searchengines-web/MeiliSearch
searchengines-web/MeiliSearch
1
0
Fork 0
mirror of https://github.com/meilisearch/MeiliSearch synced 2025-01-11 05:54:30 +01:00
Code Issues Releases Wiki Activity

Merge #225

Browse Source 
225: Introduce the error handler r=ManyTheFish a=Kerollmops

Fixes #109.

Co-authored-by: Kerollmops <clement@meilisearch.com>
Co-authored-by: many <maxime@meilisearch.com>
This commit is contained in:
bors[bot] 2021-06-16 09:46:23 +00:00 committed by GitHub
commit 02e0271e44
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
44 changed files with 1186 additions and 703 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
Cargo.lock generated
View File

@ -1377,7 +1377,6 @@ dependencies = [
name = "milli"
version = "0.3.1"
dependencies = [
"anyhow",
"big_s",
"bstr",
"byteorder",

8
benchmarks/benches/utils.rs
View File

@ -55,15 +55,15 @@ pub fn base_setup(conf: &Conf) -> Index {
options.map_size(100 * 1024 * 1024 * 1024); // 100 GB
options.max_readers(10);
let index = Index::new(options, conf.database_name).unwrap();
if let Some(primary_key) = conf.primary_key {
let mut wtxn = index.write_txn().unwrap();
index.put_primary_key(&mut wtxn, primary_key).unwrap();
}
let update_builder = UpdateBuilder::new(0);
let mut wtxn = index.write_txn().unwrap();
let mut builder = update_builder.settings(&mut wtxn, &index);
if let Some(primary_key) = conf.primary_key {
builder.set_primary_key(primary_key.to_string());
}
if let Some(criterion) = conf.criterion {
builder.reset_filterable_fields();
builder.reset_criteria();

8
http-ui/src/main.rs
View File

@ -391,7 +391,7 @@ async fn main() -> anyhow::Result<()> {
match result {
Ok(_) => wtxn.commit().map_err(Into::into),
Err(e) => Err(e)
Err(e) => Err(e.into()),
}
}
UpdateMeta::ClearDocuments => {
@ -401,7 +401,7 @@ async fn main() -> anyhow::Result<()> {
match builder.execute() {
Ok(_count) => wtxn.commit().map_err(Into::into),
Err(e) => Err(e)
Err(e) => Err(e.into()),
}
}
UpdateMeta::Settings(settings) => {
@ -471,7 +471,7 @@ async fn main() -> anyhow::Result<()> {
match result {
Ok(_count) => wtxn.commit().map_err(Into::into),
Err(e) => Err(e)
Err(e) => Err(e.into()),
}
}
UpdateMeta::Facets(levels) => {
@ -486,7 +486,7 @@ async fn main() -> anyhow::Result<()> {
}
match builder.execute() {
Ok(()) => wtxn.commit().map_err(Into::into),
Err(e) => Err(e)
Err(e) => Err(e.into()),
}
}
};

