7 +164,7 @@ pub struct FastFieldSegmentCollector {
 }
 
 impl FastFieldTestCollector {
-    pub fn for_field(field: Field) -> FastFieldTestCollector {
+    pub fn for_field(field: String) -> FastFieldTestCollector {
         FastFieldTestCollector { field }
     }
 }
@@ -180,7 +180,7 @@ impl Collector for FastFieldTestCollector {
     ) -> crate::Result<FastFieldSegmentCollector> {
         let reader = segment_reader
             .fast_fields()
-            .u64(self.field)
+            .u64(&self.field)
             .expect("Requested field is not a fast field.");
         Ok(FastFieldSegmentCollector {
             vals: Vec::new(),
@@ -238,7 +238,9 @@ impl Collector for BytesFastFieldTestCollector {
         _segment_local_id: u32,
         segment_reader: &SegmentReader,
     ) -> crate::Result<BytesFastFieldSegmentCollector> {
-        let reader = segment_reader.fast_fields().bytes(self.field)?;
+        let reader = segment_reader
+            .fast_fields()
+            .bytes(segment_reader.schema().get_field_name(self.field))?;
         Ok(BytesFastFieldSegmentCollector {
             vals: Vec::new(),
             reader,

src/collector/top_score_collector.rs

@@ -156,7 +156,7 @@ impl CustomScorer<u64> for ScorerByField {
         // The conversion will then happen only on the top-K docs.
         let ff_reader = segment_reader
             .fast_fields()
-            .typed_fast_field_reader(self.field)?;
+            .typed_fast_field_reader(segment_reader.schema().get_field_name(self.field))?;
         Ok(ScorerByFastFieldReader { ff_reader })
     }
 }
@@ -454,7 +454,7 @@ impl TopDocs {
     ///             // In our case, we will get a reader for the popularity
     ///             // fast field.
     ///             let popularity_reader =
-    ///                 segment_reader.fast_fields().u64(popularity).unwrap();
+    ///                 segment_reader.fast_fields().u64("popularity").unwrap();
     ///
     ///             // We can now define our actual scoring function
     ///             move |doc: DocId, original_score: Score| {
@@ -561,9 +561,9 @@ impl TopDocs {
     ///             // Note that this is implemented by using a `(u64, u64)`
     ///             // as a score.
     ///             let popularity_reader =
-    ///                 segment_reader.fast_fields().u64(popularity).unwrap();
+    ///                 segment_reader.fast_fields().u64("popularity").unwrap();
     ///             let boosted_reader =
-    ///                 segment_reader.fast_fields().u64(boosted).unwrap();
+    ///                 segment_reader.fast_fields().u64("boosted").unwrap();
     ///
     ///             // We can now define our actual scoring function
     ///             move |doc: DocId| {

src/core/index.rs

@@ -231,7 +231,7 @@ impl IndexBuilder {
     fn validate(&self) -> crate::Result<()> {
         if let Some(schema) = self.schema.as_ref() {
             if let Some(sort_by_field) = self.index_settings.sort_by_field.as_ref() {
-                let schema_field = schema.get_field(&sort_by_field.field).ok_or_else(|| {
+                let schema_field = schema.get_field(&sort_by_field.field).map_err(|_| {
                     TantivyError::InvalidArgument(format!(
                         "Field to sort index {} not found in schema",
                         sort_by_field.field

src/core/inverted_index_reader.rs

@@ -135,6 +135,8 @@ impl InvertedIndexReader {
         term_info: &TermInfo,
         option: IndexRecordOption,
     ) -> io::Result<SegmentPostings> {
+        let option = option.downgrade(self.record_option);
+
         let block_postings = self.read_block_postings_from_terminfo(term_info, option)?;
         let position_reader = {
             if option.has_positions() {

src/core/searcher.rs

@@ -249,7 +249,7 @@ impl SearcherInner {
         index: Index,
         segment_readers: Vec<SegmentReader>,
         generation: TrackedObject<SearcherGeneration>,
-        doc_store_cache_size: usize,
+        doc_store_cache_num_blocks: usize,
     ) -> io::Result<SearcherInner> {
         assert_eq!(
             &segment_readers
@@ -261,7 +261,7 @@ impl SearcherInner {
         );
         let store_readers: Vec<StoreReader> = segment_readers
             .iter()
-            .map(|segment_reader| segment_reader.get_store_reader(doc_store_cache_size))
+            .map(|segment_reader| segment_reader.get_store_reader(doc_store_cache_num_blocks))
             .collect::<io::Result<Vec<_>>>()?;
 
         Ok(SearcherInner {

src/core/segment_reader.rs

@@ -95,7 +95,8 @@ impl SegmentReader {
 
         match field_entry.field_type() {
             FieldType::Facet(_) => {
-                let term_ords_reader = self.fast_fields().u64s(field)?;
+                let term_ords_reader =
+                    self.fast_fields().u64s(self.schema.get_field_name(field))?;
                 let termdict = self
                     .termdict_composite
                     .open_read(field)
@@ -133,9 +134,12 @@ impl SegmentReader {
         &self.fieldnorm_readers
     }
 
-    /// Accessor to the segment's `StoreReader`.
-    pub fn get_store_reader(&self, cache_size: usize) -> io::Result<StoreReader> {
-        StoreReader::open(self.store_file.clone(), cache_size)
+    /// Accessor to the segment's [`StoreReader`](crate::store::StoreReader).
+    ///
+    /// `cache_num_blocks` sets the number of decompressed blocks to be cached in an LRU.
+    /// The size of blocks is configurable, this should be reflexted in the
+    pub fn get_store_reader(&self, cache_num_blocks: usize) -> io::Result<StoreReader> {
+        StoreReader::open(self.store_file.clone(), cache_num_blocks)
     }
 
     /// Open a new segment for reading.

src/fastfield/bytes/mod.rs

@@ -25,7 +25,7 @@ mod tests {
         index_writer.commit()?;
         let searcher = index.reader()?.searcher();
         let segment_reader = searcher.segment_reader(0);
-        let bytes_reader = segment_reader.fast_fields().bytes(bytes_field).unwrap();
+        let bytes_reader = segment_reader.fast_fields().bytes("bytesfield").unwrap();
         assert_eq!(bytes_reader.get_bytes(0), &[0u8, 1, 2, 3]);
         assert!(bytes_reader.get_bytes(1).is_empty());
         assert_eq!(bytes_reader.get_bytes(2), &[255u8]);
@@ -109,8 +109,7 @@ mod tests {
         let searcher = create_index_for_test(FAST)?;
         assert_eq!(searcher.num_docs(), 1);
         let fast_fields = searcher.segment_reader(0u32).fast_fields();
-        let field = searcher.schema().get_field("string_bytes").unwrap();
-        let fast_field_reader = fast_fields.bytes(field).unwrap();
+        let fast_field_reader = fast_fields.bytes("string_bytes").unwrap();
         assert_eq!(fast_field_reader.get_bytes(0u32), b"tantivy");
         Ok(())
     }

src/fastfield/mod.rs

@@ -226,7 +226,7 @@ mod tests {
             serializer.close().unwrap();
         }
         let file = directory.open_read(path).unwrap();
-        assert_eq!(file.len(), 34);
+        assert_eq!(file.len(), 27);
         let composite_file = CompositeFile::open(&file)?;
         let fast_field_bytes = composite_file.open_read(*FIELD).unwrap().read_bytes()?;
         let fast_field_reader = open::<u64>(fast_field_bytes)?;
@@ -275,7 +275,7 @@ mod tests {
             serializer.close()?;
         }
         let file = directory.open_read(path)?;
-        assert_eq!(file.len(), 62);
+        assert_eq!(file.len(), 55);
         {
             let fast_fields_composite = CompositeFile::open(&file)?;
             let data = fast_fields_composite
@@ -316,7 +316,7 @@ mod tests {
             serializer.close().unwrap();
         }
         let file = directory.open_read(path).unwrap();
-        assert_eq!(file.len(), 35);
+        assert_eq!(file.len(), 28);
         {
             let fast_fields_composite = CompositeFile::open(&file).unwrap();
             let data = fast_fields_composite
@@ -355,7 +355,7 @@ mod tests {
             serializer.close().unwrap();
         }
         let file = directory.open_read(path).unwrap();
-        assert_eq!(file.len(), 80049);
+        assert_eq!(file.len(), 80042);
         {
             let fast_fields_composite = CompositeFile::open(&file)?;
             let data = fast_fields_composite
@@ -397,7 +397,7 @@ mod tests {
             serializer.close().unwrap();
         }
         let file = directory.open_read(path).unwrap();
-        assert_eq!(file.len(), 49_usize);
+        assert_eq!(file.len(), 42_usize);
 
         {
             let fast_fields_composite = CompositeFile::open(&file)?;
@@ -583,7 +583,7 @@ mod tests {
             assert_eq!(searcher.segment_readers().len(), 1);
             let segment_reader = searcher.segment_reader(0);
             let fast_fields = segment_reader.fast_fields();
-            let text_fast_field = fast_fields.u64s(text_field).unwrap();
+            let text_fast_field = fast_fields.u64s("text").unwrap();
 
             assert_eq!(
                 get_vals_for_docs(&text_fast_field, 0..5),
@@ -622,7 +622,7 @@ mod tests {
             assert_eq!(searcher.segment_readers().len(), 2);
             let segment_reader = searcher.segment_reader(1);
             let fast_fields = segment_reader.fast_fields();
-            let text_fast_field = fast_fields.u64s(text_field).unwrap();
+            let text_fast_field = fast_fields.u64s("text").unwrap();
 
             assert_eq!(get_vals_for_docs(&text_fast_field, 0..3), vec![0, 1, 0]);
         }
@@ -638,7 +638,7 @@ mod tests {
         let searcher = reader.searcher();
         let segment_reader = searcher.segment_reader(0);
         let fast_fields = segment_reader.fast_fields();
-        let text_fast_field = fast_fields.u64s(text_field).unwrap();
+        let text_fast_field = fast_fields.u64s("text").unwrap();
 
         assert_eq!(
             get_vals_for_docs(&text_fast_field, 0..8),
@@ -681,7 +681,7 @@ mod tests {
             assert_eq!(searcher.segment_readers().len(), 1);
             let segment_reader = searcher.segment_reader(0);
             let fast_fields = segment_reader.fast_fields();
-            let text_fast_field = fast_fields.u64s(text_field).unwrap();
+            let text_fast_field = fast_fields.u64s("text").unwrap();
 
             assert_eq!(get_vals_for_docs(&text_fast_field, 0..6), vec![1, 0, 0, 2]);
 
@@ -712,7 +712,7 @@ mod tests {
             assert_eq!(searcher.segment_readers().len(), 2);
             let segment_reader = searcher.segment_reader(1);
             let fast_fields = segment_reader.fast_fields();
-            let text_fast_field = fast_fields.u64s(text_field).unwrap();
+            let text_fast_field = fast_fields.u64s("text").unwrap();
 
             assert_eq!(get_vals_for_docs(&text_fast_field, 0..2), vec![0, 1]);
         }
@@ -728,7 +728,7 @@ mod tests {
         let searcher = reader.searcher();
         let segment_reader = searcher.segment_reader(0);
         let fast_fields = segment_reader.fast_fields();
-        let text_fast_field = fast_fields.u64s(text_field).unwrap();
+        let text_fast_field = fast_fields.u64s("text").unwrap();
 
         assert_eq!(
             get_vals_for_docs(&text_fast_field, 0..9),
@@ -773,8 +773,8 @@ mod tests {
         assert_eq!(searcher.segment_readers().len(), 1);
         let segment_reader = searcher.segment_reader(0);
         let fast_fields = segment_reader.fast_fields();
-        let date_fast_field = fast_fields.date(date_field).unwrap();
-        let dates_fast_field = fast_fields.dates(multi_date_field).unwrap();
+        let date_fast_field = fast_fields.date("date").unwrap();
+        let dates_fast_field = fast_fields.dates("multi_date").unwrap();
         let mut dates = vec![];
         {
             assert_eq!(date_fast_field.get_val(0).into_timestamp_micros(), 1i64);
@@ -836,7 +836,7 @@ mod tests {
             serializer.close().unwrap();
         }
         let file = directory.open_read(path).unwrap();
-        assert_eq!(file.len(), 33);
+        assert_eq!(file.len(), 26);
         let composite_file = CompositeFile::open(&file)?;
         let data = composite_file.open_read(field).unwrap().read_bytes()?;
         let fast_field_reader = open::<bool>(data)?;
@@ -874,7 +874,7 @@ mod tests {
             serializer.close().unwrap();
         }
         let file = directory.open_read(path).unwrap();
-        assert_eq!(file.len(), 45);
+        assert_eq!(file.len(), 38);
         let composite_file = CompositeFile::open(&file)?;
         let data = composite_file.open_read(field).unwrap().read_bytes()?;
         let fast_field_reader = open::<bool>(data)?;
@@ -906,7 +906,7 @@ mod tests {
         }
         let file = directory.open_read(path).unwrap();
         let composite_file = CompositeFile::open(&file)?;
-        assert_eq!(file.len(), 32);
+        assert_eq!(file.len(), 25);
         let data = composite_file.open_read(field).unwrap().read_bytes()?;
         let fast_field_reader = open::<bool>(data)?;
         assert_eq!(fast_field_reader.get_val(0), false);
@@ -940,10 +940,10 @@ mod tests {
     pub fn test_gcd_date() -> crate::Result<()> {
         let size_prec_sec =
             test_gcd_date_with_codec(FastFieldCodecType::Bitpacked, DatePrecision::Seconds)?;
-        assert_eq!(size_prec_sec, 5 + 4 + 28 + (1_000 * 13) / 8); // 13 bits per val = ceil(log_2(number of seconds in 2hours);
+        assert_eq!(size_prec_sec, 5 + 4 + 21 + (1_000 * 13) / 8); // 13 bits per val = ceil(log_2(number of seconds in 2hours);
         let size_prec_micro =
             test_gcd_date_with_codec(FastFieldCodecType::Bitpacked, DatePrecision::Microseconds)?;
-        assert_eq!(size_prec_micro, 5 + 4 + 26 + (1_000 * 33) / 8); // 33 bits per val = ceil(log_2(number of microsecsseconds in 2hours);
+        assert_eq!(size_prec_micro, 5 + 4 + 19 + (1_000 * 33) / 8); // 33 bits per val = ceil(log_2(number of microsecsseconds in 2hours);
         Ok(())
     }
 
