4089: Use a bufreader and bufwriter everytime there is a grenad<file> r=curquiza a=irevoire

# Pull Request
Wrap all the files we give to a grenad in a `BufReader` or `BufWriter`.

The dump import I tried in the issue went from 2h to 10 minutes on my machine.

I also ran a bunch of benchmarks on my machine, and we're faster by a few seconds everywhere but nothing huge.

-----

The one thing I’m afraid about is if we used to get the inner file in a grenad and then do a read right after without a seek at the beginning of the file or a reopen.
Since we now use a bufreader our read would return the bytes one buffer later and probably completely corrupt what we were supposed to read.

From what I see, it looks like it works, but I may have missed something, I don't know much about this part of the codebase.

This issue should not arise on the bufwriter, though, because if we're not able to write the content of the buffer I ensured that the `into_inner` of the bufwriter should return an internal error.

## Related issue
Fixes #4087


Co-authored-by: Tamo <tamo@meilisearch.com>
This commit is contained in:
meili-bors[bot] 2023-10-11 14:27:00 +00:00 committed by GitHub
commit 67a678cfb6
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
21 changed files with 109 additions and 85 deletions

View File

@ -1,4 +1,5 @@
use std::fs::File; use std::fs::File;
use std::io::BufReader;
use std::{io, str}; use std::{io, str};
use obkv::KvReader; use obkv::KvReader;
@ -19,14 +20,14 @@ use crate::FieldId;
pub struct EnrichedDocumentsBatchReader<R> { pub struct EnrichedDocumentsBatchReader<R> {
documents: DocumentsBatchReader<R>, documents: DocumentsBatchReader<R>,
primary_key: String, primary_key: String,
external_ids: grenad::ReaderCursor<File>, external_ids: grenad::ReaderCursor<BufReader<File>>,
} }
impl<R: io::Read + io::Seek> EnrichedDocumentsBatchReader<R> { impl<R: io::Read + io::Seek> EnrichedDocumentsBatchReader<R> {
pub fn new( pub fn new(
documents: DocumentsBatchReader<R>, documents: DocumentsBatchReader<R>,
primary_key: String, primary_key: String,
external_ids: grenad::Reader<File>, external_ids: grenad::Reader<BufReader<File>>,
) -> Result<Self, Error> { ) -> Result<Self, Error> {
if documents.documents_count() as u64 == external_ids.len() { if documents.documents_count() as u64 == external_ids.len() {
Ok(EnrichedDocumentsBatchReader { Ok(EnrichedDocumentsBatchReader {
@ -75,7 +76,7 @@ pub struct EnrichedDocument<'a> {
pub struct EnrichedDocumentsBatchCursor<R> { pub struct EnrichedDocumentsBatchCursor<R> {
documents: DocumentsBatchCursor<R>, documents: DocumentsBatchCursor<R>,
primary_key: String, primary_key: String,
external_ids: grenad::ReaderCursor<File>, external_ids: grenad::ReaderCursor<BufReader<File>>,
} }
impl<R> EnrichedDocumentsBatchCursor<R> { impl<R> EnrichedDocumentsBatchCursor<R> {

View File

@ -1,5 +1,6 @@
use std::borrow::Cow; use std::borrow::Cow;
use std::fs::File; use std::fs::File;
use std::io::BufReader;
use grenad::CompressionType; use grenad::CompressionType;
use heed::types::ByteSlice; use heed::types::ByteSlice;
@ -30,7 +31,7 @@ pub struct FacetsUpdateBulk<'i> {
facet_type: FacetType, facet_type: FacetType,
field_ids: Vec<FieldId>, field_ids: Vec<FieldId>,
// None if level 0 does not need to be updated // None if level 0 does not need to be updated
new_data: Option<grenad::Reader<File>>, new_data: Option<grenad::Reader<BufReader<File>>>,
} }
impl<'i> FacetsUpdateBulk<'i> { impl<'i> FacetsUpdateBulk<'i> {
@ -38,7 +39,7 @@ impl<'i> FacetsUpdateBulk<'i> {
index: &'i Index, index: &'i Index,
field_ids: Vec<FieldId>, field_ids: Vec<FieldId>,
facet_type: FacetType, facet_type: FacetType,
new_data: grenad::Reader<File>, new_data: grenad::Reader<BufReader<File>>,
group_size: u8, group_size: u8,
min_level_size: u8, min_level_size: u8,
) -> FacetsUpdateBulk<'i> { ) -> FacetsUpdateBulk<'i> {
@ -187,7 +188,7 @@ impl<R: std::io::Read + std::io::Seek> FacetsUpdateBulkInner<R> {
&self, &self,
field_id: FieldId, field_id: FieldId,
txn: &RoTxn, txn: &RoTxn,
) -> Result<(Vec<grenad::Reader<File>>, RoaringBitmap)> { ) -> Result<(Vec<grenad::Reader<BufReader<File>>>, RoaringBitmap)> {
let mut all_docids = RoaringBitmap::new(); let mut all_docids = RoaringBitmap::new();
let subwriters = self.compute_higher_levels(txn, field_id, 32, &mut |bitmaps, _| { let subwriters = self.compute_higher_levels(txn, field_id, 32, &mut |bitmaps, _| {
for bitmap in bitmaps { for bitmap in bitmaps {
@ -259,7 +260,7 @@ impl<R: std::io::Read + std::io::Seek> FacetsUpdateBulkInner<R> {
field_id: u16, field_id: u16,
level: u8, level: u8,
handle_group: &mut dyn FnMut(&[RoaringBitmap], &'t [u8]) -> Result<()>, handle_group: &mut dyn FnMut(&[RoaringBitmap], &'t [u8]) -> Result<()>,
) -> Result<Vec<grenad::Reader<File>>> { ) -> Result<Vec<grenad::Reader<BufReader<File>>>> {
if level == 0 { if level == 0 {
self.read_level_0(rtxn, field_id, handle_group)?; self.read_level_0(rtxn, field_id, handle_group)?;
// Level 0 is already in the database // Level 0 is already in the database