100
infos/src/main.rs
View File

@ -5,55 +5,41 @@ use std::{str, io, fmt};
use anyhow::Context;
use byte_unit::Byte;
use heed::EnvOpenOptions;
use milli::facet::FacetType;
use milli::{Index, TreeLevel};
use structopt::StructOpt;
use milli::facet::FacetType;
use milli::index::db_name::*;
use milli::{Index, TreeLevel};
use Command::*;
#[cfg(target_os = "linux")]
#[global_allocator]
static ALLOC: jemallocator::Jemalloc = jemallocator::Jemalloc;
const MAIN_DB_NAME: &str = "main";
const WORD_DOCIDS_DB_NAME: &str = "word-docids";
const WORD_PREFIX_DOCIDS_DB_NAME: &str = "word-prefix-docids";
const DOCID_WORD_POSITIONS_DB_NAME: &str = "docid-word-positions";
const WORD_PAIR_PROXIMITY_DOCIDS_DB_NAME: &str = "word-pair-proximity-docids";
const WORD_PREFIX_PAIR_PROXIMITY_DOCIDS_DB_NAME: &str = "word-prefix-pair-proximity-docids";
const WORD_LEVEL_POSITION_DOCIDS_DB_NAME: &str = "word-level-position-docids";
const WORD_PREFIX_LEVEL_POSITION_DOCIDS_DB_NAME: &str = "word-prefix-level-position-docids";
const FIELD_ID_WORD_COUNT_DOCIDS_DB_NAME: &str = "field-id-word-count-docids";
const FACET_ID_F64_DOCIDS_DB_NAME: &str = "facet-id-f64-docids";
const FACET_ID_STRING_DOCIDS_DB_NAME: &str = "facet-id-string-docids";
const FIELD_ID_DOCID_FACET_F64S_DB_NAME: &str = "field-id-docid-facet-f64s";
const FIELD_ID_DOCID_FACET_STRINGS_DB_NAME: &str = "field-id-docid-facet-strings";
const DOCUMENTS_DB_NAME: &str = "documents";
const ALL_DATABASE_NAMES: &[&str] = &[
MAIN_DB_NAME,
WORD_DOCIDS_DB_NAME,
WORD_PREFIX_DOCIDS_DB_NAME,
DOCID_WORD_POSITIONS_DB_NAME,
WORD_PAIR_PROXIMITY_DOCIDS_DB_NAME,
WORD_PREFIX_PAIR_PROXIMITY_DOCIDS_DB_NAME,
WORD_LEVEL_POSITION_DOCIDS_DB_NAME,
WORD_PREFIX_LEVEL_POSITION_DOCIDS_DB_NAME,
FIELD_ID_WORD_COUNT_DOCIDS_DB_NAME,
FACET_ID_F64_DOCIDS_DB_NAME,
FACET_ID_STRING_DOCIDS_DB_NAME,
FIELD_ID_DOCID_FACET_F64S_DB_NAME,
FIELD_ID_DOCID_FACET_STRINGS_DB_NAME,
DOCUMENTS_DB_NAME,
MAIN,
WORD_DOCIDS,
WORD_PREFIX_DOCIDS,
DOCID_WORD_POSITIONS,
WORD_PAIR_PROXIMITY_DOCIDS,
WORD_PREFIX_PAIR_PROXIMITY_DOCIDS,
WORD_LEVEL_POSITION_DOCIDS,
WORD_PREFIX_LEVEL_POSITION_DOCIDS,
FIELD_ID_WORD_COUNT_DOCIDS,
FACET_ID_F64_DOCIDS,
FACET_ID_STRING_DOCIDS,
FIELD_ID_DOCID_FACET_F64S,
FIELD_ID_DOCID_FACET_STRINGS,
DOCUMENTS,
];
const POSTINGS_DATABASE_NAMES: &[&str] = &[
WORD_DOCIDS_DB_NAME,
WORD_PREFIX_DOCIDS_DB_NAME,
DOCID_WORD_POSITIONS_DB_NAME,
WORD_PAIR_PROXIMITY_DOCIDS_DB_NAME,
WORD_PREFIX_PAIR_PROXIMITY_DOCIDS_DB_NAME,
WORD_DOCIDS,
WORD_PREFIX_DOCIDS,
DOCID_WORD_POSITIONS,
WORD_PAIR_PROXIMITY_DOCIDS,
WORD_PREFIX_PAIR_PROXIMITY_DOCIDS,
];
#[derive(Debug, StructOpt)]
@ -944,21 +930,21 @@ fn size_of_databases(index: &Index, rtxn: &heed::RoTxn, names: Vec<String>) -> a
for name in names {
let database = match name.as_str() {
MAIN_DB_NAME => &main,
WORD_PREFIX_DOCIDS_DB_NAME => word_prefix_docids.as_polymorph(),
WORD_DOCIDS_DB_NAME => word_docids.as_polymorph(),
DOCID_WORD_POSITIONS_DB_NAME => docid_word_positions.as_polymorph(),