@@ -1014,7 +1014,7 @@ mod tests {
         let reader = index.reader().unwrap();
         let searcher = reader.searcher();
         let segment = &searcher.segment_readers()[0];
-        let field = segment.fast_fields().u64(num_field).unwrap();
+        let field = segment.fast_fields().u64("url_norm_hash").unwrap();
 
         let numbers = vec![100, 200, 300];
         let test_range = |range: RangeInclusive<u64>| {
@@ -1063,7 +1063,7 @@ mod tests {
         let reader = index.reader().unwrap();
         let searcher = reader.searcher();
         let segment = &searcher.segment_readers()[0];
-        let field = segment.fast_fields().u64(num_field).unwrap();
+        let field = segment.fast_fields().u64("url_norm_hash").unwrap();
 
         let numbers = vec![1000, 1001, 1003];
         let test_range = |range: RangeInclusive<u64>| {

src/fastfield/multivalued/mod.rs

@@ -52,7 +52,7 @@ mod tests {
         let searcher = index.reader()?.searcher();
         let segment_reader = searcher.segment_reader(0);
         let mut vals = Vec::new();
-        let multi_value_reader = segment_reader.fast_fields().u64s(field)?;
+        let multi_value_reader = segment_reader.fast_fields().u64s("multifield")?;
         {
             multi_value_reader.get_vals(2, &mut vals);
             assert_eq!(&vals, &[4u64]);
@@ -229,7 +229,7 @@ mod tests {
         let searcher = index.reader()?.searcher();
         let segment_reader = searcher.segment_reader(0);
         let mut vals = Vec::new();
-        let multi_value_reader = segment_reader.fast_fields().i64s(field).unwrap();
+        let multi_value_reader = segment_reader.fast_fields().i64s("multifield").unwrap();
         multi_value_reader.get_vals(2, &mut vals);
         assert_eq!(&vals, &[-4i64]);
         multi_value_reader.get_vals(0, &mut vals);
@@ -261,7 +261,7 @@ mod tests {
         let searcher = index.reader()?.searcher();
         let segment_reader = searcher.segment_reader(0);
         let mut vals = Vec::new();
-        let multi_value_reader = segment_reader.fast_fields().bools(bool_field).unwrap();
+        let multi_value_reader = segment_reader.fast_fields().bools("multifield").unwrap();
         multi_value_reader.get_vals(2, &mut vals);
         assert_eq!(&vals, &[false]);
         multi_value_reader.get_vals(0, &mut vals);

src/fastfield/multivalued/reader.rs

@@ -159,7 +159,7 @@ mod tests {
         let searcher = reader.searcher();
         let reader = searcher.segment_reader(0);
 
-        let date_ff_reader = reader.fast_fields().dates(date_field).unwrap();
+        let date_ff_reader = reader.fast_fields().dates("multi_date_field").unwrap();
         let mut docids = vec![];
         date_ff_reader.get_docids_for_value_range(
             DateTime::from_utc(first_time_stamp)..=DateTime::from_utc(two_secs_ahead),
@@ -173,7 +173,7 @@ mod tests {
 
         assert_eq!(
             count_multiples(RangeQuery::new_date(
-                date_field,
+                "multi_date_field".to_string(),
                 DateTime::from_utc(first_time_stamp)..DateTime::from_utc(two_secs_ahead)
             )),
             1
@@ -226,7 +226,7 @@ mod tests {
         let reader = searcher.segment_reader(0);
         assert_eq!(reader.num_docs(), 5);
 
-        let date_ff_reader = reader.fast_fields().dates(date_field).unwrap();
+        let date_ff_reader = reader.fast_fields().dates("multi_date_field").unwrap();
         let mut docids = vec![];
         date_ff_reader.get_docids_for_value_range(
             DateTime::from_utc(first_time_stamp)..=DateTime::from_utc(two_secs_ahead),
@@ -240,7 +240,7 @@ mod tests {
 
         assert_eq!(
             count_multiples(RangeQuery::new_date(
-                date_field,
+                "multi_date_field".to_string(),
                 DateTime::from_utc(first_time_stamp)..DateTime::from_utc(two_secs_ahead)
             )),
             2
@@ -324,7 +324,7 @@ mod tests {
         index_writer.commit()?;
         let searcher = index.reader()?.searcher();
         let segment_reader = searcher.segment_reader(0);
-        let field_reader = segment_reader.fast_fields().i64s(item_field)?;
+        let field_reader = segment_reader.fast_fields().i64s("items")?;
 
         assert_eq!(field_reader.min_value(), -2);
         assert_eq!(field_reader.max_value(), 6);

src/fastfield/readers.rs

@@ -114,9 +114,11 @@ impl FastFieldReaders {
 
     pub(crate) fn typed_fast_field_reader_with_idx<TFastValue: FastValue>(
         &self,
-        field: Field,
+        field_name: &str,
         index: usize,
     ) -> crate::Result<Arc<dyn Column<TFastValue>>> {
+        let field = self.schema.get_field(field_name)?;
+
         let fast_field_slice = self.fast_field_data(field, index)?;
         let bytes = fast_field_slice.read_bytes()?;
         let column = fastfield_codecs::open(bytes)?;
@@ -125,32 +127,37 @@ impl FastFieldReaders {
 
     pub(crate) fn typed_fast_field_reader<TFastValue: FastValue>(
         &self,
-        field: Field,
+        field_name: &str,
     ) -> crate::Result<Arc<dyn Column<TFastValue>>> {
-        self.typed_fast_field_reader_with_idx(field, 0)
+        self.typed_fast_field_reader_with_idx(field_name, 0)
     }
 
     pub(crate) fn typed_fast_field_multi_reader<TFastValue: FastValue>(
         &self,
-        field: Field,
+        field_name: &str,
     ) -> crate::Result<MultiValuedFastFieldReader<TFastValue>> {
-        let idx_reader = self.typed_fast_field_reader(field)?;
-        let vals_reader = self.typed_fast_field_reader_with_idx(field, 1)?;
+        let idx_reader = self.typed_fast_field_reader(field_name)?;
+        let vals_reader = self.typed_fast_field_reader_with_idx(field_name, 1)?;
         Ok(MultiValuedFastFieldReader::open(idx_reader, vals_reader))
     }
 
     /// Returns the `u64` fast field reader reader associated with `field`.
     ///
     /// If `field` is not a u64 fast field, this method returns an Error.
-    pub fn u64(&self, field: Field) -> crate::Result<Arc<dyn Column<u64>>> {
-        self.check_type(field, FastType::U64, Cardinality::SingleValue)?;
-        self.typed_fast_field_reader(field)
+    pub fn u64(&self, field_name: &str) -> crate::Result<Arc<dyn Column<u64>>> {
+        self.check_type(
+            self.schema.get_field(field_name)?,
+            FastType::U64,
+            Cardinality::SingleValue,
+        )?;
+        self.typed_fast_field_reader(field_name)
     }
 
     /// Returns the `ip` fast field reader reader associated to `field`.
     ///
     /// If `field` is not a u128 fast field, this method returns an Error.
-    pub fn ip_addr(&self, field: Field) -> crate::Result<Arc<dyn Column<Ipv6Addr>>> {
+    pub fn ip_addr(&self, field_name: &str) -> crate::Result<Arc<dyn Column<Ipv6Addr>>> {
+        let field = self.schema.get_field(field_name)?;
         self.check_type(field, FastType::U128, Cardinality::SingleValue)?;
         let bytes = self.fast_field_data(field, 0)?.read_bytes()?;
         Ok(open_u128::<Ipv6Addr>(bytes)?)
@@ -159,9 +166,13 @@ impl FastFieldReaders {
     /// Returns the `ip` fast field reader reader associated to `field`.
     ///
     /// If `field` is not a u128 fast field, this method returns an Error.
-    pub fn ip_addrs(&self, field: Field) -> crate::Result<MultiValuedFastFieldReader<Ipv6Addr>> {
+    pub fn ip_addrs(
+        &self,
+        field_name: &str,
+    ) -> crate::Result<MultiValuedFastFieldReader<Ipv6Addr>> {
+        let field = self.schema.get_field(field_name)?;
         self.check_type(field, FastType::U128, Cardinality::MultiValues)?;
-        let idx_reader: Arc<dyn Column<u64>> = self.typed_fast_field_reader(field)?;
+        let idx_reader: Arc<dyn Column<u64>> = self.typed_fast_field_reader(field_name)?;
 
         let bytes = self.fast_field_data(field, 1)?.read_bytes()?;
         let vals_reader = open_u128::<Ipv6Addr>(bytes)?;
@@ -172,7 +183,8 @@ impl FastFieldReaders {
     /// Returns the `u128` fast field reader reader associated to `field`.
     ///
     /// If `field` is not a u128 fast field, this method returns an Error.
-    pub(crate) fn u128(&self, field: Field) -> crate::Result<Arc<dyn Column<u128>>> {
+    pub(crate) fn u128(&self, field_name: &str) -> crate::Result<Arc<dyn Column<u128>>> {
+        let field = self.schema.get_field(field_name)?;
         self.check_type(field, FastType::U128, Cardinality::SingleValue)?;
         let bytes = self.fast_field_data(field, 0)?.read_bytes()?;
         Ok(open_u128::<u128>(bytes)?)
@@ -181,9 +193,11 @@ impl FastFieldReaders {
     /// Returns the `u128` multi-valued fast field reader reader associated to `field`.
     ///
     /// If `field` is not a u128 multi-valued fast field, this method returns an Error.
-    pub fn u128s(&self, field: Field) -> crate::Result<MultiValuedFastFieldReader<u128>> {
+    pub fn u128s(&self, field_name: &str) -> crate::Result<MultiValuedFastFieldReader<u128>> {
+        let field = self.schema.get_field(field_name)?;
         self.check_type(field, FastType::U128, Cardinality::MultiValues)?;
-        let idx_reader: Arc<dyn Column<u64>> = self.typed_fast_field_reader(field)?;
+        let idx_reader: Arc<dyn Column<u64>> =
+            self.typed_fast_field_reader(self.schema.get_field_name(field))?;
 
         let bytes = self.fast_field_data(field, 1)?.read_bytes()?;
         let vals_reader = open_u128::<u128>(bytes)?;
@@ -196,80 +210,88 @@ impl FastFieldReaders {
     ///
     /// If not, the fastfield reader will returns the u64-value associated with the original
     /// FastValue.
-    pub fn u64_lenient(&self, field: Field) -> crate::Result<Arc<dyn Column<u64>>> {
-        self.typed_fast_field_reader(field)
+    pub fn u64_lenient(&self, field_name: &str) -> crate::Result<Arc<dyn Column<u64>>> {
+        self.typed_fast_field_reader(field_name)
     }
 
     /// Returns the `i64` fast field reader reader associated with `field`.
     ///
     /// If `field` is not a i64 fast field, this method returns an Error.
-    pub fn i64(&self, field: Field) -> crate::Result<Arc<dyn Column<i64>>> {
+    pub fn i64(&self, field_name: &str) -> crate::Result<Arc<dyn Column<i64>>> {
+        let field = self.schema.get_field(field_name)?;
         self.check_type(field, FastType::I64, Cardinality::SingleValue)?;
-        self.typed_fast_field_reader(field)
+        self.typed_fast_field_reader(self.schema.get_field_name(field))
     }
 
     /// Returns the `date` fast field reader reader associated with `field`.
     ///
     /// If `field` is not a date fast field, this method returns an Error.
-    pub fn date(&self, field: Field) -> crate::Result<Arc<dyn Column<DateTime>>> {
+    pub fn date(&self, field_name: &str) -> crate::Result<Arc<dyn Column<DateTime>>> {
+        let field = self.schema.get_field(field_name)?;
         self.check_type(field, FastType::Date, Cardinality::SingleValue)?;
-        self.typed_fast_field_reader(field)
+        self.typed_fast_field_reader(field_name)
     }
 
     /// Returns the `f64` fast field reader reader associated with `field`.
     ///
     /// If `field` is not a f64 fast field, this method returns an Error.
-    pub fn f64(&self, field: Field) -> crate::Result<Arc<dyn Column<f64>>> {
+    pub fn f64(&self, field_name: &str) -> crate::Result<Arc<dyn Column<f64>>> {
+        let field = self.schema.get_field(field_name)?;
         self.check_type(field, FastType::F64, Cardinality::SingleValue)?;
-        self.typed_fast_field_reader(field)
+        self.typed_fast_field_reader(field_name)
     }
 
     /// Returns the `bool` fast field reader reader associated with `field`.
     ///
     /// If `field` is not a bool fast field, this method returns an Error.
-    pub fn bool(&self, field: Field) -> crate::Result<Arc<dyn Column<bool>>> {
+    pub fn bool(&self, field_name: &str) -> crate::Result<Arc<dyn Column<bool>>> {
+        let field = self.schema.get_field(field_name)?;
         self.check_type(field, FastType::Bool, Cardinality::SingleValue)?;
-        self.typed_fast_field_reader(field)
+        self.typed_fast_field_reader(field_name)
     }
 
     /// Returns a `u64s` multi-valued fast field reader reader associated with `field`.
     ///
     /// If `field` is not a u64 multi-valued fast field, this method returns an Error.
-    pub fn u64s(&self, field: Field) -> crate::Result<MultiValuedFastFieldReader<u64>> {
+    pub fn u64s(&self, field_name: &str) -> crate::Result<MultiValuedFastFieldReader<u64>> {
+        let field = self.schema.get_field(field_name)?;
         self.check_type(field, FastType::U64, Cardinality::MultiValues)?;
-        self.typed_fast_field_multi_reader(field)
+        self.typed_fast_field_multi_reader(field_name)
     }
 
     /// Returns a `u64s` multi-valued fast field reader reader associated with `field`, regardless
     /// of whether the given field is effectively of type `u64` or not.
     ///
     /// If `field` is not a u64 multi-valued fast field, this method returns an Error.
-    pub fn u64s_lenient(&self, field: Field) -> crate::Result<MultiValuedFastFieldReader<u64>> {
-        self.typed_fast_field_multi_reader(field)
+    pub fn u64s_lenient(&self, field_name: &str) -> crate::Result<MultiValuedFastFieldReader<u64>> {
+        self.typed_fast_field_multi_reader(field_name)
     }
 
     /// Returns a `i64s` multi-valued fast field reader reader associated with `field`.
     ///
     /// If `field` is not a i64 multi-valued fast field, this method returns an Error.
-    pub fn i64s(&self, field: Field) -> crate::Result<MultiValuedFastFieldReader<i64>> {
+    pub fn i64s(&self, field_name: &str) -> crate::Result<MultiValuedFastFieldReader<i64>> {
+        let field = self.schema.get_field(field_name)?;
         self.check_type(field, FastType::I64, Cardinality::MultiValues)?;
-        self.typed_fast_field_multi_reader(field)
+        self.typed_fast_field_multi_reader(self.schema.get_field_name(field))
     }
 
     /// Returns a `f64s` multi-valued fast field reader reader associated with `field`.
     ///
     /// If `field` is not a f64 multi-valued fast field, this method returns an Error.
-    pub fn f64s(&self, field: Field) -> crate::Result<MultiValuedFastFieldReader<f64>> {
+    pub fn f64s(&self, field_name: &str) -> crate::Result<MultiValuedFastFieldReader<f64>> {
+        let field = self.schema.get_field(field_name)?;
         self.check_type(field, FastType::F64, Cardinality::MultiValues)?;
-        self.typed_fast_field_multi_reader(field)
+        self.typed_fast_field_multi_reader(self.schema.get_field_name(field))
     }
 
     /// Returns a `bools` multi-valued fast field reader reader associated with `field`.
     ///
     /// If `field` is not a bool multi-valued fast field, this method returns an Error.
-    pub fn bools(&self, field: Field) -> crate::Result<MultiValuedFastFieldReader<bool>> {
+    pub fn bools(&self, field_name: &str) -> crate::Result<MultiValuedFastFieldReader<bool>> {
+        let field = self.schema.get_field(field_name)?;
         self.check_type(field, FastType::Bool, Cardinality::MultiValues)?;
-        self.typed_fast_field_multi_reader(field)
+        self.typed_fast_field_multi_reader(self.schema.get_field_name(field))
     }
 
