greptimedb/src/mito2/src/compaction/reader.rs

// Copyright 2023 Greptime Team
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
//     http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.

use std::collections::{BTreeMap, HashMap};
use std::sync::Arc;

use common_time::Timestamp;
use common_time::range::TimestampRange;
use common_time::timestamp::TimeUnit;
use datafusion_common::ScalarValue;
use datafusion_expr::Expr;
use datatypes::extension::json::is_json2_extension_type;
use datatypes::types::json_type::JsonNativeType;
use parquet::arrow::parquet_to_arrow_schema;
use parquet::file::metadata::{PageIndexPolicy, ParquetMetaData};
use snafu::{OptionExt, ResultExt};
use store_api::metadata::RegionMetadataRef;

use crate::access_layer::AccessLayerRef;
use crate::cache::{CacheManagerRef, CacheStrategy};
use crate::error::{
    DataTypeMismatchSnafu, ParquetToArrowSchemaSnafu, Result, TimeRangePredicateOverflowSnafu,
};
use crate::read::FlatSource;
use crate::read::flat_projection::FlatProjectionMapper;
use crate::read::read_columns::ReadColumns;
use crate::read::scan_region::{PredicateGroup, ScanInput};
use crate::read::seq_scan::SeqScan;
use crate::region::options::MergeMode;
use crate::sst::file::FileHandle;
use crate::sst::parquet::reader::MetadataCacheMetrics;

/// Builders to create [BoxedRecordBatchStream] for compaction.
pub(crate) struct CompactionSstReaderBuilder<'a> {
    pub(crate) metadata: RegionMetadataRef,
    pub(crate) sst_layer: AccessLayerRef,
    pub(crate) cache: CacheManagerRef,
    pub(crate) inputs: &'a [FileHandle],
    pub(crate) append_mode: bool,
    pub(crate) filter_deleted: bool,
    pub(crate) time_range: Option<TimestampRange>,
    pub(crate) merge_mode: MergeMode,
}

impl CompactionSstReaderBuilder<'_> {
    /// Build a [FlatSource] that yields Arrow `RecordBatch`s from reading all the input SST files,
    /// for compaction. The schema of the [FlatSource] is unified.
    pub(crate) async fn build_flat_sst_reader(self) -> Result<FlatSource> {
        let scan_input = self.build_scan_input().await?;

        let schema = scan_input.mapper.output_schema();
        let schema = schema.arrow_schema();

        let stream = SeqScan::new(scan_input)
            .build_flat_reader_for_compaction()
            .await?;
        Ok(FlatSource::new_stream(schema.clone(), stream))
    }

    async fn build_scan_input(self) -> Result<ScanInput> {
        let schema = self.metadata.schema.arrow_schema();
        let parquet_metadata = self.collect_parquet_metadata().await?;
        let batch_size = crate::batch_size::estimate_batch_size(
            parquet_metadata
                .iter()
                .flat_map(|metadata| metadata.row_groups())
                .map(|row_group| {
                    let uncompressed_bytes = row_group
                        .columns()
                        .iter()
                        .map(|column| column.uncompressed_size() as u64)
                        .sum();
                    (row_group.num_rows() as u64, uncompressed_bytes)
                }),
        );
        let json_type_hint = if schema.fields().iter().any(is_json2_extension_type) {
            let mut json_type_hint = schema
                .fields()
                .iter()
                .filter(|&field| is_json2_extension_type(field))
                .map(|field| (field.name().clone(), JsonNativeType::Null))
                .collect::<HashMap<_, _>>();

            for metadata in &parquet_metadata {
                let file_metadata = metadata.file_metadata();
                let schema = parquet_to_arrow_schema(
                    file_metadata.schema_descr(),
                    file_metadata.key_value_metadata(),
                )
                .context(ParquetToArrowSchemaSnafu {
                    file: "compaction input",
                })?;
                for field in schema.fields() {
                    let Some(merged) = json_type_hint.get_mut(field.name()) else {
                        continue;
                    };

                    let json_type = JsonNativeType::try_from(field.data_type())
                        .context(DataTypeMismatchSnafu)?;
                    merged.merge(&json_type);
                }
            }

