improve bucket size aggregation limit

postpone bucket size aggregation limit for terms aggregation
improve min_doc_count special use case 0, it loaded more texts from the
dict than segment_size limit

Note that we effectively check now that segment_size does not exceed the
bucket limit. segment_size is 10 * size. So requests would fail with
size 6500 for a bucket limit of 65000.

closes #1822

Cargo.toml

@@ -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

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/column_index/optional_index/set_block/tests.rs

@@ -1,7 +1,8 @@
 use std::collections::HashMap;
 
-use crate::column_index::optional_index::set_block::dense::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 {

columnar/src/columnar/column_type.rs

@@ -3,24 +3,22 @@ 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 {
     I64 = 0u8,
     U64 = 1u8,
     F64 = 2u8,
-    Bytes = 10u8,
-    Str = 14u8,
-    Bool = 18u8,
-    IpAddr = 22u8,
-    DateTime = 26u8,
+    Bytes = 3u8,
+    Str = 4u8,
+    Bool = 5u8,
+    IpAddr = 6u8,
+    DateTime = 7u8,
 }
 
-#[cfg(test)]
+// The order needs to match _exactly_ the order in the enum
 const COLUMN_TYPES: [ColumnType; 8] = [
     ColumnType::I64,
     ColumnType::U64,
@@ -38,18 +36,7 @@ impl ColumnType {
     }
 
     pub(crate) fn try_from_code(code: u8) -> Result<ColumnType, InvalidData> {
-        use ColumnType::*;
-        match code {
-            0u8 => Ok(I64),
-            1u8 => Ok(U64),
-            2u8 => Ok(F64),
-            10u8 => Ok(Bytes),
-            14u8 => Ok(Str),
-            18u8 => Ok(Bool),
-            22u8 => Ok(IpAddr),
-            26u8 => Ok(Self::DateTime),
-            _ => Err(InvalidData),
-        }
+        COLUMN_TYPES.get(code as usize).copied().ok_or(InvalidData)
     }
 }
 
@@ -64,18 +51,6 @@ impl From<NumericalType> for ColumnType {
 }
 
 impl ColumnType {
-    /// get column type category
-    pub(crate) fn column_type_category(self) -> ColumnTypeCategory {
-        match self {
-            ColumnType::I64 | ColumnType::U64 | ColumnType::F64 => ColumnTypeCategory::Numerical,
-            ColumnType::Bytes => ColumnTypeCategory::Bytes,
-            ColumnType::Str => ColumnTypeCategory::Str,
-            ColumnType::Bool => ColumnTypeCategory::Bool,
-            ColumnType::IpAddr => ColumnTypeCategory::IpAddr,
-            ColumnType::DateTime => ColumnTypeCategory::DateTime,
-        }
-    }
-
     pub fn numerical_type(&self) -> Option<NumericalType> {
         match self {
             ColumnType::I64 => Some(NumericalType::I64),
@@ -154,70 +129,20 @@ impl HasAssociatedColumnType for Ipv6Addr {
     }
 }
 
-/// 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, Hash, Debug)]
-#[repr(u8)]
-pub 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,
-        }
-    }
-}
-
 #[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(), super::COLUMN_TYPES.len());
-    }
-
-    #[test]
-    fn test_column_category_sort_consistent_with_column_type_sort() {
-        // This is a very important property because we
-        // we need to serialize colunmn in the right order.
-        let mut column_types: Vec<ColumnType> = super::COLUMN_TYPES.iter().copied().collect();
-        column_types.sort_by_key(|col| col.to_code());
-        let column_categories: Vec<ColumnTypeCategory> = column_types
-            .into_iter()
-            .map(ColumnTypeCategory::from)
-            .collect();
-        for (prev, next) in column_categories.iter().zip(column_categories.iter()) {
-            assert!(prev <= next);
+        for code in COLUMN_TYPES.len() as u8..=u8::MAX {
+            assert!(ColumnType::try_from_code(code as u8).is_err());
         }
     }
 

columnar/src/columnar/merge.rs

@@ -1,9 +1,9 @@
 use std::collections::HashMap;
 use std::io;
 
-use super::column_type::ColumnTypeCategory;
 use crate::columnar::ColumnarReader;
 use crate::dynamic_column::DynamicColumn;
+use crate::ColumnType;
 
 pub enum MergeDocOrder {
     /// Columnar tables are simply stacked one above the other.
@@ -35,7 +35,40 @@ pub fn merge_columnar(
     }
 }
 