WORD_PAIR_PROXIMITY_DOCIDS_DB_NAME => word_pair_proximity_docids.as_polymorph(),
WORD_PREFIX_PAIR_PROXIMITY_DOCIDS_DB_NAME => word_prefix_pair_proximity_docids.as_polymorph(),
WORD_LEVEL_POSITION_DOCIDS_DB_NAME => word_level_position_docids.as_polymorph(),
WORD_PREFIX_LEVEL_POSITION_DOCIDS_DB_NAME => word_prefix_level_position_docids.as_polymorph(),
FIELD_ID_WORD_COUNT_DOCIDS_DB_NAME => field_id_word_count_docids.as_polymorph(),
FACET_ID_F64_DOCIDS_DB_NAME => facet_id_f64_docids.as_polymorph(),
FACET_ID_STRING_DOCIDS_DB_NAME => facet_id_string_docids.as_polymorph(),
FIELD_ID_DOCID_FACET_F64S_DB_NAME => field_id_docid_facet_f64s.as_polymorph(),
FIELD_ID_DOCID_FACET_STRINGS_DB_NAME => field_id_docid_facet_strings.as_polymorph(),
MAIN => &main,
WORD_PREFIX_DOCIDS => word_prefix_docids.as_polymorph(),
WORD_DOCIDS => word_docids.as_polymorph(),
DOCID_WORD_POSITIONS => docid_word_positions.as_polymorph(),
WORD_PAIR_PROXIMITY_DOCIDS => word_pair_proximity_docids.as_polymorph(),
WORD_PREFIX_PAIR_PROXIMITY_DOCIDS => word_prefix_pair_proximity_docids.as_polymorph(),
WORD_LEVEL_POSITION_DOCIDS => word_level_position_docids.as_polymorph(),
WORD_PREFIX_LEVEL_POSITION_DOCIDS => word_prefix_level_position_docids.as_polymorph(),
FIELD_ID_WORD_COUNT_DOCIDS => field_id_word_count_docids.as_polymorph(),
FACET_ID_F64_DOCIDS => facet_id_f64_docids.as_polymorph(),
FACET_ID_STRING_DOCIDS => facet_id_string_docids.as_polymorph(),
FIELD_ID_DOCID_FACET_F64S => field_id_docid_facet_f64s.as_polymorph(),
FIELD_ID_DOCID_FACET_STRINGS => field_id_docid_facet_strings.as_polymorph(),
DOCUMENTS_DB_NAME => documents.as_polymorph(),
DOCUMENTS => documents.as_polymorph(),
unknown => anyhow::bail!("unknown database {:?}", unknown),
};
@ -1039,27 +1025,27 @@ fn database_stats(index: &Index, rtxn: &heed::RoTxn, name: &str) -> anyhow::Resu
}
match name {
WORD_DOCIDS_DB_NAME => {
WORD_DOCIDS => {
let db = index.word_docids.as_polymorph();
compute_stats::<RoaringBitmapCodec>(*db, rtxn, name)
},
WORD_PREFIX_DOCIDS_DB_NAME => {
WORD_PREFIX_DOCIDS => {
let db = index.word_prefix_docids.as_polymorph();
compute_stats::<RoaringBitmapCodec>(*db, rtxn, name)
},
DOCID_WORD_POSITIONS_DB_NAME => {
DOCID_WORD_POSITIONS => {
let db = index.docid_word_positions.as_polymorph();
compute_stats::<BoRoaringBitmapCodec>(*db, rtxn, name)
},
WORD_PAIR_PROXIMITY_DOCIDS_DB_NAME => {
WORD_PAIR_PROXIMITY_DOCIDS => {
let db = index.word_pair_proximity_docids.as_polymorph();
compute_stats::<CboRoaringBitmapCodec>(*db, rtxn, name)
},
WORD_PREFIX_PAIR_PROXIMITY_DOCIDS_DB_NAME => {
WORD_PREFIX_PAIR_PROXIMITY_DOCIDS => {
let db = index.word_prefix_pair_proximity_docids.as_polymorph();
compute_stats::<CboRoaringBitmapCodec>(*db, rtxn, name)
},
FIELD_ID_WORD_COUNT_DOCIDS_DB_NAME => {
FIELD_ID_WORD_COUNT_DOCIDS => {
let db = index.field_id_word_count_docids.as_polymorph();
compute_stats::<CboRoaringBitmapCodec>(*db, rtxn, name)
},