     /// Returns a `time::OffsetDateTime` multi-valued fast field reader reader associated with
@@ -277,15 +299,17 @@ impl FastFieldReaders {
     ///
     /// If `field` is not a `time::OffsetDateTime` multi-valued fast field, this method returns an
     /// Error.
-    pub fn dates(&self, field: Field) -> crate::Result<MultiValuedFastFieldReader<DateTime>> {
+    pub fn dates(&self, field_name: &str) -> crate::Result<MultiValuedFastFieldReader<DateTime>> {
+        let field = self.schema.get_field(field_name)?;
         self.check_type(field, FastType::Date, Cardinality::MultiValues)?;
-        self.typed_fast_field_multi_reader(field)
+        self.typed_fast_field_multi_reader(self.schema.get_field_name(field))
     }
 
     /// Returns the `bytes` fast field reader associated with `field`.
     ///
     /// If `field` is not a bytes fast field, returns an Error.
-    pub fn bytes(&self, field: Field) -> crate::Result<BytesFastFieldReader> {
+    pub fn bytes(&self, field_name: &str) -> crate::Result<BytesFastFieldReader> {
+        let field = self.schema.get_field(field_name)?;
         let field_entry = self.schema.get_field_entry(field);
         if let FieldType::Bytes(bytes_option) = field_entry.field_type() {
             if !bytes_option.is_fast() {

src/indexer/demuxer.rs

@@ -1,322 +0,0 @@
-use common::BitSet;
-use itertools::Itertools;
-
-use crate::fastfield::AliveBitSet;
-use crate::{merge_filtered_segments, Directory, Index, IndexSettings, Segment, SegmentOrdinal};
-/// DemuxMapping can be used to reorganize data from multiple segments.
-///
-/// DemuxMapping is useful in a multitenant settings, in which each document might actually belong
-/// to a different tenant. It allows to reorganize documents as follows:
-///
-/// e.g. if you have two tenant ids TENANT_A and TENANT_B and two segments with
-/// the documents (simplified)
-/// Seg 1 [TENANT_A, TENANT_B]
-/// Seg 2 [TENANT_A, TENANT_B]
-///
-/// You may want to group your documents to
-/// Seg 1 [TENANT_A, TENANT_A]
-/// Seg 2 [TENANT_B, TENANT_B]
-///
-/// Demuxing is the tool for that.
-/// Semantically you can define a mapping from [old segment ordinal, old doc_id] -> [new segment
-/// ordinal].
-#[derive(Debug, Default)]
-pub struct DemuxMapping {
-    /// [index old segment ordinal] -> [index doc_id] = new segment ordinal
-    mapping: Vec<DocIdToSegmentOrdinal>,
-}
-
-/// DocIdToSegmentOrdinal maps from doc_id within a segment to the new segment ordinal for demuxing.
-///
-/// For every source segment there is a `DocIdToSegmentOrdinal` to distribute its doc_ids.
-#[derive(Debug, Default)]
-pub struct DocIdToSegmentOrdinal {
-    doc_id_index_to_segment_ord: Vec<SegmentOrdinal>,
-}
-
-impl DocIdToSegmentOrdinal {
-    /// Creates a new DocIdToSegmentOrdinal with size of num_doc_ids.
-    /// Initially all doc_ids point to segment ordinal 0 and need to be set
-    /// the via `set` method.
-    pub fn with_max_doc(max_doc: usize) -> Self {
-        DocIdToSegmentOrdinal {
-            doc_id_index_to_segment_ord: vec![0; max_doc],
-        }
-    }
-
-    /// Returns the number of documents in this mapping.
-    /// It should be equal to the `max_doc` of the segment it targets.
-    pub fn max_doc(&self) -> u32 {
-        self.doc_id_index_to_segment_ord.len() as u32
-    }
-
-    /// Associates a doc_id with an output `SegmentOrdinal`.
-    pub fn set(&mut self, doc_id: u32, segment_ord: SegmentOrdinal) {
-        self.doc_id_index_to_segment_ord[doc_id as usize] = segment_ord;
-    }
-
-    /// Iterates over the new SegmentOrdinal in the order of the doc_id.
-    pub fn iter(&self) -> impl Iterator<Item = SegmentOrdinal> + '_ {
-        self.doc_id_index_to_segment_ord.iter().cloned()
-    }
-}
-
-impl DemuxMapping {
-    /// Adds a DocIdToSegmentOrdinal. The order of the pus calls
-    /// defines the old segment ordinal. e.g. first push = ordinal 0.
-    pub fn add(&mut self, segment_mapping: DocIdToSegmentOrdinal) {
-        self.mapping.push(segment_mapping);
-    }
-
-    /// Returns the old number of segments.
-    pub fn get_old_num_segments(&self) -> usize {
-        self.mapping.len()
-    }
-}
-
-fn docs_for_segment_ord(
-    doc_id_to_segment_ord: &DocIdToSegmentOrdinal,
-    target_segment_ord: SegmentOrdinal,
-) -> AliveBitSet {
-    let mut bitset = BitSet::with_max_value(doc_id_to_segment_ord.max_doc());
-    for doc_id in doc_id_to_segment_ord
-        .iter()
-        .enumerate()
-        .filter(|(_doc_id, new_segment_ord)| *new_segment_ord == target_segment_ord)
-        .map(|(doc_id, _)| doc_id)
-    {
-        // add document if segment ordinal = target segment ordinal
-        bitset.insert(doc_id as u32);
-    }
-    AliveBitSet::from_bitset(&bitset)
-}
-
-fn get_alive_bitsets(
-    demux_mapping: &DemuxMapping,
-    target_segment_ord: SegmentOrdinal,
-) -> Vec<AliveBitSet> {
-    demux_mapping
-        .mapping
-        .iter()
-        .map(|doc_id_to_segment_ord| {
-            docs_for_segment_ord(doc_id_to_segment_ord, target_segment_ord)
-        })
-        .collect_vec()
-}
-
-/// Demux the segments according to `demux_mapping`. See `DemuxMapping`.
-/// The number of output_directories need to match max new segment ordinal from `demux_mapping`.
-///
-/// The ordinal of `segments` need to match the ordinals provided in `demux_mapping`.
-pub fn demux(
-    segments: &[Segment],
-    demux_mapping: &DemuxMapping,
-    target_settings: IndexSettings,
-    output_directories: Vec<Box<dyn Directory>>,
-) -> crate::Result<Vec<Index>> {
-    let mut indices = vec![];
-    for (target_segment_ord, output_directory) in output_directories.into_iter().enumerate() {
-        let alive_bitset = get_alive_bitsets(demux_mapping, target_segment_ord as u32)
-            .into_iter()
-            .map(Some)
-            .collect_vec();
-        let index = merge_filtered_segments(
-            segments,
-            target_settings.clone(),
-            alive_bitset,
-            output_directory,
-        )?;
-        indices.push(index);
-    }
-    Ok(indices)
-}
-
-#[cfg(test)]
-mod tests {
-    use super::*;
-    use crate::collector::TopDocs;
-    use crate::directory::RamDirectory;
-    use crate::query::QueryParser;
-    use crate::schema::{Schema, TEXT};
-    use crate::{DocAddress, Term};
-
-    #[test]
-    fn test_demux_map_to_alive_bitset() {
-        let max_value = 2;
-        let mut demux_mapping = DemuxMapping::default();
-        // segment ordinal 0 mapping
-        let mut doc_id_to_segment = DocIdToSegmentOrdinal::with_max_doc(max_value);
-        doc_id_to_segment.set(0, 1);
-        doc_id_to_segment.set(1, 0);
-        demux_mapping.add(doc_id_to_segment);
-
-        // segment ordinal 1 mapping
-        let mut doc_id_to_segment = DocIdToSegmentOrdinal::with_max_doc(max_value);
-        doc_id_to_segment.set(0, 1);
-        doc_id_to_segment.set(1, 1);
-        demux_mapping.add(doc_id_to_segment);
-        {
-            let bit_sets_for_demuxing_to_segment_ord_0 = get_alive_bitsets(&demux_mapping, 0);
-
-            assert_eq!(
-                bit_sets_for_demuxing_to_segment_ord_0[0].is_deleted(0),
-                true
-            );
-            assert_eq!(
-                bit_sets_for_demuxing_to_segment_ord_0[0].is_deleted(1),
-                false
-            );
-            assert_eq!(
-                bit_sets_for_demuxing_to_segment_ord_0[1].is_deleted(0),
-                true
-            );
-            assert_eq!(
-                bit_sets_for_demuxing_to_segment_ord_0[1].is_deleted(1),
-                true
-            );
-        }
-
-        {
-            let bit_sets_for_demuxing_to_segment_ord_1 = get_alive_bitsets(&demux_mapping, 1);
-
-            assert_eq!(
-                bit_sets_for_demuxing_to_segment_ord_1[0].is_deleted(0),
-                false
-            );
-            assert_eq!(
-                bit_sets_for_demuxing_to_segment_ord_1[0].is_deleted(1),
-                true
-            );
-            assert_eq!(
-                bit_sets_for_demuxing_to_segment_ord_1[1].is_deleted(0),
-                false
-            );
-            assert_eq!(
-                bit_sets_for_demuxing_to_segment_ord_1[1].is_deleted(1),
-                false
-            );
-        }
-    }
-
-    #[test]
-    fn test_demux_segments() -> crate::Result<()> {
-        let first_index = {
-            let mut schema_builder = Schema::builder();
-            let text_field = schema_builder.add_text_field("text", TEXT);
-            let index = Index::create_in_ram(schema_builder.build());
-            let mut index_writer = index.writer_for_tests()?;
-            index_writer.add_document(doc!(text_field=>"texto1"))?;
-            index_writer.add_document(doc!(text_field=>"texto2"))?;
-            index_writer.commit()?;
-            index
-        };
-
-        let second_index = {
-            let mut schema_builder = Schema::builder();
-            let text_field = schema_builder.add_text_field("text", TEXT);
-            let index = Index::create_in_ram(schema_builder.build());
-            let mut index_writer = index.writer_for_tests()?;
-            index_writer.add_document(doc!(text_field=>"texto3"))?;
-            index_writer.add_document(doc!(text_field=>"texto4"))?;
-            index_writer.delete_term(Term::from_field_text(text_field, "4"));
-
-            index_writer.commit()?;
-            index
-        };
-
-        let mut segments: Vec<Segment> = Vec::new();
-        segments.extend(first_index.searchable_segments()?);
-        segments.extend(second_index.searchable_segments()?);
-
-        let target_settings = first_index.settings().clone();
-
-        let mut demux_mapping = DemuxMapping::default();
-        {
-            let max_value = 2;
-            // segment ordinal 0 mapping
-            let mut doc_id_to_segment = DocIdToSegmentOrdinal::with_max_doc(max_value);
-            doc_id_to_segment.set(0, 1);
-            doc_id_to_segment.set(1, 0);
-            demux_mapping.add(doc_id_to_segment);
-
-            // segment ordinal 1 mapping
-            let mut doc_id_to_segment = DocIdToSegmentOrdinal::with_max_doc(max_value);
-            doc_id_to_segment.set(0, 1);
-            doc_id_to_segment.set(1, 1);
-            demux_mapping.add(doc_id_to_segment);
-        }
-        assert_eq!(demux_mapping.get_old_num_segments(), 2);
-
-        let demuxed_indices = demux(
-            &segments,
-            &demux_mapping,
-            target_settings,
-            vec![
-                Box::<RamDirectory>::default(),
-                Box::<RamDirectory>::default(),
-            ],
-        )?;
-
-        {
-            let index = &demuxed_indices[0];
-
-            let segments = index.searchable_segments()?;
-            assert_eq!(segments.len(), 1);
-
-            let segment_metas = segments[0].meta();
-            assert_eq!(segment_metas.num_deleted_docs(), 0);
-            assert_eq!(segment_metas.num_docs(), 1);
-
-            let searcher = index.reader().unwrap().searcher();
-            {
-                let text_field = index.schema().get_field("text").unwrap();
-
-                let do_search = |term: &str| {
-                    let query = QueryParser::for_index(index, vec![text_field])
-                        .parse_query(term)
-                        .unwrap();
-                    let top_docs: Vec<(f32, DocAddress)> =
-                        searcher.search(&query, &TopDocs::with_limit(3)).unwrap();
-
-                    top_docs.iter().map(|el| el.1.doc_id).collect::<Vec<_>>()
-                };
-
-                assert_eq!(do_search("texto1"), vec![] as Vec<u32>);
-                assert_eq!(do_search("texto2"), vec![0]);
-            }
-        }
-
-        {
-            let index = &demuxed_indices[1];
-
-            let segments = index.searchable_segments()?;
-            assert_eq!(segments.len(), 1);
-
-            let segment_metas = segments[0].meta();
-            assert_eq!(segment_metas.num_deleted_docs(), 0);
-            assert_eq!(segment_metas.num_docs(), 3);
-
-            let searcher = index.reader().unwrap().searcher();
-            {
-                let text_field = index.schema().get_field("text").unwrap();
-
-                let do_search = |term: &str| {
-                    let query = QueryParser::for_index(index, vec![text_field])
-                        .parse_query(term)
-                        .unwrap();
-                    let top_docs: Vec<(f32, DocAddress)> =
-                        searcher.search(&query, &TopDocs::with_limit(3)).unwrap();
-
-                    top_docs.iter().map(|el| el.1.doc_id).collect::<Vec<_>>()
-                };
-
-                assert_eq!(do_search("texto1"), vec![0]);
-                assert_eq!(do_search("texto2"), vec![] as Vec<u32>);
-                assert_eq!(do_search("texto3"), vec![1]);
-                assert_eq!(do_search("texto4"), vec![2]);
-            }
-        }
-
-        Ok(())
-    }
-}

src/indexer/doc_id_mapping.rs

@@ -99,7 +99,7 @@ pub(crate) fn expect_field_id_for_sort_field(
     schema: &Schema,
     sort_by_field: &IndexSortByField,
 ) -> crate::Result<Field> {
-    schema.get_field(&sort_by_field.field).ok_or_else(|| {
+    schema.get_field(&sort_by_field.field).map_err(|_| {
         TantivyError::InvalidArgument(format!(
             "field to sort index by not found: {:?}",
             sort_by_field.field
@@ -462,15 +462,14 @@ mod tests_indexsorting {
         assert_eq!(searcher.segment_readers().len(), 1);
         let segment_reader = searcher.segment_reader(0);
         let fast_fields = segment_reader.fast_fields();
-        let my_number = index.schema().get_field("my_number").unwrap();
+        index.schema().get_field("my_number").unwrap();
 
-        let fast_field = fast_fields.u64(my_number).unwrap();
+        let fast_field = fast_fields.u64("my_number").unwrap();
         assert_eq!(fast_field.get_val(0), 10u64);
         assert_eq!(fast_field.get_val(1), 20u64);
         assert_eq!(fast_field.get_val(2), 30u64);
 