View File

@ -1,5 +1,6 @@
use std::collections::HashMap; use std::collections::HashMap;
use std::fs::File; use std::fs::File;
use std::io::BufReader;
use heed::types::{ByteSlice, DecodeIgnore}; use heed::types::{ByteSlice, DecodeIgnore};
use heed::{BytesDecode, Error, RoTxn, RwTxn}; use heed::{BytesDecode, Error, RoTxn, RwTxn};
@ -34,14 +35,14 @@ pub struct FacetsUpdateIncremental<'i> {
index: &'i Index, index: &'i Index,
inner: FacetsUpdateIncrementalInner, inner: FacetsUpdateIncrementalInner,
facet_type: FacetType, facet_type: FacetType,
new_data: grenad::Reader<File>, new_data: grenad::Reader<BufReader<File>>,
} }
impl<'i> FacetsUpdateIncremental<'i> { impl<'i> FacetsUpdateIncremental<'i> {
pub fn new( pub fn new(
index: &'i Index, index: &'i Index,
facet_type: FacetType, facet_type: FacetType,
new_data: grenad::Reader<File>, new_data: grenad::Reader<BufReader<File>>,
group_size: u8, group_size: u8,
min_level_size: u8, min_level_size: u8,
max_group_size: u8, max_group_size: u8,

View File

@ -78,6 +78,7 @@ pub const FACET_MIN_LEVEL_SIZE: u8 = 5;
use std::collections::BTreeSet; use std::collections::BTreeSet;
use std::fs::File; use std::fs::File;
use std::io::BufReader;
use std::iter::FromIterator; use std::iter::FromIterator;
use charabia::normalizer::{Normalize, NormalizerOption}; use charabia::normalizer::{Normalize, NormalizerOption};
@ -108,13 +109,17 @@ pub struct FacetsUpdate<'i> {
index: &'i Index, index: &'i Index,
database: heed::Database<FacetGroupKeyCodec<ByteSliceRefCodec>, FacetGroupValueCodec>, database: heed::Database<FacetGroupKeyCodec<ByteSliceRefCodec>, FacetGroupValueCodec>,
facet_type: FacetType, facet_type: FacetType,
new_data: grenad::Reader<File>, new_data: grenad::Reader<BufReader<File>>,
group_size: u8, group_size: u8,
max_group_size: u8, max_group_size: u8,
min_level_size: u8, min_level_size: u8,
} }
impl<'i> FacetsUpdate<'i> { impl<'i> FacetsUpdate<'i> {
pub fn new(index: &'i Index, facet_type: FacetType, new_data: grenad::Reader<File>) -> Self { pub fn new(
index: &'i Index,
facet_type: FacetType,
new_data: grenad::Reader<BufReader<File>>,
) -> Self {
let database = match facet_type { let database = match facet_type {
FacetType::String => index FacetType::String => index
.facet_id_string_docids .facet_id_string_docids

View File

@ -1,4 +1,4 @@
use std::io::{Read, Seek}; use std::io::{BufWriter, Read, Seek};
use std::result::Result as StdResult; use std::result::Result as StdResult;
use std::{fmt, iter}; use std::{fmt, iter};
@ -35,7 +35,7 @@ pub fn enrich_documents_batch<R: Read + Seek>(
let (mut cursor, mut documents_batch_index) = reader.into_cursor_and_fields_index(); let (mut cursor, mut documents_batch_index) = reader.into_cursor_and_fields_index();
let mut external_ids = tempfile::tempfile().map(grenad::Writer::new)?; let mut external_ids = tempfile::tempfile().map(BufWriter::new).map(grenad::Writer::new)?;
let mut uuid_buffer = [0; uuid::fmt::Hyphenated::LENGTH]; let mut uuid_buffer = [0; uuid::fmt::Hyphenated::LENGTH];
// The primary key *field id* that has already been set for this index or the one // The primary key *field id* that has already been set for this index or the one