1
milli/Cargo.toml
View File

@ -5,7 +5,6 @@ authors = ["Kerollmops <clement@meilisearch.com>"]
edition = "2018"
[dependencies]
anyhow = "1.0.38"
bstr = "0.2.15"
byteorder = "1.4.2"
chrono = { version = "0.4.19", features = ["serde"] }

11
milli/src/criterion.rs
View File

@ -1,11 +1,12 @@
use std::fmt;
use std::str::FromStr;
use anyhow::{Context, bail};
use regex::Regex;
use serde::{Serialize, Deserialize};
use once_cell::sync::Lazy;
use crate::error::{Error, UserError};
static ASC_DESC_REGEX: Lazy<Regex> = Lazy::new(|| {
Regex::new(r#"(asc|desc)\(([\w_-]+)\)"#).unwrap()
});
@ -41,7 +42,7 @@ impl Criterion {
}
impl FromStr for Criterion {
type Err = anyhow::Error;
type Err = Error;
fn from_str(txt: &str) -> Result<Criterion, Self::Err> {
match txt {
@ -51,13 +52,15 @@ impl FromStr for Criterion {
"attribute" => Ok(Criterion::Attribute),
"exactness" => Ok(Criterion::Exactness),
text => {
let caps = ASC_DESC_REGEX.captures(text).with_context(|| format!("unknown criterion name: {}", text))?;
let caps = ASC_DESC_REGEX.captures(text).ok_or_else(|| {
UserError::InvalidCriterionName { name: text.to_string() }
})?;
let order = caps.get(1).unwrap().as_str();
let field_name = caps.get(2).unwrap().as_str();
match order {
"asc" => Ok(Criterion::Asc(field_name.to_string())),
"desc" => Ok(Criterion::Desc(field_name.to_string())),
otherwise => bail!("unknown criterion name: {}", otherwise),
text => return Err(UserError::InvalidCriterionName { name: text.to_string() }.into()),
}
},
}