-        let multi_numbers = index.schema().get_field("multi_numbers").unwrap();
-        let multifield = fast_fields.u64s(multi_numbers).unwrap();
+        let multifield = fast_fields.u64s("multi_numbers").unwrap();
         let mut vals = vec![];
         multifield.get_vals(0u32, &mut vals);
         assert_eq!(vals, &[] as &[u64]);

src/indexer/index_writer.rs

@@ -1465,7 +1465,7 @@ mod tests {
         let segment_reader = searcher.segment_reader(0);
         assert_eq!(segment_reader.num_docs(), 8);
         assert_eq!(segment_reader.max_doc(), 10);
-        let fast_field_reader = segment_reader.fast_fields().u64(id_field)?;
+        let fast_field_reader = segment_reader.fast_fields().u64("id")?;
         let in_order_alive_ids: Vec<u64> = segment_reader
             .doc_ids_alive()
             .map(|doc| fast_field_reader.get_val(doc))
@@ -1526,7 +1526,7 @@ mod tests {
         let segment_reader = searcher.segment_reader(0);
         assert_eq!(segment_reader.num_docs(), 8);
         assert_eq!(segment_reader.max_doc(), 10);
-        let fast_field_reader = segment_reader.fast_fields().u64(id_field)?;
+        let fast_field_reader = segment_reader.fast_fields().u64("id")?;
         let in_order_alive_ids: Vec<u64> = segment_reader
             .doc_ids_alive()
             .map(|doc| fast_field_reader.get_val(doc))
@@ -1778,7 +1778,7 @@ mod tests {
             .segment_readers()
             .iter()
             .flat_map(|segment_reader| {
-                let ff_reader = segment_reader.fast_fields().u64(id_field).unwrap();
+                let ff_reader = segment_reader.fast_fields().u64("id").unwrap();
                 segment_reader
                     .doc_ids_alive()
                     .map(move |doc| ff_reader.get_val(doc))
@@ -1789,7 +1789,7 @@ mod tests {
             .segment_readers()
             .iter()
             .flat_map(|segment_reader| {
-                let ff_reader = segment_reader.fast_fields().u64(id_field).unwrap();
+                let ff_reader = segment_reader.fast_fields().u64("id").unwrap();
                 segment_reader
                     .doc_ids_alive()
                     .map(move |doc| ff_reader.get_val(doc))
@@ -1804,7 +1804,7 @@ mod tests {
         let mut all_ips = Vec::new();
         let mut num_ips = 0;
         for segment_reader in searcher.segment_readers().iter() {
-            let ip_reader = segment_reader.fast_fields().ip_addrs(ips_field).unwrap();
+            let ip_reader = segment_reader.fast_fields().ip_addrs("ips").unwrap();
             for doc in segment_reader.doc_ids_alive() {
                 let mut vals = vec![];
                 ip_reader.get_vals(doc, &mut vals);
@@ -1851,7 +1851,7 @@ mod tests {
                 .segment_readers()
                 .iter()
                 .map(|segment_reader| {
-                    let ff_reader = segment_reader.fast_fields().ip_addrs(ips_field).unwrap();
+                    let ff_reader = segment_reader.fast_fields().ip_addrs("ips").unwrap();
                     ff_reader.get_index_reader().num_docs() as usize
                 })
                 .sum();
@@ -1863,7 +1863,7 @@ mod tests {
             .segment_readers()
             .iter()
             .flat_map(|segment_reader| {
-                let ff_reader = segment_reader.fast_fields().ip_addr(ip_field).unwrap();
+                let ff_reader = segment_reader.fast_fields().ip_addr("ip").unwrap();
                 segment_reader.doc_ids_alive().flat_map(move |doc| {
                     let val = ff_reader.get_val(doc);
                     if val == Ipv6Addr::from_u128(0) {
@@ -1902,7 +1902,7 @@ mod tests {
             .segment_readers()
             .iter()
             .flat_map(|segment_reader| {
-                let ff_reader = segment_reader.fast_fields().ip_addrs(ips_field).unwrap();
+                let ff_reader = segment_reader.fast_fields().ip_addrs("ips").unwrap();
                 segment_reader.doc_ids_alive().flat_map(move |doc| {
                     let mut vals = vec![];
                     ff_reader.get_vals(doc, &mut vals);
@@ -1914,9 +1914,9 @@ mod tests {
 
         // multivalue fast field tests
         for segment_reader in searcher.segment_readers().iter() {
-            let id_reader = segment_reader.fast_fields().u64(id_field).unwrap();
-            let ff_reader = segment_reader.fast_fields().u64s(multi_numbers).unwrap();
-            let bool_ff_reader = segment_reader.fast_fields().bools(multi_bools).unwrap();
+            let id_reader = segment_reader.fast_fields().u64("id").unwrap();
+            let ff_reader = segment_reader.fast_fields().u64s("multi_numbers").unwrap();
+            let bool_ff_reader = segment_reader.fast_fields().bools("multi_bools").unwrap();
             for doc in segment_reader.doc_ids_alive() {
                 let mut vals = vec![];
                 ff_reader.get_vals(doc, &mut vals);
@@ -2109,7 +2109,7 @@ mod tests {
         // test facets
         for segment_reader in searcher.segment_readers().iter() {
             let mut facet_reader = segment_reader.facet_reader(facet_field).unwrap();
-            let ff_reader = segment_reader.fast_fields().u64(id_field).unwrap();
+            let ff_reader = segment_reader.fast_fields().u64("id").unwrap();
             for doc_id in segment_reader.doc_ids_alive() {
                 let mut facet_ords = Vec::new();
                 facet_reader.facet_ords(doc_id, &mut facet_ords);

src/indexer/merger.rs

@@ -16,7 +16,7 @@ use crate::fastfield::{
     MultiValueIndex, MultiValuedFastFieldReader,
 };
 use crate::fieldnorm::{FieldNormReader, FieldNormReaders, FieldNormsSerializer, FieldNormsWriter};
-use crate::indexer::doc_id_mapping::{expect_field_id_for_sort_field, SegmentDocIdMapping};
+use crate::indexer::doc_id_mapping::SegmentDocIdMapping;
 use crate::indexer::sorted_doc_id_column::RemappedDocIdColumn;
 use crate::indexer::sorted_doc_id_multivalue_column::RemappedDocIdMultiValueColumn;
 use crate::indexer::SegmentSerializer;
@@ -335,8 +335,10 @@ impl IndexMerger {
             .readers
             .iter()
             .map(|segment_reader| {
-                let ff_reader: MultiValuedFastFieldReader<u128> =
-                    segment_reader.fast_fields().u128s(field).expect(
+                let ff_reader: MultiValuedFastFieldReader<u128> = segment_reader
+                    .fast_fields()
+                    .u128s(self.schema.get_field_name(field))
+                    .expect(
                         "Failed to find index for multivalued field. This is a bug in tantivy, \
                          please report.",
                     );
@@ -401,10 +403,13 @@ impl IndexMerger {
             .readers
             .iter()
             .map(|reader| {
-                let u128_reader: Arc<dyn Column<u128>> = reader.fast_fields().u128(field).expect(
-                    "Failed to find a reader for single fast field. This is a tantivy bug and it \
-                     should never happen.",
-                );
+                let u128_reader: Arc<dyn Column<u128>> = reader
+                    .fast_fields()
+                    .u128(self.schema.get_field_name(field))
+                    .expect(
+                        "Failed to find a reader for single fast field. This is a tantivy bug and \
+                         it should never happen.",
+                    );
                 u128_reader
             })
             .collect::<Vec<_>>();
@@ -431,7 +436,11 @@ impl IndexMerger {
         fast_field_serializer: &mut CompositeFastFieldSerializer,
         doc_id_mapping: &SegmentDocIdMapping,
     ) -> crate::Result<()> {
-        let fast_field_accessor = RemappedDocIdColumn::new(&self.readers, doc_id_mapping, field);
+        let fast_field_accessor = RemappedDocIdColumn::new(
+            &self.readers,
+            doc_id_mapping,
+            self.schema.get_field_name(field),
+        );
         fast_field_serializer.create_auto_detect_u64_fast_field(field, fast_field_accessor)?;
 
         Ok(())
@@ -464,8 +473,8 @@ impl IndexMerger {
         reader: &SegmentReader,
         sort_by_field: &IndexSortByField,
     ) -> crate::Result<Arc<dyn Column>> {
-        let field_id = expect_field_id_for_sort_field(reader.schema(), sort_by_field)?; // for now expect fastfield, but not strictly required
-        let value_accessor = reader.fast_fields().u64_lenient(field_id)?;
+        reader.schema().get_field(&sort_by_field.field)?;
+        let value_accessor = reader.fast_fields().u64_lenient(&sort_by_field.field)?;
         Ok(value_accessor)
     }
     /// Collecting value_accessors into a vec to bind the lifetime.
@@ -569,7 +578,7 @@ impl IndexMerger {
             .map(|reader| {
                 let u64s_reader: MultiValuedFastFieldReader<u64> = reader
                     .fast_fields()
-                    .typed_fast_field_multi_reader::<u64>(field)
+                    .typed_fast_field_multi_reader::<u64>(self.schema.get_field_name(field))
                     .expect(
                         "Failed to find index for multivalued field. This is a bug in tantivy, \
                          please report.",
@@ -613,7 +622,7 @@ impl IndexMerger {
             .map(|reader| {
                 let ff_reader: MultiValuedFastFieldReader<u64> = reader
                     .fast_fields()
-                    .u64s(field)
+                    .u64s(self.schema.get_field_name(field))
                     .expect("Could not find multivalued u64 fast value reader.");
                 ff_reader
             })
@@ -684,8 +693,11 @@ impl IndexMerger {
 
         self.write_multi_value_fast_field_idx(field, fast_field_serializer, doc_id_mapping)?;
 
-        let fastfield_accessor =
-            RemappedDocIdMultiValueColumn::new(&self.readers, doc_id_mapping, field);
+        let fastfield_accessor = RemappedDocIdMultiValueColumn::new(
+            &self.readers,
+            doc_id_mapping,
+            self.schema.get_field_name(field),
+        );
         fast_field_serializer.create_auto_detect_u64_fast_field_with_idx_and_codecs(
             field,
             fastfield_accessor,
@@ -706,10 +718,13 @@ impl IndexMerger {
             .readers
             .iter()
             .map(|reader| {
-                let bytes_reader = reader.fast_fields().bytes(field).expect(
-                    "Failed to find index for bytes field. This is a bug in tantivy, please \
-                     report.",
-                );
+                let bytes_reader = reader
+                    .fast_fields()
+                    .bytes(self.schema.get_field_name(field))
+                    .expect(
+                        "Failed to find index for bytes field. This is a bug in tantivy, please \
+                         report.",
+                    );
                 (reader, bytes_reader)
             })
             .collect::<Vec<_>>();
@@ -1206,7 +1221,10 @@ mod tests {
             {
                 let get_fast_vals = |terms: Vec<Term>| {
                     let query = BooleanQuery::new_multiterms_query(terms);
-                    searcher.search(&query, &FastFieldTestCollector::for_field(score_field))
+                    searcher.search(
+                        &query,
+                        &FastFieldTestCollector::for_field("score".to_string()),
+                    )
                 };
                 let get_fast_vals_bytes = |terms: Vec<Term>| {
                     let query = BooleanQuery::new_multiterms_query(terms);
@@ -1244,7 +1262,7 @@ mod tests {
         let mut index_writer = index.writer_for_tests()?;
         let reader = index.reader().unwrap();
         let search_term = |searcher: &Searcher, term: Term| {
-            let collector = FastFieldTestCollector::for_field(score_field);
+            let collector = FastFieldTestCollector::for_field("score".to_string());
             let bytes_collector = BytesFastFieldTestCollector::for_field(bytes_score_field);
             let term_query = TermQuery::new(term, IndexRecordOption::Basic);
             searcher
@@ -1366,7 +1384,7 @@ mod tests {
             let score_field_reader = searcher
                 .segment_reader(0)
                 .fast_fields()
-                .u64(score_field)
+                .u64("score")
                 .unwrap();
             assert_eq!(score_field_reader.min_value(), 4000);
             assert_eq!(score_field_reader.max_value(), 7000);
@@ -1374,7 +1392,7 @@ mod tests {
             let score_field_reader = searcher
                 .segment_reader(1)
                 .fast_fields()
-                .u64(score_field)
+                .u64("score")
                 .unwrap();
             assert_eq!(score_field_reader.min_value(), 1);
             assert_eq!(score_field_reader.max_value(), 3);
@@ -1420,7 +1438,7 @@ mod tests {
             let score_field_reader = searcher
                 .segment_reader(0)
                 .fast_fields()
-                .u64(score_field)
+                .u64("score")
                 .unwrap();
             assert_eq!(score_field_reader.min_value(), 3);
             assert_eq!(score_field_reader.max_value(), 7000);
@@ -1467,7 +1485,7 @@ mod tests {
             let score_field_reader = searcher
                 .segment_reader(0)
                 .fast_fields()
-                .u64(score_field)
+                .u64("score")
                 .unwrap();
             assert_eq!(score_field_reader.min_value(), 3);
             assert_eq!(score_field_reader.max_value(), 7000);
@@ -1514,7 +1532,7 @@ mod tests {
             let score_field_reader = searcher
                 .segment_reader(0)
                 .fast_fields()
-                .u64(score_field)
+                .u64("score")
                 .unwrap();
             assert_eq!(score_field_reader.min_value(), 6000);
             assert_eq!(score_field_reader.max_value(), 7000);
@@ -1836,7 +1854,7 @@ mod tests {
 
         {
             let segment = searcher.segment_reader(0u32);
-            let ff_reader = segment.fast_fields().u64s(int_field).unwrap();
+            let ff_reader = segment.fast_fields().u64s("intvals").unwrap();
 
             ff_reader.get_vals(0, &mut vals);
             assert_eq!(&vals, &[1, 2]);
@@ -1862,7 +1880,7 @@ mod tests {
 
         {
             let segment = searcher.segment_reader(1u32);
-            let ff_reader = segment.fast_fields().u64s(int_field).unwrap();
+            let ff_reader = segment.fast_fields().u64s("intvals").unwrap();
             ff_reader.get_vals(0, &mut vals);
             assert_eq!(&vals, &[28, 27]);
 
@@ -1872,7 +1890,7 @@ mod tests {
 
         {
             let segment = searcher.segment_reader(2u32);
-            let ff_reader = segment.fast_fields().u64s(int_field).unwrap();
+            let ff_reader = segment.fast_fields().u64s("intvals").unwrap();
             ff_reader.get_vals(0, &mut vals);
             assert_eq!(&vals, &[20]);
         }
@@ -1889,7 +1907,7 @@ mod tests {
         {
             let searcher = reader.searcher();
             let segment = searcher.segment_reader(0u32);
-            let ff_reader = segment.fast_fields().u64s(int_field).unwrap();
+            let ff_reader = segment.fast_fields().u64s("intvals").unwrap();
 
             ff_reader.get_vals(0, &mut vals);
             assert_eq!(&vals, &[1, 2]);

src/indexer/merger_sorted_index_test.rs

@@ -185,7 +185,7 @@ mod tests {
         let segment_reader = searcher.segment_readers().last().unwrap();
 
         let fast_fields = segment_reader.fast_fields();
-        let fast_field = fast_fields.u64(int_field).unwrap();
+        let fast_field = fast_fields.u64("intval").unwrap();
         assert_eq!(fast_field.get_val(5), 1u64);
         assert_eq!(fast_field.get_val(4), 2u64);
         assert_eq!(fast_field.get_val(3), 3u64);
@@ -364,15 +364,13 @@ mod tests {
         .unwrap();
 
         let int_field = index.schema().get_field("intval").unwrap();
-        let multi_numbers = index.schema().get_field("multi_numbers").unwrap();
-        let bytes_field = index.schema().get_field("bytes").unwrap();
         let reader = index.reader().unwrap();
         let searcher = reader.searcher();
         assert_eq!(searcher.segment_readers().len(), 1);
         let segment_reader = searcher.segment_readers().last().unwrap();
 
         let fast_fields = segment_reader.fast_fields();
-        let fast_field = fast_fields.u64(int_field).unwrap();
+        let fast_field = fast_fields.u64("intval").unwrap();
         assert_eq!(fast_field.get_val(0), 1u64);
         assert_eq!(fast_field.get_val(1), 2u64);
         assert_eq!(fast_field.get_val(2), 3u64);
@@ -386,7 +384,7 @@ mod tests {
             vals
         };
         let fast_fields = segment_reader.fast_fields();
-        let fast_field = fast_fields.u64s(multi_numbers).unwrap();
+        let fast_field = fast_fields.u64s("multi_numbers").unwrap();
         assert_eq!(&get_vals(&fast_field, 0), &[] as &[u64]);
         assert_eq!(&get_vals(&fast_field, 1), &[2, 3]);
         assert_eq!(&get_vals(&fast_field, 2), &[3, 4]);
@@ -394,7 +392,7 @@ mod tests {
         assert_eq!(&get_vals(&fast_field, 4), &[20]);
         assert_eq!(&get_vals(&fast_field, 5), &[1001, 1002]);
 