View File

@ -1,6 +1,7 @@
use std::collections::{HashMap, HashSet}; use std::collections::{HashMap, HashSet};
use std::convert::TryInto; use std::convert::TryInto;
use std::fs::File; use std::fs::File;
use std::io::BufReader;
use std::{io, mem, str}; use std::{io, mem, str};
use charabia::{Language, Script, SeparatorKind, Token, TokenKind, Tokenizer, TokenizerBuilder}; use charabia::{Language, Script, SeparatorKind, Token, TokenKind, Tokenizer, TokenizerBuilder};
@ -31,7 +32,7 @@ pub fn extract_docid_word_positions<R: io::Read + io::Seek>(
allowed_separators: Option<&[&str]>, allowed_separators: Option<&[&str]>,
dictionary: Option<&[&str]>, dictionary: Option<&[&str]>,
max_positions_per_attributes: Option<u32>, max_positions_per_attributes: Option<u32>,
) -> Result<(RoaringBitmap, grenad::Reader<File>, ScriptLanguageDocidsMap)> { ) -> Result<(RoaringBitmap, grenad::Reader<BufReader<File>>, ScriptLanguageDocidsMap)> {
puffin::profile_function!(); puffin::profile_function!();
let max_positions_per_attributes = max_positions_per_attributes let max_positions_per_attributes = max_positions_per_attributes

View File

@ -1,5 +1,5 @@
use std::fs::File; use std::fs::File;
use std::io; use std::io::{self, BufReader};
use heed::{BytesDecode, BytesEncode}; use heed::{BytesDecode, BytesEncode};
@ -19,7 +19,7 @@ use crate::Result;
pub fn extract_facet_number_docids<R: io::Read + io::Seek>( pub fn extract_facet_number_docids<R: io::Read + io::Seek>(
docid_fid_facet_number: grenad::Reader<R>, docid_fid_facet_number: grenad::Reader<R>,
indexer: GrenadParameters, indexer: GrenadParameters,
) -> Result<grenad::Reader<File>> { ) -> Result<grenad::Reader<BufReader<File>>> {
puffin::profile_function!(); puffin::profile_function!();
let max_memory = indexer.max_memory_by_thread(); let max_memory = indexer.max_memory_by_thread();

View File

@ -1,5 +1,5 @@
use std::fs::File; use std::fs::File;
use std::io; use std::io::{self, BufReader};
use heed::BytesEncode; use heed::BytesEncode;
@ -17,7 +17,7 @@ use crate::{FieldId, Result, MAX_FACET_VALUE_LENGTH};
pub fn extract_facet_string_docids<R: io::Read + io::Seek>( pub fn extract_facet_string_docids<R: io::Read + io::Seek>(
docid_fid_facet_string: grenad::Reader<R>, docid_fid_facet_string: grenad::Reader<R>,
indexer: GrenadParameters, indexer: GrenadParameters,
) -> Result<grenad::Reader<File>> { ) -> Result<grenad::Reader<BufReader<File>>> {
puffin::profile_function!(); puffin::profile_function!();
let max_memory = indexer.max_memory_by_thread(); let max_memory = indexer.max_memory_by_thread();

View File

@ -1,7 +1,7 @@
use std::collections::{BTreeMap, HashSet}; use std::collections::{BTreeMap, HashSet};
use std::convert::TryInto; use std::convert::TryInto;
use std::fs::File; use std::fs::File;
use std::io; use std::io::{self, BufReader};
use std::mem::size_of; use std::mem::size_of;
use heed::zerocopy::AsBytes; use heed::zerocopy::AsBytes;
@ -17,11 +17,11 @@ use crate::{CboRoaringBitmapCodec, DocumentId, FieldId, Result, BEU32, MAX_FACET
/// The extracted facet values stored in grenad files by type. /// The extracted facet values stored in grenad files by type.
pub struct ExtractedFacetValues { pub struct ExtractedFacetValues {
pub docid_fid_facet_numbers_chunk: grenad::Reader<File>, pub docid_fid_facet_numbers_chunk: grenad::Reader<BufReader<File>>,
pub docid_fid_facet_strings_chunk: grenad::Reader<File>, pub docid_fid_facet_strings_chunk: grenad::Reader<BufReader<File>>,
pub fid_facet_is_null_docids_chunk: grenad::Reader<File>, pub fid_facet_is_null_docids_chunk: grenad::Reader<BufReader<File>>,
pub fid_facet_is_empty_docids_chunk: grenad::Reader<File>, pub fid_facet_is_empty_docids_chunk: grenad::Reader<BufReader<File>>,
pub fid_facet_exists_docids_chunk: grenad::Reader<File>, pub fid_facet_exists_docids_chunk: grenad::Reader<BufReader<File>>,
} }
/// Extracts the facet values of each faceted field of each document. /// Extracts the facet values of each faceted field of each document.