-pub fn collect_columns(
+/// 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.
@@ -51,7 +84,7 @@ pub fn collect_columns(
                 .or_default();
 
             let columns = column_type_to_handles
-                .entry(handle.column_type().column_type_category())
+                .entry(handle.column_type().into())
                 .or_default();
             columns.push(handle.open()?);
         }
@@ -62,10 +95,9 @@ pub fn collect_columns(
     Ok(field_name_to_group)
 }
 
-/// Cast numerical type columns to the same type
-pub(crate) fn normalize_columns(
-    map: &mut HashMap<String, HashMap<ColumnTypeCategory, Vec<DynamicColumn>>>,
-) {
+/// 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 {

columnar/src/columnar/writer/column_writers.rs

@@ -184,10 +184,12 @@ impl CompatibleNumericalTypes {
 }
 
 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(

columnar/src/columnar/writer/mod.rs

@@ -15,7 +15,7 @@ use crate::column_index::SerializableColumnIndex;
 use crate::column_values::{
     ColumnValues, MonotonicallyMappableToU128, MonotonicallyMappableToU64, VecColumn,
 };
-use crate::columnar::column_type::{ColumnType, ColumnTypeCategory};
+use crate::columnar::column_type::ColumnType;
 use crate::columnar::writer::column_writers::{
     ColumnWriter, NumericalColumnWriter, StrOrBytesColumnWriter,
 };
@@ -276,35 +276,40 @@ impl ColumnarWriter {
     }
     pub fn serialize(&mut self, num_docs: RowId, wrt: &mut dyn io::Write) -> io::Result<()> {
         let mut serializer = ColumnarSerializer::new(wrt);
-        let mut columns: Vec<(&[u8], ColumnTypeCategory, Addr)> = self
+        let mut columns: Vec<(&[u8], ColumnType, Addr)> = self
             .numerical_field_hash_map
             .iter()
-            .map(|(column_name, addr, _)| (column_name, 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();
         columns.extend(
             self.bytes_field_hash_map
                 .iter()
-                .map(|(term, addr, _)| (term, ColumnTypeCategory::Bytes, addr)),
+                .map(|(term, addr, _)| (term, ColumnType::Bytes, addr)),
         );
         columns.extend(
             self.str_field_hash_map
                 .iter()
-                .map(|(column_name, addr, _)| (column_name, ColumnTypeCategory::Str, addr)),
+                .map(|(column_name, addr, _)| (column_name, ColumnType::Str, addr)),
         );
         columns.extend(
             self.bool_field_hash_map
                 .iter()
-                .map(|(column_name, addr, _)| (column_name, ColumnTypeCategory::Bool, addr)),
+                .map(|(column_name, addr, _)| (column_name, ColumnType::Bool, addr)),
         );
         columns.extend(
             self.ip_addr_field_hash_map
                 .iter()
-                .map(|(column_name, addr, _)| (column_name, ColumnTypeCategory::IpAddr, addr)),
+                .map(|(column_name, addr, _)| (column_name, ColumnType::IpAddr, addr)),
         );
         columns.extend(
             self.datetime_field_hash_map
                 .iter()
-                .map(|(column_name, addr, _)| (column_name, ColumnTypeCategory::DateTime, addr)),
+                .map(|(column_name, addr, _)| (column_name, ColumnType::DateTime, addr)),
         );
         columns.sort_unstable_by_key(|(column_name, col_type, _)| (*column_name, *col_type));
 