-        let fast_field = fast_fields.bytes(bytes_field).unwrap();
+        let fast_field = fast_fields.bytes("bytes").unwrap();
         assert_eq!(fast_field.get_bytes(0), &[] as &[u8]);
         assert_eq!(fast_field.get_bytes(2), &[1, 2, 3]);
         assert_eq!(fast_field.get_bytes(5), &[5, 5]);
@@ -527,7 +525,6 @@ mod bench_sorted_index_merge {
             order: Order::Desc,
         };
         let index = create_index(Some(sort_by_field.clone()));
-        let field = index.schema().get_field("intval").unwrap();
         let segments = index.searchable_segments().unwrap();
         let merger: IndexMerger =
             IndexMerger::open(index.schema(), index.settings().clone(), &segments[..])?;
@@ -535,8 +532,10 @@ mod bench_sorted_index_merge {
         b.iter(|| {
             let sorted_doc_ids = doc_id_mapping.iter_old_doc_addrs().map(|doc_addr| {
                 let reader = &merger.readers[doc_addr.segment_ord as usize];
-                let u64_reader: Arc<dyn Column<u64>> =
-                    reader.fast_fields().typed_fast_field_reader(field).expect(
+                let u64_reader: Arc<dyn Column<u64>> = reader
+                    .fast_fields()
+                    .typed_fast_field_reader("intval")
+                    .expect(
                         "Failed to find a reader for single fast field. This is a tantivy bug and \
                          it should never happen.",
                     );

src/indexer/mod.rs

@@ -1,6 +1,5 @@
 pub mod delete_queue;
 
-pub mod demuxer;
 pub mod doc_id_mapping;
 mod doc_opstamp_mapping;
 mod flat_map_with_buffer;

src/indexer/segment_writer.rs

@@ -834,20 +834,23 @@ mod tests {
         // This is a bit of a contrived example.
         let tokens = PreTokenizedString {
             text: "contrived-example".to_string(), //< I can't think of a use case where this corner case happens in real life.
-            tokens: vec![Token { // Not the last token, yet ends after the last token.
-                offset_from: 0,
-                offset_to: 14,
-                position: 0,
-                text: "long_token".to_string(),
-                position_length: 3,
-            },
-            Token {
-                offset_from: 0,
-                offset_to: 14,
-                position: 1,
-                text: "short".to_string(),
-                position_length: 1,
-            }],
+            tokens: vec![
+                Token {
+                    // Not the last token, yet ends after the last token.
+                    offset_from: 0,
+                    offset_to: 14,
+                    position: 0,
+                    text: "long_token".to_string(),
+                    position_length: 3,
+                },
+                Token {
+                    offset_from: 0,
+                    offset_to: 14,
+                    position: 1,
+                    text: "short".to_string(),
+                    position_length: 1,
+                },
+            ],
         };
         doc.add_pre_tokenized_text(text, tokens);
         doc.add_text(text, "hello");

src/indexer/sorted_doc_id_column.rs

@@ -4,7 +4,6 @@ use fastfield_codecs::Column;
 use itertools::Itertools;
 
 use crate::indexer::doc_id_mapping::SegmentDocIdMapping;
-use crate::schema::Field;
 use crate::SegmentReader;
 
 pub(crate) struct RemappedDocIdColumn<'a> {
@@ -41,7 +40,7 @@ impl<'a> RemappedDocIdColumn<'a> {
     pub(crate) fn new(
         readers: &'a [SegmentReader],
         doc_id_mapping: &'a SegmentDocIdMapping,
-        field: Field,
+        field: &str,
     ) -> Self {
         let (min_value, max_value) = readers
             .iter()

src/indexer/sorted_doc_id_multivalue_column.rs

@@ -5,7 +5,6 @@ use fastfield_codecs::Column;
 use super::flat_map_with_buffer::FlatMapWithBufferIter;
 use crate::fastfield::{MultiValueIndex, MultiValuedFastFieldReader};
 use crate::indexer::doc_id_mapping::SegmentDocIdMapping;
-use crate::schema::Field;
 use crate::{DocAddress, SegmentReader};
 
 pub(crate) struct RemappedDocIdMultiValueColumn<'a> {
@@ -20,7 +19,7 @@ impl<'a> RemappedDocIdMultiValueColumn<'a> {
     pub(crate) fn new(
         readers: &'a [SegmentReader],
         doc_id_mapping: &'a SegmentDocIdMapping,
-        field: Field,
+        field: &str,
     ) -> Self {
         // Our values are bitpacked and we need to know what should be
         // our bitwidth and our minimum value before serializing any values.

src/lib.rs

@@ -299,7 +299,6 @@ pub use crate::core::{
     SegmentReader, SingleSegmentIndexWriter,
 };
 pub use crate::directory::Directory;
-pub use crate::indexer::demuxer::*;
 pub use crate::indexer::operation::UserOperation;
 pub use crate::indexer::{merge_filtered_segments, merge_indices, IndexWriter, PreparedCommit};
 pub use crate::postings::Postings;
@@ -995,8 +994,8 @@ pub mod tests {
         let fast_field_unsigned = schema_builder.add_u64_field("unsigned", FAST);
         let fast_field_signed = schema_builder.add_i64_field("signed", FAST);
         let fast_field_float = schema_builder.add_f64_field("float", FAST);
-        let text_field = schema_builder.add_text_field("text", TEXT);
-        let stored_int_field = schema_builder.add_u64_field("stored_int", STORED);
+        schema_builder.add_text_field("text", TEXT);
+        schema_builder.add_u64_field("stored_int", STORED);
         let schema = schema_builder.build();
 
         let index = Index::create_in_ram(schema);
@@ -1011,37 +1010,37 @@ pub mod tests {
         let searcher = reader.searcher();
         let segment_reader: &SegmentReader = searcher.segment_reader(0);
         {
-            let fast_field_reader_res = segment_reader.fast_fields().u64(text_field);
+            let fast_field_reader_res = segment_reader.fast_fields().u64("text");
             assert!(fast_field_reader_res.is_err());
         }
         {
-            let fast_field_reader_opt = segment_reader.fast_fields().u64(stored_int_field);
+            let fast_field_reader_opt = segment_reader.fast_fields().u64("stored_int");
             assert!(fast_field_reader_opt.is_err());
         }
         {
-            let fast_field_reader_opt = segment_reader.fast_fields().u64(fast_field_signed);
+            let fast_field_reader_opt = segment_reader.fast_fields().u64("signed");
             assert!(fast_field_reader_opt.is_err());
         }
         {
-            let fast_field_reader_opt = segment_reader.fast_fields().u64(fast_field_float);
+            let fast_field_reader_opt = segment_reader.fast_fields().u64("float");
             assert!(fast_field_reader_opt.is_err());
         }
         {
-            let fast_field_reader_opt = segment_reader.fast_fields().u64(fast_field_unsigned);
+            let fast_field_reader_opt = segment_reader.fast_fields().u64("unsigned");
             assert!(fast_field_reader_opt.is_ok());
             let fast_field_reader = fast_field_reader_opt.unwrap();
             assert_eq!(fast_field_reader.get_val(0), 4u64)
         }
 
         {
-            let fast_field_reader_res = segment_reader.fast_fields().i64(fast_field_signed);
+            let fast_field_reader_res = segment_reader.fast_fields().i64("signed");
             assert!(fast_field_reader_res.is_ok());
             let fast_field_reader = fast_field_reader_res.unwrap();
             assert_eq!(fast_field_reader.get_val(0), 4i64)
         }
 
         {
-            let fast_field_reader_res = segment_reader.fast_fields().f64(fast_field_float);
+            let fast_field_reader_res = segment_reader.fast_fields().f64("float");
             assert!(fast_field_reader_res.is_ok());
             let fast_field_reader = fast_field_reader_res.unwrap();
             assert_eq!(fast_field_reader.get_val(0), 4f64)

src/query/query_parser/logical_ast.rs

@@ -10,7 +10,7 @@ pub enum LogicalLiteral {
     Term(Term),
     Phrase(Vec<(usize, Term)>, u32),
     Range {
-        field: Field,
+        field: String,
         value_type: Type,
         lower: Bound<Term>,
         upper: Bound<Term>,

src/query/query_parser/query_parser.rs

@@ -672,7 +672,7 @@ impl QueryParser {
                 let field_entry = self.schema.get_field_entry(field);
                 let value_type = field_entry.field_type().value_type();
                 let logical_ast = LogicalAst::Leaf(Box::new(LogicalLiteral::Range {
-                    field,
+                    field: self.schema.get_field_name(field).to_string(),
                     value_type,
                     lower: self.resolve_bound(field, json_path, &lower)?,
                     upper: self.resolve_bound(field, json_path, &upper)?,
@@ -964,7 +964,7 @@ mod test {
         let query = make_query_parser().parse_query("title:[A TO B]").unwrap();
         assert_eq!(
             format!("{:?}", query),
-            "RangeQuery { field: Field(0), value_type: Str, left_bound: Included([97]), \
+            "RangeQuery { field: \"title\", value_type: Str, left_bound: Included([97]), \
              right_bound: Included([98]) }"
         );
     }

src/query/range_query/range_query.rs

@@ -67,7 +67,7 @@ pub(crate) fn map_bound<TFrom, TTo, Transform: Fn(&TFrom) -> TTo>(
 ///
 /// let reader = index.reader()?;
 /// let searcher = reader.searcher();
-/// let docs_in_the_sixties = RangeQuery::new_u64(year_field, 1960..1970);
+/// let docs_in_the_sixties = RangeQuery::new_u64("year".to_string(), 1960..1970);
 /// let num_60s_books = searcher.search(&docs_in_the_sixties, &Count)?;
 /// assert_eq!(num_60s_books, 2285);
 /// Ok(())
@@ -76,7 +76,7 @@ pub(crate) fn map_bound<TFrom, TTo, Transform: Fn(&TFrom) -> TTo>(
 /// ```
 #[derive(Clone, Debug)]
 pub struct RangeQuery {
-    field: Field,
+    field: String,
     value_type: Type,
     left_bound: Bound<Vec<u8>>,
     right_bound: Bound<Vec<u8>>,
@@ -88,15 +88,12 @@ impl RangeQuery {
     /// If the value type is not correct, something may go terribly wrong when
     /// the `Weight` object is created.
     pub fn new_term_bounds(
-        field: Field,
+        field: String,
         value_type: Type,
         left_bound: &Bound<Term>,
         right_bound: &Bound<Term>,
     ) -> RangeQuery {
-        let verify_and_unwrap_term = |val: &Term| {
-            assert_eq!(field, val.field());
-            val.value_bytes().to_owned()
-        };
+        let verify_and_unwrap_term = |val: &Term| val.value_bytes().to_owned();
         RangeQuery {
             field,
             value_type,
@@ -109,7 +106,7 @@ impl RangeQuery {
     ///
     /// If the field is not of the type `i64`, tantivy
     /// will panic when the `Weight` object is created.
-    pub fn new_i64(field: Field, range: Range<i64>) -> RangeQuery {
+    pub fn new_i64(field: String, range: Range<i64>) -> RangeQuery {
         RangeQuery::new_i64_bounds(
             field,
             Bound::Included(range.start),
@@ -125,11 +122,15 @@ impl RangeQuery {
     /// If the field is not of the type `i64`, tantivy
     /// will panic when the `Weight` object is created.
     pub fn new_i64_bounds(
-        field: Field,
+        field: String,
         left_bound: Bound<i64>,
         right_bound: Bound<i64>,
     ) -> RangeQuery {
-        let make_term_val = |val: &i64| Term::from_field_i64(field, *val).value_bytes().to_owned();
+        let make_term_val = |val: &i64| {
+            Term::from_field_i64(Field::from_field_id(0), *val)
+                .value_bytes()
+                .to_owned()
+        };
         RangeQuery {
             field,
             value_type: Type::I64,
@@ -142,7 +143,7 @@ impl RangeQuery {
     ///
     /// If the field is not of the type `f64`, tantivy
     /// will panic when the `Weight` object is created.
-    pub fn new_f64(field: Field, range: Range<f64>) -> RangeQuery {
+    pub fn new_f64(field: String, range: Range<f64>) -> RangeQuery {
         RangeQuery::new_f64_bounds(
             field,
             Bound::Included(range.start),
@@ -158,11 +159,15 @@ impl RangeQuery {
     /// If the field is not of the type `f64`, tantivy
     /// will panic when the `Weight` object is created.
     pub fn new_f64_bounds(
-        field: Field,
+        field: String,
         left_bound: Bound<f64>,
         right_bound: Bound<f64>,
     ) -> RangeQuery {
-        let make_term_val = |val: &f64| Term::from_field_f64(field, *val).value_bytes().to_owned();
+        let make_term_val = |val: &f64| {
+            Term::from_field_f64(Field::from_field_id(0), *val)
+                .value_bytes()
+                .to_owned()
+        };
         RangeQuery {
             field,
             value_type: Type::F64,
@@ -179,11 +184,15 @@ impl RangeQuery {
     /// If the field is not of the type `u64`, tantivy
     /// will panic when the `Weight` object is created.
     pub fn new_u64_bounds(
-        field: Field,
+        field: String,
         left_bound: Bound<u64>,
         right_bound: Bound<u64>,
     ) -> RangeQuery {
-        let make_term_val = |val: &u64| Term::from_field_u64(field, *val).value_bytes().to_owned();
+        let make_term_val = |val: &u64| {
+            Term::from_field_u64(Field::from_field_id(0), *val)
+                .value_bytes()
+                .to_owned()
+        };
         RangeQuery {
             field,
             value_type: Type::U64,
@@ -196,7 +205,7 @@ impl RangeQuery {
     ///
     /// If the field is not of the type `u64`, tantivy
     /// will panic when the `Weight` object is created.
-    pub fn new_u64(field: Field, range: Range<u64>) -> RangeQuery {
+    pub fn new_u64(field: String, range: Range<u64>) -> RangeQuery {
         RangeQuery::new_u64_bounds(
             field,
             Bound::Included(range.start),
@@ -212,12 +221,15 @@ impl RangeQuery {
     /// If the field is not of the type `date`, tantivy
     /// will panic when the `Weight` object is created.
     pub fn new_date_bounds(
-        field: Field,
+        field: String,
         left_bound: Bound<DateTime>,
         right_bound: Bound<DateTime>,
     ) -> RangeQuery {
-        let make_term_val =
-            |val: &DateTime| Term::from_field_date(field, *val).value_bytes().to_owned();
+        let make_term_val = |val: &DateTime| {
+            Term::from_field_date(Field::from_field_id(0), *val)
+                .value_bytes()
+                .to_owned()
+        };
         RangeQuery {
             field,
             value_type: Type::Date,
@@ -230,7 +242,7 @@ impl RangeQuery {
     ///
     /// If the field is not of the type `date`, tantivy
     /// will panic when the `Weight` object is created.
-    pub fn new_date(field: Field, range: Range<DateTime>) -> RangeQuery {
+    pub fn new_date(field: String, range: Range<DateTime>) -> RangeQuery {
         RangeQuery::new_date_bounds(
             field,
             Bound::Included(range.start),
@@ -245,7 +257,7 @@ impl RangeQuery {
     ///
     /// If the field is not of the type `Str`, tantivy
     /// will panic when the `Weight` object is created.
-    pub fn new_str_bounds(field: Field, left: Bound<&str>, right: Bound<&str>) -> RangeQuery {
+    pub fn new_str_bounds(field: String, left: Bound<&str>, right: Bound<&str>) -> RangeQuery {
         let make_term_val = |val: &&str| val.as_bytes().to_vec();
         RangeQuery {
             field,
@@ -259,7 +271,7 @@ impl RangeQuery {
     ///
     /// If the field is not of the type `Str`, tantivy
     /// will panic when the `Weight` object is created.
-    pub fn new_str(field: Field, range: Range<&str>) -> RangeQuery {
+    pub fn new_str(field: String, range: Range<&str>) -> RangeQuery {
         RangeQuery::new_str_bounds(
             field,
             Bound::Included(range.start),
@@ -268,22 +280,8 @@ impl RangeQuery {
     }
 