View File

@ -1,6 +1,6 @@
use std::collections::HashMap; use std::collections::HashMap;
use std::fs::File; use std::fs::File;
use std::io; use std::io::{self, BufReader};
use grenad::Sorter; use grenad::Sorter;
@ -21,7 +21,7 @@ use crate::{relative_from_absolute_position, DocumentId, FieldId, Result};
pub fn extract_fid_word_count_docids<R: io::Read + io::Seek>( pub fn extract_fid_word_count_docids<R: io::Read + io::Seek>(
docid_word_positions: grenad::Reader<R>, docid_word_positions: grenad::Reader<R>,
indexer: GrenadParameters, indexer: GrenadParameters,
) -> Result<grenad::Reader<File>> { ) -> Result<grenad::Reader<BufReader<File>>> {
puffin::profile_function!(); puffin::profile_function!();
let max_memory = indexer.max_memory_by_thread(); let max_memory = indexer.max_memory_by_thread();

View File

@ -1,5 +1,5 @@
use std::fs::File; use std::fs::File;
use std::io; use std::io::{self, BufReader};
use concat_arrays::concat_arrays; use concat_arrays::concat_arrays;
use serde_json::Value; use serde_json::Value;
@ -18,7 +18,7 @@ pub fn extract_geo_points<R: io::Read + io::Seek>(
indexer: GrenadParameters, indexer: GrenadParameters,
primary_key_id: FieldId, primary_key_id: FieldId,
(lat_fid, lng_fid): (FieldId, FieldId), (lat_fid, lng_fid): (FieldId, FieldId),
) -> Result<grenad::Reader<File>> { ) -> Result<grenad::Reader<BufReader<File>>> {
puffin::profile_function!(); puffin::profile_function!();
let mut writer = create_writer( let mut writer = create_writer(

View File

@ -1,6 +1,6 @@
use std::convert::TryFrom; use std::convert::TryFrom;
use std::fs::File; use std::fs::File;
use std::io; use std::io::{self, BufReader};
use bytemuck::cast_slice; use bytemuck::cast_slice;
use serde_json::{from_slice, Value}; use serde_json::{from_slice, Value};
@ -18,7 +18,7 @@ pub fn extract_vector_points<R: io::Read + io::Seek>(
indexer: GrenadParameters, indexer: GrenadParameters,
primary_key_id: FieldId, primary_key_id: FieldId,
vectors_fid: FieldId, vectors_fid: FieldId,
) -> Result<grenad::Reader<File>> { ) -> Result<grenad::Reader<BufReader<File>>> {
puffin::profile_function!(); puffin::profile_function!();
let mut writer = create_writer( let mut writer = create_writer(

View File

@ -1,6 +1,6 @@
use std::collections::HashSet; use std::collections::HashSet;
use std::fs::File; use std::fs::File;
use std::io; use std::io::{self, BufReader};
use std::iter::FromIterator; use std::iter::FromIterator;
use roaring::RoaringBitmap; use roaring::RoaringBitmap;
@ -26,7 +26,7 @@ pub fn extract_word_docids<R: io::Read + io::Seek>(
docid_word_positions: grenad::Reader<R>, docid_word_positions: grenad::Reader<R>,
indexer: GrenadParameters, indexer: GrenadParameters,
exact_attributes: &HashSet<FieldId>, exact_attributes: &HashSet<FieldId>,
) -> Result<(grenad::Reader<File>, grenad::Reader<File>)> { ) -> Result<(grenad::Reader<BufReader<File>>, grenad::Reader<BufReader<File>>)> {
puffin::profile_function!(); puffin::profile_function!();
let max_memory = indexer.max_memory_by_thread(); let max_memory = indexer.max_memory_by_thread();

View File

@ -1,5 +1,5 @@
use std::fs::File; use std::fs::File;
use std::io; use std::io::{self, BufReader};
use super::helpers::{ use super::helpers::{
create_sorter, merge_cbo_roaring_bitmaps, read_u32_ne_bytes, sorter_into_reader, create_sorter, merge_cbo_roaring_bitmaps, read_u32_ne_bytes, sorter_into_reader,
@ -14,7 +14,7 @@ use crate::{relative_from_absolute_position, DocumentId, Result};
pub fn extract_word_fid_docids<R: io::Read + io::Seek>( pub fn extract_word_fid_docids<R: io::Read + io::Seek>(
docid_word_positions: grenad::Reader<R>, docid_word_positions: grenad::Reader<R>,
indexer: GrenadParameters, indexer: GrenadParameters,
) -> Result<grenad::Reader<File>> { ) -> Result<grenad::Reader<BufReader<File>>> {
puffin::profile_function!(); puffin::profile_function!();
let max_memory = indexer.max_memory_by_thread(); let max_memory = indexer.max_memory_by_thread();

View File

@ -1,6 +1,7 @@
use std::cmp::Ordering; use std::cmp::Ordering;
use std::collections::{BinaryHeap, HashMap}; use std::collections::{BinaryHeap, HashMap};
use std::fs::File; use std::fs::File;
use std::io::BufReader;
use std::{cmp, io, mem, str, vec}; use std::{cmp, io, mem, str, vec};
use super::helpers::{ use super::helpers::{
@ -20,7 +21,7 @@ use crate::{DocumentId, Result};
pub fn extract_word_pair_proximity_docids<R: io::Read + io::Seek>( pub fn extract_word_pair_proximity_docids<R: io::Read + io::Seek>(
docid_word_positions: grenad::Reader<R>, docid_word_positions: grenad::Reader<R>,
indexer: GrenadParameters, indexer: GrenadParameters,
) -> Result<grenad::Reader<File>> { ) -> Result<grenad::Reader<BufReader<File>>> {
puffin::profile_function!(); puffin::profile_function!();
let max_memory = indexer.max_memory_by_thread(); let max_memory = indexer.max_memory_by_thread();