            Some(json_type_hint)
        } else {
            None
        };

        let projection = (0..self.metadata.column_metadatas.len()).collect();
        let read_column_ids = self
            .metadata
            .column_metadatas
            .iter()
            .map(|x| x.column_id)
            .collect::<Vec<_>>();
        let json_target_types = json_type_hint
            .as_ref()
            .map(|hint| {
                hint.iter()
                    .filter_map(|(col_name, json_type)| {
                        self.metadata
                            .column_by_name(col_name)
                            .map(|col| (col.column_id, json_type.clone()))
                    })
                    .collect::<BTreeMap<_, _>>()
            })
            .unwrap_or_default();
        let read_columns =
            ReadColumns::new(read_column_ids).with_json_target_types(json_target_types);
        let mapper =
            FlatProjectionMapper::new_with_read_columns(&self.metadata, projection, read_columns)?;

        let mut scan_input = ScanInput::new(self.sst_layer, mapper)
            .with_files(self.inputs.to_vec())
            .with_compaction(true)
            .with_batch_size(batch_size)
            .with_append_mode(self.append_mode)
            // We use special cache strategy for compaction.
            .with_cache(CacheStrategy::Compaction(self.cache))
            .with_filter_deleted(self.filter_deleted)
            // We ignore file not found error during compaction.
            .with_ignore_file_not_found(true)
            .with_merge_mode(self.merge_mode);

        // This serves as a workaround of https://github.com/GreptimeTeam/greptimedb/issues/3944
        // by converting time ranges into predicate.
        if let Some(time_range) = self.time_range {
            scan_input =
                scan_input.with_predicate(time_range_to_predicate(time_range, &self.metadata)?);
        }

        Ok(scan_input)
    }

    async fn collect_parquet_metadata(&self) -> Result<Vec<Arc<ParquetMetaData>>> {
        let mut metadata = Vec::with_capacity(self.inputs.len());

        for file_handle in self.inputs {
            let file_path =
                file_handle.file_path(self.sst_layer.table_dir(), self.sst_layer.path_type());
            let file_size = file_handle.meta_ref().file_size;
            let parquet_metadata = match self
                .sst_layer
                .read_sst(file_handle.clone())
                .cache(CacheStrategy::Compaction(self.cache.clone()))
                .read_parquet_metadata(
                    &file_path,
                    file_size,
                    &mut MetadataCacheMetrics::default(),
                    PageIndexPolicy::default(),
                )
                .await
                .map(|x| x.0.parquet_metadata())
            {
                Ok(x) => x,
                Err(e) if e.is_object_not_found() => continue,
                Err(e) => return Err(e),
            };
            metadata.push(parquet_metadata);
        }
        Ok(metadata)
    }
}

/// Converts time range to predicates so that rows outside the range will be filtered.
fn time_range_to_predicate(
    range: TimestampRange,
    metadata: &RegionMetadataRef,
) -> Result<PredicateGroup> {
    let ts_col = metadata.time_index_column();

    // safety: time index column's type must be a valid timestamp type.
    let ts_col_unit = ts_col
        .column_schema
        .data_type
        .as_timestamp()
        .unwrap()
        .unit();

    let exprs = match (range.start(), range.end()) {
        (Some(start), Some(end)) => {
            vec![
                datafusion_expr::col(ts_col.column_schema.name.clone())
                    .gt_eq(ts_to_lit(*start, ts_col_unit)?),
                datafusion_expr::col(ts_col.column_schema.name.clone())
                    .lt(ts_to_lit(*end, ts_col_unit)?),
            ]
        }
        (Some(start), None) => {
            vec![
                datafusion_expr::col(ts_col.column_schema.name.clone())
                    .gt_eq(ts_to_lit(*start, ts_col_unit)?),
            ]
        }

        (None, Some(end)) => {
            vec![
                datafusion_expr::col(ts_col.column_schema.name.clone())
                    .lt(ts_to_lit(*end, ts_col_unit)?),
            ]
        }
        (None, None) => {
            return Ok(PredicateGroup::default());
        }
    };

    let predicate = PredicateGroup::new(metadata, &exprs)?;
    Ok(predicate)
}

fn ts_to_lit(ts: Timestamp, ts_col_unit: TimeUnit) -> Result<Expr> {
    let ts = ts
        .convert_to(ts_col_unit)
        .context(TimeRangePredicateOverflowSnafu {
            timestamp: ts,
            unit: ts_col_unit,
        })?;
    let val = ts.value();
    let scalar_value = match ts_col_unit {
        TimeUnit::Second => ScalarValue::TimestampSecond(Some(val), None),
        TimeUnit::Millisecond => ScalarValue::TimestampMillisecond(Some(val), None),
        TimeUnit::Microsecond => ScalarValue::TimestampMicrosecond(Some(val), None),
        TimeUnit::Nanosecond => ScalarValue::TimestampNanosecond(Some(val), None),
    };
    Ok(datafusion_expr::lit(scalar_value))
}