     /// Field to search over
-    pub fn field(&self) -> Field {
-        self.field
-    }
-
-    /// Lower bound of range
-    pub fn left_bound(&self) -> Bound<Term> {
-        map_bound(&self.left_bound, &|bytes| {
-            Term::from_field_bytes(self.field, bytes)
-        })
-    }
-
-    /// Upper bound of range
-    pub fn right_bound(&self) -> Bound<Term> {
-        map_bound(&self.right_bound, &|bytes| {
-            Term::from_field_bytes(self.field, bytes)
-        })
+    pub fn field(&self) -> &str {
+        &self.field
     }
 }
 
@@ -307,7 +305,9 @@ pub(crate) fn maps_to_u64_fastfield(typ: Type) -> bool {
 impl Query for RangeQuery {
     fn weight(&self, enable_scoring: EnableScoring<'_>) -> crate::Result<Box<dyn Weight>> {
         let schema = enable_scoring.schema();
-        let field_type = schema.get_field_entry(self.field).field_type();
+        let field_type = schema
+            .get_field_entry(schema.get_field(&self.field)?)
+            .field_type();
         let value_type = field_type.value_type();
         if value_type != self.value_type {
             let err_msg = format!(
@@ -320,7 +320,7 @@ impl Query for RangeQuery {
         if field_type.is_fast() && is_type_valid_for_fastfield_range_query(self.value_type) {
             if field_type.is_ip_addr() {
                 Ok(Box::new(IPFastFieldRangeWeight::new(
-                    self.field,
+                    self.field.to_string(),
                     &self.left_bound,
                     &self.right_bound,
                 )))
@@ -335,14 +335,14 @@ impl Query for RangeQuery {
                 let left_bound = map_bound(&self.left_bound, &parse_from_bytes);
                 let right_bound = map_bound(&self.right_bound, &parse_from_bytes);
                 Ok(Box::new(FastFieldRangeWeight::new(
-                    self.field,
+                    self.field.to_string(),
                     left_bound,
                     right_bound,
                 )))
             }
         } else {
             Ok(Box::new(RangeWeight {
-                field: self.field,
+                field: self.field.to_string(),
                 left_bound: self.left_bound.clone(),
                 right_bound: self.right_bound.clone(),
             }))
@@ -351,7 +351,7 @@ impl Query for RangeQuery {
 }
 
 pub struct RangeWeight {
-    field: Field,
+    field: String,
     left_bound: Bound<Vec<u8>>,
     right_bound: Bound<Vec<u8>>,
 }
@@ -379,7 +379,7 @@ impl Weight for RangeWeight {
         let max_doc = reader.max_doc();
         let mut doc_bitset = BitSet::with_max_value(max_doc);
 
-        let inverted_index = reader.inverted_index(self.field)?;
+        let inverted_index = reader.inverted_index(reader.schema().get_field(&self.field)?)?;
         let term_dict = inverted_index.terms();
         let mut term_range = self.term_range(term_dict)?;
         while term_range.advance() {
@@ -443,7 +443,7 @@ mod tests {
         let reader = index.reader()?;
         let searcher = reader.searcher();
 
-        let docs_in_the_sixties = RangeQuery::new_u64(year_field, 1960u64..1970u64);
+        let docs_in_the_sixties = RangeQuery::new_u64("year".to_string(), 1960u64..1970u64);
 
         // ... or `1960..=1969` if inclusive range is enabled.
         let count = searcher.search(&docs_in_the_sixties, &Count)?;
@@ -481,10 +481,13 @@ mod tests {
         let count_multiples =
             |range_query: RangeQuery| searcher.search(&range_query, &Count).unwrap();
 
-        assert_eq!(count_multiples(RangeQuery::new_i64(int_field, 10..11)), 9);
+        assert_eq!(
+            count_multiples(RangeQuery::new_i64("intfield".to_string(), 10..11)),
+            9
+        );
         assert_eq!(
             count_multiples(RangeQuery::new_i64_bounds(
-                int_field,
+                "intfield".to_string(),
                 Bound::Included(10),
                 Bound::Included(11)
             )),
@@ -492,7 +495,7 @@ mod tests {
         );
         assert_eq!(
             count_multiples(RangeQuery::new_i64_bounds(
-                int_field,
+                "intfield".to_string(),
                 Bound::Excluded(9),
                 Bound::Included(10)
             )),
@@ -500,7 +503,7 @@ mod tests {
         );
         assert_eq!(
             count_multiples(RangeQuery::new_i64_bounds(
-                int_field,
+                "intfield".to_string(),
                 Bound::Included(9),
                 Bound::Unbounded
             )),
@@ -540,12 +543,12 @@ mod tests {
             |range_query: RangeQuery| searcher.search(&range_query, &Count).unwrap();
 
         assert_eq!(
-            count_multiples(RangeQuery::new_f64(float_field, 10.0..11.0)),
+            count_multiples(RangeQuery::new_f64("floatfield".to_string(), 10.0..11.0)),
             9
         );
         assert_eq!(
             count_multiples(RangeQuery::new_f64_bounds(
-                float_field,
+                "floatfield".to_string(),
                 Bound::Included(10.0),
                 Bound::Included(11.0)
             )),
@@ -553,7 +556,7 @@ mod tests {
         );
         assert_eq!(
             count_multiples(RangeQuery::new_f64_bounds(
-                float_field,
+                "floatfield".to_string(),
                 Bound::Excluded(9.0),
                 Bound::Included(10.0)
             )),
@@ -561,7 +564,7 @@ mod tests {
         );
         assert_eq!(
             count_multiples(RangeQuery::new_f64_bounds(
-                float_field,
+                "floatfield".to_string(),
                 Bound::Included(9.0),
                 Bound::Unbounded
             )),

src/query/range_query/range_query_ip_fastfield.rs

@@ -11,18 +11,18 @@ use fastfield_codecs::MonotonicallyMappableToU128;
 use super::fast_field_range_query::{FastFieldCardinality, RangeDocSet};
 use super::range_query::map_bound;
 use crate::query::{ConstScorer, Explanation, Scorer, Weight};
-use crate::schema::{Cardinality, Field};
+use crate::schema::Cardinality;
 use crate::{DocId, DocSet, Score, SegmentReader, TantivyError};
 
 /// `IPFastFieldRangeWeight` uses the ip address fast field to execute range queries.
 pub struct IPFastFieldRangeWeight {
-    field: Field,
+    field: String,
     left_bound: Bound<Ipv6Addr>,
     right_bound: Bound<Ipv6Addr>,
 }
 
 impl IPFastFieldRangeWeight {
-    pub fn new(field: Field, left_bound: &Bound<Vec<u8>>, right_bound: &Bound<Vec<u8>>) -> Self {
+    pub fn new(field: String, left_bound: &Bound<Vec<u8>>, right_bound: &Bound<Vec<u8>>) -> Self {
         let parse_ip_from_bytes = |data: &Vec<u8>| {
             let ip_u128: u128 =
                 u128::from_be(BinarySerializable::deserialize(&mut &data[..]).unwrap());
@@ -40,10 +40,13 @@ impl IPFastFieldRangeWeight {
 
 impl Weight for IPFastFieldRangeWeight {
     fn scorer(&self, reader: &SegmentReader, boost: Score) -> crate::Result<Box<dyn Scorer>> {
-        let field_type = reader.schema().get_field_entry(self.field).field_type();
+        let field_type = reader
+            .schema()
+            .get_field_entry(reader.schema().get_field(&self.field)?)
+            .field_type();
         match field_type.fastfield_cardinality().unwrap() {
             Cardinality::SingleValue => {
-                let ip_addr_fast_field = reader.fast_fields().ip_addr(self.field)?;
+                let ip_addr_fast_field = reader.fast_fields().ip_addr(&self.field)?;
                 let value_range = bound_to_value_range(
                     &self.left_bound,
                     &self.right_bound,
@@ -57,7 +60,7 @@ impl Weight for IPFastFieldRangeWeight {
                 Ok(Box::new(ConstScorer::new(docset, boost)))
             }
             Cardinality::MultiValues => {
-                let ip_addr_fast_field = reader.fast_fields().ip_addrs(self.field)?;
+                let ip_addr_fast_field = reader.fast_fields().ip_addrs(&self.field)?;
                 let value_range = bound_to_value_range(
                     &self.left_bound,
                     &self.right_bound,

src/query/range_query/range_query_u64_fastfield.rs

@@ -9,18 +9,18 @@ use fastfield_codecs::MonotonicallyMappableToU64;
 use super::fast_field_range_query::{FastFieldCardinality, RangeDocSet};
 use super::range_query::map_bound;
 use crate::query::{ConstScorer, Explanation, Scorer, Weight};
-use crate::schema::{Cardinality, Field};
+use crate::schema::Cardinality;
 use crate::{DocId, DocSet, Score, SegmentReader, TantivyError};
 
 /// `FastFieldRangeWeight` uses the fast field to execute range queries.
 pub struct FastFieldRangeWeight {
-    field: Field,
+    field: String,
     left_bound: Bound<u64>,
     right_bound: Bound<u64>,
 }
 
 impl FastFieldRangeWeight {
-    pub fn new(field: Field, left_bound: Bound<u64>, right_bound: Bound<u64>) -> Self {
+    pub fn new(field: String, left_bound: Bound<u64>, right_bound: Bound<u64>) -> Self {
         let left_bound = map_bound(&left_bound, &|val| *val);
         let right_bound = map_bound(&right_bound, &|val| *val);
         Self {
@@ -33,10 +33,13 @@ impl FastFieldRangeWeight {
 
 impl Weight for FastFieldRangeWeight {
     fn scorer(&self, reader: &SegmentReader, boost: Score) -> crate::Result<Box<dyn Scorer>> {
-        let field_type = reader.schema().get_field_entry(self.field).field_type();
+        let field_type = reader
+            .schema()
+            .get_field_entry(reader.schema().get_field(&self.field)?)
+            .field_type();
         match field_type.fastfield_cardinality().unwrap() {
             Cardinality::SingleValue => {
-                let fast_field = reader.fast_fields().u64_lenient(self.field)?;
+                let fast_field = reader.fast_fields().u64_lenient(&self.field)?;
                 let value_range = bound_to_value_range(
                     &self.left_bound,
                     &self.right_bound,
@@ -48,7 +51,7 @@ impl Weight for FastFieldRangeWeight {
                 Ok(Box::new(ConstScorer::new(docset, boost)))
             }
             Cardinality::MultiValues => {
-                let fast_field = reader.fast_fields().u64s_lenient(self.field)?;
+                let fast_field = reader.fast_fields().u64s_lenient(&self.field)?;
                 let value_range = bound_to_value_range(
                     &self.left_bound,
                     &self.right_bound,

src/query/term_query/term_query.rs

@@ -109,6 +109,7 @@ impl TermQuery {
         } else {
             IndexRecordOption::Basic
         };
+
         Ok(TermWeight::new(
             self.term.clone(),
             index_record_option,

src/reader/mod.rs

@@ -44,7 +44,7 @@ pub struct IndexReaderBuilder {
     index: Index,
     warmers: Vec<Weak<dyn Warmer>>,
     num_warming_threads: usize,
-    doc_store_cache_size: usize,
+    doc_store_cache_num_blocks: usize,
 }
 
 impl IndexReaderBuilder {
@@ -55,7 +55,7 @@ impl IndexReaderBuilder {
             index,
             warmers: Vec::new(),
             num_warming_threads: 1,
-            doc_store_cache_size: DOCSTORE_CACHE_CAPACITY,
+            doc_store_cache_num_blocks: DOCSTORE_CACHE_CAPACITY,
         }
     }
 
@@ -72,7 +72,7 @@ impl IndexReaderBuilder {
             searcher_generation_inventory.clone(),
         )?;
         let inner_reader = InnerIndexReader::new(
-            self.doc_store_cache_size,
+            self.doc_store_cache_num_blocks,
             self.index,
             warming_state,
             searcher_generation_inventory,
@@ -119,8 +119,11 @@ impl IndexReaderBuilder {
     ///
     /// The doc store readers cache by default DOCSTORE_CACHE_CAPACITY(100) decompressed blocks.
     #[must_use]
-    pub fn doc_store_cache_size(mut self, doc_store_cache_size: usize) -> IndexReaderBuilder {
-        self.doc_store_cache_size = doc_store_cache_size;
+    pub fn doc_store_cache_num_blocks(
+        mut self,
+        doc_store_cache_num_blocks: usize,
+    ) -> IndexReaderBuilder {
+        self.doc_store_cache_num_blocks = doc_store_cache_num_blocks;
         self
     }
 