View File

@ -1,5 +1,5 @@
use std::fs::File; use std::fs::File;
use std::io; use std::io::{self, BufReader};
use super::helpers::{ use super::helpers::{
create_sorter, merge_cbo_roaring_bitmaps, read_u32_ne_bytes, sorter_into_reader, create_sorter, merge_cbo_roaring_bitmaps, read_u32_ne_bytes, sorter_into_reader,
@ -17,7 +17,7 @@ use crate::{bucketed_position, relative_from_absolute_position, DocumentId, Resu
pub fn extract_word_position_docids<R: io::Read + io::Seek>( pub fn extract_word_position_docids<R: io::Read + io::Seek>(
docid_word_positions: grenad::Reader<R>, docid_word_positions: grenad::Reader<R>,
indexer: GrenadParameters, indexer: GrenadParameters,
) -> Result<grenad::Reader<File>> { ) -> Result<grenad::Reader<BufReader<File>>> {
puffin::profile_function!(); puffin::profile_function!();
let max_memory = indexer.max_memory_by_thread(); let max_memory = indexer.max_memory_by_thread();

View File

@ -12,6 +12,7 @@ mod extract_word_position_docids;
use std::collections::HashSet; use std::collections::HashSet;
use std::fs::File; use std::fs::File;
use std::io::BufReader;
use crossbeam_channel::Sender; use crossbeam_channel::Sender;
use log::debug; use log::debug;
@ -39,8 +40,8 @@ use crate::{FieldId, Result};
/// Send data in grenad file over provided Sender. /// Send data in grenad file over provided Sender.
#[allow(clippy::too_many_arguments)] #[allow(clippy::too_many_arguments)]
pub(crate) fn data_from_obkv_documents( pub(crate) fn data_from_obkv_documents(
original_obkv_chunks: impl Iterator<Item = Result<grenad::Reader<File>>> + Send, original_obkv_chunks: impl Iterator<Item = Result<grenad::Reader<BufReader<File>>>> + Send,
flattened_obkv_chunks: impl Iterator<Item = Result<grenad::Reader<File>>> + Send, flattened_obkv_chunks: impl Iterator<Item = Result<grenad::Reader<BufReader<File>>>> + Send,
indexer: GrenadParameters, indexer: GrenadParameters,
lmdb_writer_sx: Sender<Result<TypedChunk>>, lmdb_writer_sx: Sender<Result<TypedChunk>>,
searchable_fields: Option<HashSet<FieldId>>, searchable_fields: Option<HashSet<FieldId>>,
@ -152,7 +153,7 @@ pub(crate) fn data_from_obkv_documents(
}); });
} }
spawn_extraction_task::<_, _, Vec<grenad::Reader<File>>>( spawn_extraction_task::<_, _, Vec<grenad::Reader<BufReader<File>>>>(
docid_word_positions_chunks.clone(), docid_word_positions_chunks.clone(),
indexer, indexer,
lmdb_writer_sx.clone(), lmdb_writer_sx.clone(),
@ -162,7 +163,7 @@ pub(crate) fn data_from_obkv_documents(
"word-pair-proximity-docids", "word-pair-proximity-docids",
); );
spawn_extraction_task::<_, _, Vec<grenad::Reader<File>>>( spawn_extraction_task::<_, _, Vec<grenad::Reader<BufReader<File>>>>(
docid_word_positions_chunks.clone(), docid_word_positions_chunks.clone(),
indexer, indexer,
lmdb_writer_sx.clone(), lmdb_writer_sx.clone(),
@ -172,7 +173,11 @@ pub(crate) fn data_from_obkv_documents(
"field-id-wordcount-docids", "field-id-wordcount-docids",
); );
spawn_extraction_task::<_, _, Vec<(grenad::Reader<File>, grenad::Reader<File>)>>( spawn_extraction_task::<
_,
_,
Vec<(grenad::Reader<BufReader<File>>, grenad::Reader<BufReader<File>>)>,
>(
docid_word_positions_chunks.clone(), docid_word_positions_chunks.clone(),
indexer, indexer,