@@ -312,8 +317,12 @@ impl ColumnarWriter {
         let mut symbol_byte_buffer: Vec<u8> = Vec::new();
         for (column_name, column_type, addr) in columns {
             match column_type {
-                ColumnTypeCategory::Bool => {
-                    let column_writer: ColumnWriter = self.bool_field_hash_map.read(addr);
+                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);
@@ -325,7 +334,7 @@ impl ColumnarWriter {
                         &mut column_serializer,
                     )?;
                 }
-                ColumnTypeCategory::IpAddr => {
+                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 =
@@ -338,32 +347,35 @@ impl ColumnarWriter {
                         &mut column_serializer,
                     )?;
                 }
-                ColumnTypeCategory::Bytes | ColumnTypeCategory::Str => {
-                    let (column_type, str_column_writer): (ColumnType, StrOrBytesColumnWriter) =
-                        if column_type == ColumnTypeCategory::Bytes {
-                            (ColumnType::Bytes, self.bytes_field_hash_map.read(addr))
+                ColumnType::Bytes | ColumnType::Str => {
+                    let str_or_bytes_column_writer: StrOrBytesColumnWriter =
+                        if column_type == ColumnType::Bytes {
+                            self.bytes_field_hash_map.read(addr)
                         } else {
-                            (ColumnType::Str, self.str_field_hash_map.read(addr))
+                            self.str_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);
+                        &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_column_writer.operation_iterator(arena, &mut symbol_byte_buffer),
+                        str_or_bytes_column_writer
+                            .operation_iterator(arena, &mut symbol_byte_buffer),
                         buffers,
                         &mut column_serializer,
                     )?;
                 }
-                ColumnTypeCategory::Numerical => {
+                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 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(
@@ -375,20 +387,6 @@ impl ColumnarWriter {
                         &mut column_serializer,
                     )?;
                 }
-                ColumnTypeCategory::DateTime => {
-                    let column_writer: ColumnWriter = 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::DateTime);
-                    serialize_numerical_column(
-                        cardinality,
-                        num_docs,
-                        NumericalType::I64,
-                        column_writer.operation_iterator(arena, &mut symbol_byte_buffer),
-                        buffers,
-                        &mut column_serializer,
-                    )?;
-                }
             };
         }
         serializer.finalize()?;

src/aggregation/bucket/term_agg.rs

@@ -11,7 +11,7 @@ use crate::aggregation::agg_req_with_accessor::{
 use crate::aggregation::intermediate_agg_result::{
     IntermediateBucketResult, IntermediateTermBucketEntry, IntermediateTermBucketResult,
 };
-use crate::aggregation::segment_agg_result::{BucketCount, SegmentAggregationResultsCollector};
+use crate::aggregation::segment_agg_result::SegmentAggregationResultsCollector;
 use crate::error::DataCorruption;
 use crate::fastfield::MultiValuedFastFieldReader;
 use crate::schema::Type;
@@ -268,21 +268,18 @@ impl TermBuckets {
         term_ids: &[u64],
         doc: DocId,
         sub_aggregation: &AggregationsWithAccessor,
-        bucket_count: &BucketCount,
         blueprint: &Option<SegmentAggregationResultsCollector>,
     ) -> crate::Result<()> {
         for &term_id in term_ids {
-            let entry = self.entries.entry(term_id as u32).or_insert_with(|| {
-                bucket_count.add_count(1);
-
-                TermBucketEntry::from_blueprint(blueprint)
-            });
+            let entry = self
+                .entries
+                .entry(term_id as u32)
+                .or_insert_with(|| TermBucketEntry::from_blueprint(blueprint));
             entry.doc_count += 1;
             if let Some(sub_aggregations) = entry.sub_aggregations.as_mut() {
                 sub_aggregations.collect(doc, sub_aggregation)?;
             }
         }
-        bucket_count.validate_bucket_count()?;
 
         Ok(())
     }
@@ -372,7 +369,7 @@ impl SegmentTermCollector {
             }
             OrderTarget::SubAggregation(_name) => {
                 // don't sort and cut off since it's hard to make assumptions on the quality of the
-                // results when cutting off du to unknown nature of the sub_aggregation (possible
+                // results when cutting off due to unknown nature of the sub_aggregation (possible
                 // to check).
             }
             OrderTarget::Count => {
@@ -412,6 +409,10 @@ 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) {
@@ -433,6 +434,8 @@ impl SegmentTermCollector {
             sum_other_doc_count += sum_other_docs;
             dict = dict_entries.into_iter().collect();
         }
+        agg_with_accessor.bucket_count.add_count(dict.len() as u32);
+        agg_with_accessor.bucket_count.validate_bucket_count()?;
 