@@ -151,7 +154,7 @@ impl TryInto<IndexReader> for IndexReaderBuilder {
 }
 
 struct InnerIndexReader {
-    doc_store_cache_size: usize,
+    doc_store_cache_num_blocks: usize,
     index: Index,
     warming_state: WarmingState,
     searcher: arc_swap::ArcSwap<SearcherInner>,
@@ -161,7 +164,7 @@ struct InnerIndexReader {
 
 impl InnerIndexReader {
     fn new(
-        doc_store_cache_size: usize,
+        doc_store_cache_num_blocks: usize,
         index: Index,
         warming_state: WarmingState,
         // The searcher_generation_inventory is not used as source, but as target to track the
@@ -172,13 +175,13 @@ impl InnerIndexReader {
 
         let searcher = Self::create_searcher(
             &index,
-            doc_store_cache_size,
+            doc_store_cache_num_blocks,
             &warming_state,
             &searcher_generation_counter,
             &searcher_generation_inventory,
         )?;
         Ok(InnerIndexReader {
-            doc_store_cache_size,
+            doc_store_cache_num_blocks,
             index,
             warming_state,
             searcher: ArcSwap::from(searcher),
@@ -214,7 +217,7 @@ impl InnerIndexReader {
 
     fn create_searcher(
         index: &Index,
-        doc_store_cache_size: usize,
+        doc_store_cache_num_blocks: usize,
         warming_state: &WarmingState,
         searcher_generation_counter: &Arc<AtomicU64>,
         searcher_generation_inventory: &Inventory<SearcherGeneration>,
@@ -232,7 +235,7 @@ impl InnerIndexReader {
             index.clone(),
             segment_readers,
             searcher_generation,
-            doc_store_cache_size,
+            doc_store_cache_num_blocks,
         )?);
 
         warming_state.warm_new_searcher_generation(&searcher.clone().into())?;
@@ -242,7 +245,7 @@ impl InnerIndexReader {
     fn reload(&self) -> crate::Result<()> {
         let searcher = Self::create_searcher(
             &self.index,
-            self.doc_store_cache_size,
+            self.doc_store_cache_num_blocks,
             &self.warming_state,
             &self.searcher_generation_counter,
             &self.searcher_generation_inventory,

src/schema/index_record_option.rs

@@ -49,4 +49,17 @@ impl IndexRecordOption {
             IndexRecordOption::WithFreqsAndPositions => true,
         }
     }
+
+    /// Downgrades to the next level if provided `IndexRecordOption` is unavailable.
+    pub fn downgrade(&self, other: IndexRecordOption) -> IndexRecordOption {
+        use IndexRecordOption::*;
+
+        match (other, self) {
+            (WithFreqsAndPositions, WithFreqsAndPositions) => WithFreqsAndPositions,
+            (WithFreqs, WithFreqs) => WithFreqs,
+            (WithFreqsAndPositions, WithFreqs) => WithFreqs,
+            (WithFreqs, WithFreqsAndPositions) => WithFreqs,
+            _ => Basic,
+        }
+    }
 }

src/schema/schema.rs

@@ -11,6 +11,7 @@ use super::ip_options::IpAddrOptions;
 use super::*;
 use crate::schema::bytes_options::BytesOptions;
 use crate::schema::field_type::ValueParsingError;
+use crate::TantivyError;
 
 /// Tantivy has a very strict schema.
 /// You need to specify in advance whether a field is indexed or not,
@@ -308,8 +309,12 @@ impl Schema {
     }
 
     /// Returns the field option associated with a given name.
-    pub fn get_field(&self, field_name: &str) -> Option<Field> {
-        self.0.fields_map.get(field_name).cloned()
+    pub fn get_field(&self, field_name: &str) -> crate::Result<Field> {
+        self.0
+            .fields_map
+            .get(field_name)
+            .cloned()
+            .ok_or_else(|| TantivyError::FieldNotFound(field_name.to_string()))
     }
 
     /// Create document from a named doc.
@@ -319,7 +324,7 @@ impl Schema {
     ) -> Result<Document, DocParsingError> {
         let mut document = Document::new();
         for (field_name, values) in named_doc.0 {
-            if let Some(field) = self.get_field(&field_name) {
+            if let Ok(field) = self.get_field(&field_name) {
                 for value in values {
                     document.add_field_value(field, value);
                 }
@@ -360,7 +365,7 @@ impl Schema {
     ) -> Result<Document, DocParsingError> {
         let mut doc = Document::default();
         for (field_name, json_value) in json_obj {
-            if let Some(field) = self.get_field(&field_name) {
+            if let Ok(field) = self.get_field(&field_name) {
                 let field_entry = self.get_field_entry(field);
                 let field_type = field_entry.field_type();
                 match json_value {

src/schema/term.rs

@@ -375,7 +375,8 @@ where B: AsRef<[u8]>
     ///
     /// Do NOT rely on this byte representation in the index.
     /// This value is likely to change in the future.
-    pub(crate) fn as_slice(&self) -> &[u8] {
+    #[inline(always)]
+    pub fn as_slice(&self) -> &[u8] {
         self.0.as_ref()
     }
 }

src/schema/text_options.rs

@@ -30,7 +30,7 @@ impl TextOptions {
         self.stored
     }
 
-    /// Returns true iff the value is a fast field.
+    /// Returns true if and only if the value is a fast field.
     pub fn is_fast(&self) -> bool {
         self.fast
     }

src/store/index/block.rs

@@ -90,7 +90,7 @@ impl CheckpointBlock {
             return Ok(());
         }
         let mut doc = read_u32_vint(data);
-        let mut start_offset = read_u32_vint(data) as usize;
+        let mut start_offset = VInt::deserialize_u64(data)? as usize;
         for _ in 0..len {
             let num_docs = read_u32_vint(data);
             let block_num_bytes = read_u32_vint(data) as usize;
@@ -147,6 +147,15 @@ mod tests {
         test_aux_ser_deser(&checkpoints)
     }
 
+    #[test]
+    fn test_block_serialize_large_byte_range() -> io::Result<()> {
+        let checkpoints = vec![Checkpoint {
+            doc_range: 10..12,
+            byte_range: 8_000_000_000..9_000_000_000,
+        }];
+        test_aux_ser_deser(&checkpoints)
+    }
+
     #[test]
     fn test_block_serialize() -> io::Result<()> {
         let offsets: Vec<usize> = (0..11).map(|i| i * i * i).collect();

src/store/mod.rs

@@ -4,8 +4,8 @@
 //! order to be handled in the `Store`.
 //!
 //! Internally, documents (or rather their stored fields) are serialized to a buffer.
-//! When the buffer exceeds 16K, the buffer is compressed using `brotli`, `LZ4` or `snappy`
-//! and the resulting block is written to disk.
+//! When the buffer exceeds `block_size` (defaults to 16K), the buffer is compressed using `brotli`,
+//! `LZ4` or `snappy` and the resulting block is written to disk.
 //!
 //! One can then request for a specific `DocId`.
 //! A skip list helps navigating to the right block,
@@ -28,8 +28,6 @@
 //! - at the segment level, the
 //! [`SegmentReader`'s `doc` method](../struct.SegmentReader.html#method.doc)
 //! - at the index level, the [`Searcher::doc()`](crate::Searcher::doc) method
-//!
-//! !
 
 mod compressors;
 mod decompressors;

src/store/reader.rs

@@ -114,7 +114,10 @@ impl Sum for CacheStats {
 
 impl StoreReader {
     /// Opens a store reader
-    pub fn open(store_file: FileSlice, cache_size: usize) -> io::Result<StoreReader> {
+    ///
+    /// `cache_num_blocks` sets the number of decompressed blocks to be cached in an LRU.
+    /// The size of blocks is configurable, this should be reflexted in the
+    pub fn open(store_file: FileSlice, cache_num_blocks: usize) -> io::Result<StoreReader> {
         let (footer, data_and_offset) = DocStoreFooter::extract_footer(store_file)?;
 
         let (data_file, offset_index_file) = data_and_offset.split(footer.offset as usize);
@@ -125,8 +128,8 @@ impl StoreReader {
             decompressor: footer.decompressor,
             data: data_file,
             cache: BlockCache {
-                cache: NonZeroUsize::new(cache_size)
-                    .map(|cache_size| Mutex::new(LruCache::new(cache_size))),
+                cache: NonZeroUsize::new(cache_num_blocks)
+                    .map(|cache_num_blocks| Mutex::new(LruCache::new(cache_num_blocks))),
                 cache_hits: Default::default(),
                 cache_misses: Default::default(),
             },

src/termdict/fst_termdict/term_info_store.rs

@@ -312,7 +312,7 @@ mod tests {
         bitpack.write(51, 6, &mut buffer).unwrap();
         assert_eq!(compute_num_bits(51), 6);
         bitpack.close(&mut buffer).unwrap();
-        assert_eq!(buffer.len(), 3 + 7);
+        assert_eq!(buffer.len(), 3);
         assert_eq!(extract_bits(&buffer[..], 0, 9), 321u64);
         assert_eq!(extract_bits(&buffer[..], 9, 2), 2u64);
         assert_eq!(extract_bits(&buffer[..], 11, 6), 51u64);

sstable/src/dictionary.rs

@@ -69,14 +69,16 @@ impl<TSSTable: SSTable> Dictionary<TSSTable> {
     pub(crate) fn sstable_delta_reader_for_key_range(
         &self,
         key_range: impl RangeBounds<[u8]>,
+        limit: Option<u64>,
     ) -> io::Result<DeltaReader<'static, TSSTable::ValueReader>> {
-        let slice = self.file_slice_for_range(key_range);
+        let slice = self.file_slice_for_range(key_range, limit);
         let data = slice.read_bytes()?;
         Ok(TSSTable::delta_reader(data))
     }
 
     /// This function returns a file slice covering a set of sstable blocks
-    /// that include the key range passed in arguments.
+    /// that include the key range passed in arguments. Optionally returns
+    /// only block for up to `limit` matching terms.
     ///
     /// It works by identifying
     /// - `first_block`: the block containing the start boudary key
@@ -92,26 +94,56 @@ impl<TSSTable: SSTable> Dictionary<TSSTable> {
     /// On the rare edge case where a user asks for `(start_key, end_key]`
     /// and `start_key` happens to be the last key of a block, we return a
     /// slice that is the first block was not necessary.
-    pub fn file_slice_for_range(&self, key_range: impl RangeBounds<[u8]>) -> FileSlice {
-        let start_bound: Bound<usize> = match key_range.start_bound() {
+    pub fn file_slice_for_range(
+        &self,
+        key_range: impl RangeBounds<[u8]>,
+        limit: Option<u64>,
+    ) -> FileSlice {
+        let first_block_id = match key_range.start_bound() {
             Bound::Included(key) | Bound::Excluded(key) => {
-                let Some(first_block_addr) = self.sstable_index.search_block(key) else {
+                let Some(first_block_id) = self.sstable_index.locate_with_key(key) else {
                     return FileSlice::empty();
                 };
-                Bound::Included(first_block_addr.byte_range.start)
+                Some(first_block_id)
             }
-            Bound::Unbounded => Bound::Unbounded,
+            Bound::Unbounded => None,
         };
-        let end_bound: Bound<usize> = match key_range.end_bound() {
-            Bound::Included(key) | Bound::Excluded(key) => {
-                if let Some(block_addr) = self.sstable_index.search_block(key) {
-                    Bound::Excluded(block_addr.byte_range.end)
+
+        let last_block_id = match key_range.end_bound() {
+            Bound::Included(key) | Bound::Excluded(key) => self.sstable_index.locate_with_key(key),
+            Bound::Unbounded => None,
+        };
+
+        let start_bound = if let Some(first_block_id) = first_block_id {
+            let Some(block_addr) = self.sstable_index.get_block(first_block_id) else {
+                return FileSlice::empty();
+            };
+            Bound::Included(block_addr.byte_range.start)
+        } else {
+            Bound::Unbounded
+        };
+
+        let last_block_id = if let Some(limit) = limit {
+            let second_block_id = first_block_id.map(|id| id + 1).unwrap_or(0);
+            if let Some(block_addr) = self.sstable_index.get_block(second_block_id) {
+                let ordinal_limit = block_addr.first_ordinal + limit;
+                let last_block_limit = self.sstable_index.locate_with_ord(ordinal_limit);
+                if let Some(last_block_id) = last_block_id {
+                    Some(last_block_id.min(last_block_limit))
                 } else {
-                    Bound::Unbounded
+                    Some(last_block_limit)
                 }
+            } else {
+                last_block_id
             }
-            Bound::Unbounded => Bound::Unbounded,
+        } else {
+            last_block_id
         };
+        let end_bound = last_block_id
+            .and_then(|block_id| self.sstable_index.get_block(block_id))
+            .map(|block_addr| Bound::Excluded(block_addr.byte_range.end))
+            .unwrap_or(Bound::Unbounded);
+
         self.sstable_slice.slice((start_bound, end_bound))
     }
 
@@ -156,10 +188,15 @@ impl<TSSTable: SSTable> Dictionary<TSSTable> {
 
     /// Returns the ordinal associated with a given term.
     pub fn term_ord<K: AsRef<[u8]>>(&self, key: K) -> io::Result<Option<TermOrdinal>> {
-        let mut term_ord = 0u64;
         let key_bytes = key.as_ref();
-        let mut sstable_reader = self.sstable_reader()?;
-        while sstable_reader.advance().unwrap_or(false) {
+
+        let Some(block_addr) = self.sstable_index.get_block_with_key(key_bytes) else {
+            return Ok(None);
+        };
+
+        let mut term_ord = block_addr.first_ordinal;
+        let mut sstable_reader = self.sstable_reader_block(block_addr)?;
+        while sstable_reader.advance()? {
             if sstable_reader.key() == key_bytes {
                 return Ok(Some(term_ord));
             }
@@ -178,22 +215,32 @@ impl<TSSTable: SSTable> Dictionary<TSSTable> {
     /// Regardless of whether the term is found or not,
     /// the buffer may be modified.
     pub fn ord_to_term(&self, ord: TermOrdinal, bytes: &mut Vec<u8>) -> io::Result<bool> {
-        let mut sstable_reader = self.sstable_reader()?;
-        bytes.clear();
-        for _ in 0..(ord + 1) {
-            if !sstable_reader.advance().unwrap_or(false) {
+        // find block in which the term would be
+        let block_addr = self.sstable_index.get_block_with_ord(ord);
+        let first_ordinal = block_addr.first_ordinal;
+
+        // then search inside that block only
+        let mut sstable_reader = self.sstable_reader_block(block_addr)?;
+        for _ in first_ordinal..=ord {
+            if !sstable_reader.advance()? {
                 return Ok(false);
             }
         }
+        bytes.clear();
         bytes.extend_from_slice(sstable_reader.key());
         Ok(true)
     }
 
     /// Returns the number of terms in the dictionary.
     pub fn term_info_from_ord(&self, term_ord: TermOrdinal) -> io::Result<Option<TSSTable::Value>> {
-        let mut sstable_reader = self.sstable_reader()?;
-        for _ in 0..(term_ord + 1) {
-            if !sstable_reader.advance().unwrap_or(false) {
+        // find block in which the term would be
+        let block_addr = self.sstable_index.get_block_with_ord(term_ord);
+        let first_ordinal = block_addr.first_ordinal;
+
+        // then search inside that block only
+        let mut sstable_reader = self.sstable_reader_block(block_addr)?;
+        for _ in first_ordinal..=term_ord {
+            if !sstable_reader.advance()? {
                 return Ok(None);
             }
         }
@@ -202,10 +249,10 @@ impl<TSSTable: SSTable> Dictionary<TSSTable> {
 
     /// Lookups the value corresponding to the key.
     pub fn get<K: AsRef<[u8]>>(&self, key: K) -> io::Result<Option<TSSTable::Value>> {
-        if let Some(block_addr) = self.sstable_index.search_block(key.as_ref()) {
+        if let Some(block_addr) = self.sstable_index.get_block_with_key(key.as_ref()) {
             let mut sstable_reader = self.sstable_reader_block(block_addr)?;
             let key_bytes = key.as_ref();
-            while sstable_reader.advance().unwrap_or(false) {
+            while sstable_reader.advance()? {
                 if sstable_reader.key() == key_bytes {
                     let value = sstable_reader.value().clone();
                     return Ok(Some(value));
@@ -217,10 +264,10 @@ impl<TSSTable: SSTable> Dictionary<TSSTable> {
 