lmdb_writer_sx.clone(), lmdb_writer_sx.clone(),
@ -185,7 +190,7 @@ pub(crate) fn data_from_obkv_documents(
"word-docids", "word-docids",
); );
spawn_extraction_task::<_, _, Vec<grenad::Reader<File>>>( spawn_extraction_task::<_, _, Vec<grenad::Reader<BufReader<File>>>>(
docid_word_positions_chunks.clone(), docid_word_positions_chunks.clone(),
indexer, indexer,
lmdb_writer_sx.clone(), lmdb_writer_sx.clone(),
@ -194,7 +199,7 @@ pub(crate) fn data_from_obkv_documents(
TypedChunk::WordPositionDocids, TypedChunk::WordPositionDocids,
"word-position-docids", "word-position-docids",
); );
spawn_extraction_task::<_, _, Vec<grenad::Reader<File>>>( spawn_extraction_task::<_, _, Vec<grenad::Reader<BufReader<File>>>>(
docid_word_positions_chunks, docid_word_positions_chunks,
indexer, indexer,
lmdb_writer_sx.clone(), lmdb_writer_sx.clone(),
@ -204,7 +209,7 @@ pub(crate) fn data_from_obkv_documents(
"word-fid-docids", "word-fid-docids",
); );
spawn_extraction_task::<_, _, Vec<grenad::Reader<File>>>( spawn_extraction_task::<_, _, Vec<grenad::Reader<BufReader<File>>>>(
docid_fid_facet_strings_chunks, docid_fid_facet_strings_chunks,
indexer, indexer,
lmdb_writer_sx.clone(), lmdb_writer_sx.clone(),
@ -214,7 +219,7 @@ pub(crate) fn data_from_obkv_documents(
"field-id-facet-string-docids", "field-id-facet-string-docids",
); );
spawn_extraction_task::<_, _, Vec<grenad::Reader<File>>>( spawn_extraction_task::<_, _, Vec<grenad::Reader<BufReader<File>>>>(
docid_fid_facet_numbers_chunks, docid_fid_facet_numbers_chunks,
indexer, indexer,
lmdb_writer_sx, lmdb_writer_sx,
@ -269,7 +274,7 @@ fn spawn_extraction_task<FE, FS, M>(
/// Extract chunked data and send it into lmdb_writer_sx sender: /// Extract chunked data and send it into lmdb_writer_sx sender:
/// - documents /// - documents
fn send_original_documents_data( fn send_original_documents_data(
original_documents_chunk: Result<grenad::Reader<File>>, original_documents_chunk: Result<grenad::Reader<BufReader<File>>>,
indexer: GrenadParameters, indexer: GrenadParameters,
lmdb_writer_sx: Sender<Result<TypedChunk>>, lmdb_writer_sx: Sender<Result<TypedChunk>>,
vectors_field_id: Option<FieldId>, vectors_field_id: Option<FieldId>,
@ -311,7 +316,7 @@ fn send_original_documents_data(
#[allow(clippy::too_many_arguments)] #[allow(clippy::too_many_arguments)]
#[allow(clippy::type_complexity)] #[allow(clippy::type_complexity)]
fn send_and_extract_flattened_documents_data( fn send_and_extract_flattened_documents_data(
flattened_documents_chunk: Result<grenad::Reader<File>>, flattened_documents_chunk: Result<grenad::Reader<BufReader<File>>>,
indexer: GrenadParameters, indexer: GrenadParameters,
lmdb_writer_sx: Sender<Result<TypedChunk>>, lmdb_writer_sx: Sender<Result<TypedChunk>>,
searchable_fields: &Option<HashSet<FieldId>>, searchable_fields: &Option<HashSet<FieldId>>,
@ -328,7 +333,10 @@ fn send_and_extract_flattened_documents_data(
grenad::Reader<CursorClonableMmap>, grenad::Reader<CursorClonableMmap>,
( (
grenad::Reader<CursorClonableMmap>, grenad::Reader<CursorClonableMmap>,
(grenad::Reader<File>, (grenad::Reader<File>, grenad::Reader<File>)), (
grenad::Reader<BufReader<File>>,
(grenad::Reader<BufReader<File>>, grenad::Reader<BufReader<File>>),
),
), ),
), ),
)> { )> {