         Ok(IntermediateBucketResult::Terms(
             IntermediateTermBucketResult {
@@ -469,28 +472,24 @@ impl SegmentTermCollector {
                 &vals1,
                 docs[0],
                 &bucket_with_accessor.sub_aggregation,
-                &bucket_with_accessor.bucket_count,
                 &self.blueprint,
             )?;
             self.term_buckets.increment_bucket(
                 &vals2,
                 docs[1],
                 &bucket_with_accessor.sub_aggregation,
-                &bucket_with_accessor.bucket_count,
                 &self.blueprint,
             )?;
             self.term_buckets.increment_bucket(
                 &vals3,
                 docs[2],
                 &bucket_with_accessor.sub_aggregation,
-                &bucket_with_accessor.bucket_count,
                 &self.blueprint,
             )?;
             self.term_buckets.increment_bucket(
                 &vals4,
                 docs[3],
                 &bucket_with_accessor.sub_aggregation,
-                &bucket_with_accessor.bucket_count,
                 &self.blueprint,
             )?;
         }
@@ -501,7 +500,6 @@ impl SegmentTermCollector {
                 &vals1,
                 doc,
                 &bucket_with_accessor.sub_aggregation,
-                &bucket_with_accessor.bucket_count,
                 &self.blueprint,
             )?;
         }
@@ -1136,6 +1134,33 @@ mod tests {
         assert_eq!(res["my_texts"]["sum_other_doc_count"], 0);
         assert_eq!(res["my_texts"]["doc_count_error_upper_bound"], 0);
 
+        let agg_req: Aggregations = vec![(
+            "my_texts".to_string(),
+            Aggregation::Bucket(BucketAggregation {
+                bucket_agg: BucketAggregationType::Terms(TermsAggregation {
+                    field: "string_id".to_string(),
+                    min_doc_count: Some(0),
+                    size: Some(1),
+                    ..Default::default()
+                }),
+                sub_aggregation: Default::default(),
+            }),
+        )]
+        .into_iter()
+        .collect();
+
+        // searching for terma, but min_doc_count will return all terms
+        let res = exec_request_with_query(agg_req, &index, Some(("string_id", "terma")))?;
+
+        assert_eq!(res["my_texts"]["buckets"][0]["key"], "terma");
+        assert_eq!(res["my_texts"]["buckets"][0]["doc_count"], 4);
+        assert_eq!(
+            res["my_texts"]["buckets"][1]["key"],
+            serde_json::Value::Null
+        );
+        assert_eq!(res["my_texts"]["sum_other_doc_count"], 0);
+        assert_eq!(res["my_texts"]["doc_count_error_upper_bound"], 0);
+
         Ok(())
     }
 
@@ -1214,6 +1239,27 @@ mod tests {
 
         let index = get_test_index_from_terms(true, &terms_per_segment)?;
 
+        let agg_req: Aggregations = vec![(
+            "my_texts".to_string(),
+            Aggregation::Bucket(BucketAggregation {
+                bucket_agg: BucketAggregationType::Terms(TermsAggregation {
+                    field: "string_id".to_string(),
+                    // min_doc_count: Some(0),
+                    ..Default::default()
+                }),
+                sub_aggregation: Default::default(),
+            }),
+        )]
+        .into_iter()
+        .collect();
+
+        let res = exec_request_with_query(agg_req, &index, None);
+
+        assert!(res.is_ok());
+
+        // This request has min_doc_count set to 0
+        // That means we load potentially the whole dict
+        // Make sure the bucket count is still fine
         let agg_req: Aggregations = vec![(
             "my_texts".to_string(),
             Aggregation::Bucket(BucketAggregation {
@@ -1228,6 +1274,24 @@ mod tests {
         .into_iter()
         .collect();
 
+        let res = exec_request_with_query(agg_req, &index, None);
+        assert!(res.is_ok());
+
+        let agg_req: Aggregations = vec![(
+            "my_texts".to_string(),
+            Aggregation::Bucket(BucketAggregation {
+                bucket_agg: BucketAggregationType::Terms(TermsAggregation {
+                    field: "string_id".to_string(),
+                    // min_doc_count: Some(0),
+                    size: Some(70_000),
+                    ..Default::default()
+                }),
+                sub_aggregation: Default::default(),
+            }),
+        )]
+        .into_iter()
+        .collect();
+
         let res = exec_request_with_query(agg_req, &index, None);
 
         assert!(res.is_err());
@@ -1384,14 +1448,10 @@ mod bench {
         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)
+                    .increment_bucket(&[val], 0, &aggregations_with_accessor, &None)
                     .unwrap();
             }
         })