     /// Lookups the value corresponding to the key.
     pub async fn get_async<K: AsRef<[u8]>>(&self, key: K) -> io::Result<Option<TSSTable::Value>> {
-        if let Some(block_addr) = self.sstable_index.search_block(key.as_ref()) {
+        if let Some(block_addr) = self.sstable_index.get_block_with_key(key.as_ref()) {
             let mut sstable_reader = self.sstable_reader_block_async(block_addr).await?;
             let key_bytes = key.as_ref();
-            while sstable_reader.advance().unwrap_or(false) {
+            while sstable_reader.advance()? {
                 if sstable_reader.key() == key_bytes {
                     let value = sstable_reader.value().clone();
                     return Ok(Some(value));
@@ -259,3 +306,192 @@ impl<TSSTable: SSTable> Dictionary<TSSTable> {
         Ok(())
     }
 }
+
+#[cfg(test)]
+mod tests {
+    use std::ops::Range;
+    use std::sync::{Arc, Mutex};
+
+    use common::OwnedBytes;
+
+    use super::Dictionary;
+    use crate::MonotonicU64SSTable;
+
+    #[derive(Debug)]
+    struct PermissionedHandle {
+        bytes: OwnedBytes,
+        allowed_range: Mutex<Range<usize>>,
+    }
+
+    impl PermissionedHandle {
+        fn new(bytes: Vec<u8>) -> Self {
+            let bytes = OwnedBytes::new(bytes);
+            PermissionedHandle {
+                allowed_range: Mutex::new(0..bytes.len()),
+                bytes,
+            }
+        }
+
+        fn restrict(&self, range: Range<usize>) {
+            *self.allowed_range.lock().unwrap() = range;
+        }
+    }
+
+    impl common::HasLen for PermissionedHandle {
+        fn len(&self) -> usize {
+            self.bytes.len()
+        }
+    }
+
+    impl common::file_slice::FileHandle for PermissionedHandle {
+        fn read_bytes(&self, range: Range<usize>) -> std::io::Result<OwnedBytes> {
+            let allowed_range = self.allowed_range.lock().unwrap();
+            if !allowed_range.contains(&range.start) || !allowed_range.contains(&(range.end - 1)) {
+                return Err(std::io::Error::new(
+                    std::io::ErrorKind::Other,
+                    format!("invalid range, allowed {allowed_range:?}, requested {range:?}"),
+                ));
+            }
+
+            Ok(self.bytes.slice(range))
+        }
+    }
+
+    fn make_test_sstable() -> (Dictionary<MonotonicU64SSTable>, Arc<PermissionedHandle>) {
+        let mut builder = Dictionary::<MonotonicU64SSTable>::builder(Vec::new()).unwrap();
+
+        // this makes 256k keys, enough to fill multiple blocks.
+        for elem in 0..0x3ffff {
+            let key = format!("{elem:05X}").into_bytes();
+            builder.insert_cannot_fail(&key, &elem);
+        }
+
+        let table = builder.finish().unwrap();
+        let table = Arc::new(PermissionedHandle::new(table));
+        let slice = common::file_slice::FileSlice::new(table.clone());
+
+        let dictionary = Dictionary::<MonotonicU64SSTable>::open(slice).unwrap();
+
+        // if the last block is id 0, tests are meaningless
+        assert_ne!(dictionary.sstable_index.locate_with_ord(u64::MAX), 0);
+        assert_eq!(dictionary.num_terms(), 0x3ffff);
+        (dictionary, table)
+    }
+
+    #[test]
+    fn test_ord_term_conversion() {
+        let (dic, slice) = make_test_sstable();
+
+        let block = dic.sstable_index.get_block_with_ord(100_000);
+        slice.restrict(block.byte_range);
+
+        let mut res = Vec::new();
+
+        // middle of a block
+        assert!(dic.ord_to_term(100_000, &mut res).unwrap());
+        assert_eq!(res, format!("{:05X}", 100_000).into_bytes());
+        assert_eq!(dic.term_info_from_ord(100_000).unwrap().unwrap(), 100_000);
+        assert_eq!(dic.get(&res).unwrap().unwrap(), 100_000);
+        assert_eq!(dic.term_ord(&res).unwrap().unwrap(), 100_000);
+
+        // start of a block
+        assert!(dic.ord_to_term(block.first_ordinal, &mut res).unwrap());
+        assert_eq!(res, format!("{:05X}", block.first_ordinal).into_bytes());
+        assert_eq!(
+            dic.term_info_from_ord(block.first_ordinal)
+                .unwrap()
+                .unwrap(),
+            block.first_ordinal
+        );
+        assert_eq!(dic.get(&res).unwrap().unwrap(), block.first_ordinal);
+        assert_eq!(dic.term_ord(&res).unwrap().unwrap(), block.first_ordinal);
+
+        // end of a block
+        let ordinal = block.first_ordinal - 1;
+        let new_range = dic.sstable_index.get_block_with_ord(ordinal).byte_range;
+        slice.restrict(new_range);
+        assert!(dic.ord_to_term(ordinal, &mut res).unwrap());
+        assert_eq!(res, format!("{:05X}", ordinal).into_bytes());
+        assert_eq!(dic.term_info_from_ord(ordinal).unwrap().unwrap(), ordinal);
+        assert_eq!(dic.get(&res).unwrap().unwrap(), ordinal);
+        assert_eq!(dic.term_ord(&res).unwrap().unwrap(), ordinal);
+
+        // before first block
+        // 1st block must be loaded for key-related operations
+        let block = dic.sstable_index.get_block_with_ord(0);
+        slice.restrict(block.byte_range);
+
+        assert!(dic.get(&b"$$$").unwrap().is_none());
+        assert!(dic.term_ord(&b"$$$").unwrap().is_none());
+
+        // after last block
+        // last block must be loaded for ord related operations
+        let ordinal = 0x40000 + 10;
+        let new_range = dic.sstable_index.get_block_with_ord(ordinal).byte_range;
+        slice.restrict(new_range);
+        assert!(!dic.ord_to_term(ordinal, &mut res).unwrap());
+        assert!(dic.term_info_from_ord(ordinal).unwrap().is_none());
+
+        // last block isn't required to be loaded for key related operations
+        slice.restrict(0..0);
+        assert!(dic.get(&b"~~~").unwrap().is_none());
+        assert!(dic.term_ord(&b"~~~").unwrap().is_none());
+    }
+
+    #[test]
+    fn test_range() {
+        let (dic, slice) = make_test_sstable();
+
+        let start = dic
+            .sstable_index
+            .get_block_with_key(b"10000")
+            .unwrap()
+            .byte_range;
+        let end = dic
+            .sstable_index
+            .get_block_with_key(b"18000")
+            .unwrap()
+            .byte_range;
+        slice.restrict(start.start..end.end);
+
+        let mut stream = dic.range().ge(b"10000").lt(b"18000").into_stream().unwrap();
+
+        for i in 0x10000..0x18000 {
+            assert!(stream.advance());
+            assert_eq!(stream.term_ord(), i);
+            assert_eq!(stream.value(), &i);
+            assert_eq!(stream.key(), format!("{i:05X}").into_bytes());
+        }
+        assert!(!stream.advance());
+
+        // verify limiting the number of results reduce the size read
+        slice.restrict(start.start..(end.end - 1));
+
+        let mut stream = dic
+            .range()
+            .ge(b"10000")
+            .lt(b"18000")
+            .limit(0xfff)
+            .into_stream()
+            .unwrap();
+
+        for i in 0x10000..0x10fff {
+            assert!(stream.advance());
+            assert_eq!(stream.term_ord(), i);
+            assert_eq!(stream.value(), &i);
+            assert_eq!(stream.key(), format!("{i:05X}").into_bytes());
+        }
+        // there might be more successful elements after, though how many is undefined
+
+        slice.restrict(0..slice.bytes.len());
+
+        let mut stream = dic.stream().unwrap();
+        for i in 0..0x3ffff {
+            assert!(stream.advance());
+            assert_eq!(stream.term_ord(), i);
+            assert_eq!(stream.value(), &i);
+            assert_eq!(stream.key(), format!("{i:05X}").into_bytes());
+        }
+        assert!(!stream.advance());
+    }
+}

sstable/src/sstable_index.rs

@@ -3,7 +3,7 @@ use std::ops::Range;
 
 use serde::{Deserialize, Serialize};
 
-use crate::{common_prefix_len, SSTableDataCorruption};
+use crate::{common_prefix_len, SSTableDataCorruption, TermOrdinal};
 
 #[derive(Default, Debug, Serialize, Deserialize)]
 pub struct SSTableIndex {
@@ -11,15 +11,61 @@ pub struct SSTableIndex {
 }
 
 impl SSTableIndex {
+    /// Load an index from its binary representation
     pub fn load(data: &[u8]) -> Result<SSTableIndex, SSTableDataCorruption> {
         ciborium::de::from_reader(data).map_err(|_| SSTableDataCorruption)
     }
 
-    pub fn search_block(&self, key: &[u8]) -> Option<BlockAddr> {
+    /// Get the [`BlockAddr`] of the requested block.
+    pub(crate) fn get_block(&self, block_id: usize) -> Option<BlockAddr> {
         self.blocks
-            .iter()
-            .find(|block| &block.last_key_or_greater[..] >= key)
-            .map(|block| block.block_addr.clone())
+            .get(block_id)
+            .map(|block_meta| block_meta.block_addr.clone())
+    }
+
+    /// Get the block id of the block that woudl contain `key`.
+    ///
+    /// Returns None if `key` is lexicographically after the last key recorded.
+    pub(crate) fn locate_with_key(&self, key: &[u8]) -> Option<usize> {
+        let pos = self
+            .blocks
+            .binary_search_by_key(&key, |block| &block.last_key_or_greater);
+        match pos {
+            Ok(pos) => Some(pos),
+            Err(pos) => {
+                if pos < self.blocks.len() {
+                    Some(pos)
+                } else {
+                    // after end of last block: no block matches
+                    None
+                }
+            }
+        }
+    }
+
+    /// Get the [`BlockAddr`] of the block that would contain `key`.
+    ///
+    /// Returns None if `key` is lexicographically after the last key recorded.
+    pub fn get_block_with_key(&self, key: &[u8]) -> Option<BlockAddr> {
+        self.locate_with_key(key).and_then(|id| self.get_block(id))
+    }
+
+    pub(crate) fn locate_with_ord(&self, ord: TermOrdinal) -> usize {
+        let pos = self
+            .blocks
+            .binary_search_by_key(&ord, |block| block.block_addr.first_ordinal);
+
+        match pos {
+            Ok(pos) => pos,
+            // Err(0) can't happen as the sstable starts with ordinal zero
+            Err(pos) => pos - 1,
+        }
+    }
+
+    /// Get the [`BlockAddr`] of the block containing the `ord`-th term.
+    pub(crate) fn get_block_with_ord(&self, ord: TermOrdinal) -> BlockAddr {
+        // locate_with_ord always returns an index within range
+        self.get_block(self.locate_with_ord(ord)).unwrap()
     }
 }
 
@@ -30,7 +76,7 @@ pub struct BlockAddr {
 }
 
 #[derive(Debug, Serialize, Deserialize)]
-struct BlockMeta {
+pub(crate) struct BlockMeta {
     /// Any byte string that is lexicographically greater or equal to
     /// the last key in the block,
     /// and yet strictly smaller than the first key in the next block.
@@ -98,26 +144,38 @@ mod tests {
     fn test_sstable_index() {
         let mut sstable_builder = SSTableIndexBuilder::default();
         sstable_builder.add_block(b"aaa", 10..20, 0u64);
-        sstable_builder.add_block(b"bbbbbbb", 20..30, 564);
+        sstable_builder.add_block(b"bbbbbbb", 20..30, 5u64);
         sstable_builder.add_block(b"ccc", 30..40, 10u64);
         sstable_builder.add_block(b"dddd", 40..50, 15u64);
         let mut buffer: Vec<u8> = Vec::new();
         sstable_builder.serialize(&mut buffer).unwrap();
         let sstable_index = SSTableIndex::load(&buffer[..]).unwrap();
         assert_eq!(
-            sstable_index.search_block(b"bbbde"),
+            sstable_index.get_block_with_key(b"bbbde"),
             Some(BlockAddr {
                 first_ordinal: 10u64,
                 byte_range: 30..40
             })
         );
+
+        assert_eq!(sstable_index.locate_with_key(b"aa").unwrap(), 0);
+        assert_eq!(sstable_index.locate_with_key(b"aaa").unwrap(), 0);
+        assert_eq!(sstable_index.locate_with_key(b"aab").unwrap(), 1);
+        assert_eq!(sstable_index.locate_with_key(b"ccc").unwrap(), 2);
+        assert!(sstable_index.locate_with_key(b"e").is_none());
+
+        assert_eq!(sstable_index.locate_with_ord(0), 0);
+        assert_eq!(sstable_index.locate_with_ord(1), 0);
+        assert_eq!(sstable_index.locate_with_ord(4), 0);
+        assert_eq!(sstable_index.locate_with_ord(5), 1);
+        assert_eq!(sstable_index.locate_with_ord(100), 3);
     }
 
     #[test]
     fn test_sstable_with_corrupted_data() {
         let mut sstable_builder = SSTableIndexBuilder::default();
         sstable_builder.add_block(b"aaa", 10..20, 0u64);
-        sstable_builder.add_block(b"bbbbbbb", 20..30, 564);
+        sstable_builder.add_block(b"bbbbbbb", 20..30, 5u64);
         sstable_builder.add_block(b"ccc", 30..40, 10u64);
         sstable_builder.add_block(b"dddd", 40..50, 15u64);
         let mut buffer: Vec<u8> = Vec::new();

sstable/src/streamer.rs

@@ -19,6 +19,7 @@ where
     automaton: A,
     lower: Bound<Vec<u8>>,
     upper: Bound<Vec<u8>>,
+    limit: Option<u64>,
 }
 
 fn bound_as_byte_slice(bound: &Bound<Vec<u8>>) -> Bound<&[u8]> {
@@ -41,6 +42,7 @@ where
             automaton,
             lower: Bound::Unbounded,
             upper: Bound::Unbounded,
+            limit: None,
         }
     }
 
@@ -68,24 +70,46 @@ where
         self
     }
 
+    /// Load no more data than what's required to to get `limit`
+    /// matching entries.
+    ///
+    /// The resulting [`Streamer`] can still return marginaly
+    /// more than `limit` elements.
+    pub fn limit(mut self, limit: u64) -> Self {
+        self.limit = Some(limit);
+        self
+    }
+
     /// Creates the stream corresponding to the range
     /// of terms defined using the `StreamerBuilder`.
     pub fn into_stream(self) -> io::Result<Streamer<'a, TSSTable, A>> {
         // TODO Optimize by skipping to the right first block.
         let start_state = self.automaton.start();
+
         let key_range = (
             bound_as_byte_slice(&self.lower),
             bound_as_byte_slice(&self.upper),
         );
+
+        let first_term = match &key_range.0 {
+            Bound::Included(key) | Bound::Excluded(key) => self
+                .term_dict
+                .sstable_index
+                .get_block_with_key(key)
+                .map(|block| block.first_ordinal)
+                .unwrap_or(0),
+            Bound::Unbounded => 0,
+        };
+
         let delta_reader = self
             .term_dict
-            .sstable_delta_reader_for_key_range(key_range)?;
+            .sstable_delta_reader_for_key_range(key_range, self.limit)?;
         Ok(Streamer {
             automaton: self.automaton,
             states: vec![start_state],
             delta_reader,
             key: Vec::new(),
-            term_ord: None,
+            term_ord: first_term.checked_sub(1),
             lower_bound: self.lower,
             upper_bound: self.upper,
         })

sstable/src/value/mod.rs

@@ -16,7 +16,7 @@ pub trait ValueReader: Default {
 
     /// Loads a block.
     ///
-    /// Returns the number of bytes that were written.
+    /// Returns the number of bytes that were read.
     fn load(&mut self, data: &[u8]) -> io::Result<usize>;
 }