View File

@ -1,6 +1,6 @@
use std::borrow::Cow; use std::borrow::Cow;
use std::fs::File; use std::fs::File;
use std::io::{self, Seek}; use std::io::{self, BufReader, BufWriter, Seek};
use std::time::Instant; use std::time::Instant;
use grenad::{CompressionType, Sorter}; use grenad::{CompressionType, Sorter};
@ -17,13 +17,13 @@ pub fn create_writer<R: io::Write>(
typ: grenad::CompressionType, typ: grenad::CompressionType,
level: Option<u32>, level: Option<u32>,
file: R, file: R,
) -> grenad::Writer<R> { ) -> grenad::Writer<BufWriter<R>> {
let mut builder = grenad::Writer::builder(); let mut builder = grenad::Writer::builder();
builder.compression_type(typ); builder.compression_type(typ);
if let Some(level) = level { if let Some(level) = level {
builder.compression_level(level); builder.compression_level(level);
} }
builder.build(file) builder.build(BufWriter::new(file))
} }
pub fn create_sorter( pub fn create_sorter(
@ -53,7 +53,7 @@ pub fn create_sorter(
pub fn sorter_into_reader( pub fn sorter_into_reader(
sorter: grenad::Sorter<MergeFn>, sorter: grenad::Sorter<MergeFn>,
indexer: GrenadParameters, indexer: GrenadParameters,
) -> Result<grenad::Reader<File>> { ) -> Result<grenad::Reader<BufReader<File>>> {
let mut writer = create_writer( let mut writer = create_writer(
indexer.chunk_compression_type, indexer.chunk_compression_type,
indexer.chunk_compression_level, indexer.chunk_compression_level,
@ -64,16 +64,18 @@ pub fn sorter_into_reader(
writer_into_reader(writer) writer_into_reader(writer)
} }
pub fn writer_into_reader(writer: grenad::Writer<File>) -> Result<grenad::Reader<File>> { pub fn writer_into_reader(
let mut file = writer.into_inner()?; writer: grenad::Writer<BufWriter<File>>,
) -> Result<grenad::Reader<BufReader<File>>> {
let mut file = writer.into_inner()?.into_inner().map_err(|err| err.into_error())?;
file.rewind()?; file.rewind()?;
grenad::Reader::new(file).map_err(Into::into) grenad::Reader::new(BufReader::new(file)).map_err(Into::into)
} }
pub unsafe fn as_cloneable_grenad( pub unsafe fn as_cloneable_grenad(
reader: &grenad::Reader<File>, reader: &grenad::Reader<BufReader<File>>,
) -> Result<grenad::Reader<CursorClonableMmap>> { ) -> Result<grenad::Reader<CursorClonableMmap>> {
let file = reader.get_ref(); let file = reader.get_ref().get_ref();
let mmap = memmap2::Mmap::map(file)?; let mmap = memmap2::Mmap::map(file)?;
let cursor = io::Cursor::new(ClonableMmap::from(mmap)); let cursor = io::Cursor::new(ClonableMmap::from(mmap));
let reader = grenad::Reader::new(cursor)?; let reader = grenad::Reader::new(cursor)?;
@ -89,8 +91,8 @@ where
fn merge(self, merge_fn: MergeFn, indexer: &GrenadParameters) -> Result<Self::Output>; fn merge(self, merge_fn: MergeFn, indexer: &GrenadParameters) -> Result<Self::Output>;
} }
impl MergeableReader for Vec<grenad::Reader<File>> { impl MergeableReader for Vec<grenad::Reader<BufReader<File>>> {
type Output = grenad::Reader<File>; type Output = grenad::Reader<BufReader<File>>;
fn merge(self, merge_fn: MergeFn, params: &GrenadParameters) -> Result<Self::Output> { fn merge(self, merge_fn: MergeFn, params: &GrenadParameters) -> Result<Self::Output> {
let mut merger = MergerBuilder::new(merge_fn); let mut merger = MergerBuilder::new(merge_fn);
@ -99,8 +101,8 @@ impl MergeableReader for Vec<grenad::Reader<File>> {
} }
} }
impl MergeableReader for Vec<(grenad::Reader<File>, grenad::Reader<File>)> { impl MergeableReader for Vec<(grenad::Reader<BufReader<File>>, grenad::Reader<BufReader<File>>)> {
type Output = (grenad::Reader<File>, grenad::Reader<File>); type Output = (grenad::Reader<BufReader<File>>, grenad::Reader<BufReader<File>>);
fn merge(self, merge_fn: MergeFn, params: &GrenadParameters) -> Result<Self::Output> { fn merge(self, merge_fn: MergeFn, params: &GrenadParameters) -> Result<Self::Output> {
let mut m1 = MergerBuilder::new(merge_fn); let mut m1 = MergerBuilder::new(merge_fn);
@ -125,7 +127,7 @@ impl<R: io::Read + io::Seek> MergerBuilder<R> {
Ok(()) Ok(())
} }
fn finish(self, params: &GrenadParameters) -> Result<grenad::Reader<File>> { fn finish(self, params: &GrenadParameters) -> Result<grenad::Reader<BufReader<File>>> {
let merger = self.0.build(); let merger = self.0.build();
let mut writer = create_writer( let mut writer = create_writer(
params.chunk_compression_type, params.chunk_compression_type,
@ -176,7 +178,7 @@ pub fn grenad_obkv_into_chunks<R: io::Read + io::Seek>(
reader: grenad::Reader<R>, reader: grenad::Reader<R>,
indexer: GrenadParameters, indexer: GrenadParameters,
documents_chunk_size: usize, documents_chunk_size: usize,
) -> Result<impl Iterator<Item = Result<grenad::Reader<File>>>> { ) -> Result<impl Iterator<Item = Result<grenad::Reader<BufReader<File>>>>> {
let mut continue_reading = true; let mut continue_reading = true;
let mut cursor = reader.into_cursor()?; let mut cursor = reader.into_cursor()?;

View File

@ -659,8 +659,10 @@ impl<'a, 'i> Transform<'a, 'i> {
new_documents_ids: self.new_documents_ids, new_documents_ids: self.new_documents_ids,
replaced_documents_ids: self.replaced_documents_ids, replaced_documents_ids: self.replaced_documents_ids,
documents_count: self.documents_count, documents_count: self.documents_count,
original_documents, original_documents: original_documents.into_inner().map_err(|err| err.into_error())?,
flattened_documents, flattened_documents: flattened_documents
.into_inner()
.map_err(|err| err.into_error())?,
}) })
} }
@ -779,8 +781,10 @@ impl<'a, 'i> Transform<'a, 'i> {
new_documents_ids: documents_ids, new_documents_ids: documents_ids,
replaced_documents_ids: RoaringBitmap::default(), replaced_documents_ids: RoaringBitmap::default(),
documents_count, documents_count,
original_documents, original_documents: original_documents.into_inner().map_err(|err| err.into_error())?,
flattened_documents, flattened_documents: flattened_documents
.into_inner()
.map_err(|err| err.into_error())?,
}; };
let new_facets = output.compute_real_facets(wtxn, self.index)?; let new_facets = output.compute_real_facets(wtxn, self.index)?;

View File

@ -2,7 +2,7 @@ use std::borrow::Cow;
use std::collections::HashMap; use std::collections::HashMap;
use std::convert::TryInto; use std::convert::TryInto;
use std::fs::File; use std::fs::File;
use std::io; use std::io::{self, BufReader};
use bytemuck::allocation::pod_collect_to_vec; use bytemuck::allocation::pod_collect_to_vec;
use charabia::{Language, Script}; use charabia::{Language, Script};
@ -27,22 +27,22 @@ pub(crate) enum TypedChunk {
FieldIdDocidFacetStrings(grenad::Reader<CursorClonableMmap>), FieldIdDocidFacetStrings(grenad::Reader<CursorClonableMmap>),
FieldIdDocidFacetNumbers(grenad::Reader<CursorClonableMmap>), FieldIdDocidFacetNumbers(grenad::Reader<CursorClonableMmap>),
Documents(grenad::Reader<CursorClonableMmap>), Documents(grenad::Reader<CursorClonableMmap>),
FieldIdWordcountDocids(grenad::Reader<File>), FieldIdWordcountDocids(grenad::Reader<BufReader<File>>),
NewDocumentsIds(RoaringBitmap), NewDocumentsIds(RoaringBitmap),
WordDocids { WordDocids {
word_docids_reader: grenad::Reader<File>, word_docids_reader: grenad::Reader<BufReader<File>>,
exact_word_docids_reader: grenad::Reader<File>, exact_word_docids_reader: grenad::Reader<BufReader<File>>,
}, },
WordPositionDocids(grenad::Reader<File>), WordPositionDocids(grenad::Reader<BufReader<File>>),
WordFidDocids(grenad::Reader<File>), WordFidDocids(grenad::Reader<BufReader<File>>),
WordPairProximityDocids(grenad::Reader<File>), WordPairProximityDocids(grenad::Reader<BufReader<File>>),
FieldIdFacetStringDocids(grenad::Reader<File>), FieldIdFacetStringDocids(grenad::Reader<BufReader<File>>),
FieldIdFacetNumberDocids(grenad::Reader<File>), FieldIdFacetNumberDocids(grenad::Reader<BufReader<File>>),
FieldIdFacetExistsDocids(grenad::Reader<File>), FieldIdFacetExistsDocids(grenad::Reader<BufReader<File>>),
FieldIdFacetIsNullDocids(grenad::Reader<File>), FieldIdFacetIsNullDocids(grenad::Reader<BufReader<File>>),
FieldIdFacetIsEmptyDocids(grenad::Reader<File>), FieldIdFacetIsEmptyDocids(grenad::Reader<BufReader<File>>),
GeoPoints(grenad::Reader<File>), GeoPoints(grenad::Reader<BufReader<File>>),
VectorPoints(grenad::Reader<File>), VectorPoints(grenad::Reader<BufReader<File>>),
ScriptLanguageDocids(HashMap<(Script, Language), RoaringBitmap>), ScriptLanguageDocids(HashMap<(Script, Language), RoaringBitmap>),
} }

View File

@ -1,6 +1,6 @@
use std::borrow::Cow; use std::borrow::Cow;
use std::collections::HashSet; use std::collections::HashSet;
use std::io::BufReader; use std::io::{BufReader, BufWriter};
use grenad::CompressionType; use grenad::CompressionType;
use heed::types::ByteSlice; use heed::types::ByteSlice;
@ -119,9 +119,9 @@ pub fn insert_into_database(
pub fn write_into_lmdb_database_without_merging( pub fn write_into_lmdb_database_without_merging(
wtxn: &mut heed::RwTxn, wtxn: &mut heed::RwTxn,
database: heed::PolyDatabase, database: heed::PolyDatabase,
writer: grenad::Writer<std::fs::File>, writer: grenad::Writer<BufWriter<std::fs::File>>,
) -> Result<()> { ) -> Result<()> {
let file = writer.into_inner()?; let file = writer.into_inner()?.into_inner().map_err(|err| err.into_error())?;
let reader = grenad::Reader::new(BufReader::new(file))?; let reader = grenad::Reader::new(BufReader::new(file))?;
if database.is_empty(wtxn)? { if database.is_empty(wtxn)? {
let mut out_iter = database.iter_mut::<_, ByteSlice, ByteSlice>(wtxn)?; let mut out_iter = database.iter_mut::<_, ByteSlice, ByteSlice>(wtxn)?;