265
milli/src/error.rs Normal file
View File

@ -0,0 +1,265 @@
use std::convert::Infallible;
use std::error::Error as StdError;
use std::{fmt, io, str};
use heed::{MdbError, Error as HeedError};
use rayon::ThreadPoolBuildError;
use serde_json::{Map, Value};
use crate::search::ParserRule;
use crate::{DocumentId, FieldId};
pub type Object = Map<String, Value>;
#[derive(Debug)]
pub enum Error {
InternalError(InternalError),
IoError(io::Error),
UserError(UserError),
}
#[derive(Debug)]
pub enum InternalError {
DatabaseClosing,
DatabaseMissingEntry { db_name: &'static str, key: Option<&'static str> },
FieldIdMapMissingEntry(FieldIdMapMissingEntry),
Fst(fst::Error),
GrenadInvalidCompressionType,
IndexingMergingKeys { process: &'static str },
InvalidDatabaseTyping,
RayonThreadPool(ThreadPoolBuildError),
SerdeJson(serde_json::Error),
Serialization(SerializationError),
Store(MdbError),
Utf8(str::Utf8Error),
}
#[derive(Debug)]
pub enum SerializationError {
Decoding { db_name: Option<&'static str> },
Encoding { db_name: Option<&'static str> },
InvalidNumberSerialization,
}
#[derive(Debug)]
pub enum FieldIdMapMissingEntry {
FieldId { field_id: FieldId, process: &'static str },
FieldName { field_name: String, process: &'static str },
}
#[derive(Debug)]
pub enum UserError {
AttributeLimitReached,
Csv(csv::Error),
MaxDatabaseSizeReached,
DocumentLimitReached,
InvalidFilter(pest::error::Error<ParserRule>),
InvalidCriterionName { name: String },
InvalidDocumentId { document_id: Value },
InvalidFilterAttribute(pest::error::Error<ParserRule>),
InvalidStoreFile,
MissingDocumentId { document: Object },
MissingPrimaryKey,
NoSpaceLeftOnDevice,
PrimaryKeyCannotBeChanged,
PrimaryKeyCannotBeReset,
SerdeJson(serde_json::Error),
UnknownInternalDocumentId { document_id: DocumentId },
}
impl From<io::Error> for Error {
fn from(error: io::Error) -> Error {
// TODO must be improved and more precise
Error::IoError(error)
}
}
impl From<fst::Error> for Error {
fn from(error: fst::Error) -> Error {
Error::InternalError(InternalError::Fst(error))
}
}
impl<E> From<grenad::Error<E>> for Error where Error: From<E> {
fn from(error: grenad::Error<E>) -> Error {
match error {
grenad::Error::Io(error) => Error::IoError(error),
grenad::Error::Merge(error) => Error::from(error),
grenad::Error::InvalidCompressionType => {
Error::InternalError(InternalError::GrenadInvalidCompressionType)
},
}
}
}
impl From<str::Utf8Error> for Error {
fn from(error: str::Utf8Error) -> Error {
Error::InternalError(InternalError::Utf8(error))
}
}
impl From<Infallible> for Error {
fn from(_error: Infallible) -> Error {
unreachable!()
}
}
impl From<HeedError> for Error {
fn from(error: HeedError) -> Error {
use self::Error::*;
use self::InternalError::*;
use self::SerializationError::*;
use self::UserError::*;
match error {
HeedError::Io(error) => Error::from(error),
HeedError::Mdb(MdbError::MapFull) => UserError(MaxDatabaseSizeReached),
HeedError::Mdb(MdbError::Invalid) => UserError(InvalidStoreFile),
HeedError::Mdb(error) => InternalError(Store(error)),
HeedError::Encoding => InternalError(Serialization(Encoding { db_name: None })),
HeedError::Decoding => InternalError(Serialization(Decoding { db_name: None })),
HeedError::InvalidDatabaseTyping => InternalError(InvalidDatabaseTyping),
HeedError::DatabaseClosing => InternalError(DatabaseClosing),
}
}
}
impl From<ThreadPoolBuildError> for Error {
fn from(error: ThreadPoolBuildError) -> Error {
Error::InternalError(InternalError::RayonThreadPool(error))
}
}
impl From<FieldIdMapMissingEntry> for Error {
fn from(error: FieldIdMapMissingEntry) -> Error {
Error::InternalError(InternalError::FieldIdMapMissingEntry(error))
}
}
impl From<InternalError> for Error {
fn from(error: InternalError) -> Error {
Error::InternalError(error)
}
}
impl From<UserError> for Error {
fn from(error: UserError) -> Error {
Error::UserError(error)
}
}
impl From<SerializationError> for Error {
fn from(error: SerializationError) -> Error {
Error::InternalError(InternalError::Serialization(error))
}
}
impl fmt::Display for Error {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
match self {
Self::InternalError(error) => write!(f, "internal: {}", error),
Self::IoError(error) => error.fmt(f),
Self::UserError(error) => error.fmt(f),
}
}
}
impl StdError for Error {}
impl fmt::Display for InternalError {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
match self {
Self::DatabaseMissingEntry { db_name, key } => {
write!(f, "missing {} in the {} database", key.unwrap_or("key"), db_name)
},
Self::FieldIdMapMissingEntry(error) => error.fmt(f),
Self::Fst(error) => error.fmt(f),
Self::GrenadInvalidCompressionType => {
f.write_str("invalid compression type have been specified to grenad")
},
Self::IndexingMergingKeys { process } => {
write!(f, "invalid merge while processing {}", process)
},
Self::Serialization(error) => error.fmt(f),
Self::InvalidDatabaseTyping => HeedError::InvalidDatabaseTyping.fmt(f),
Self::RayonThreadPool(error) => error.fmt(f),
Self::SerdeJson(error) => error.fmt(f),
Self::DatabaseClosing => HeedError::DatabaseClosing.fmt(f),
Self::Store(error) => error.fmt(f),
Self::Utf8(error) => error.fmt(f),
}
}
}
impl StdError for InternalError {}
impl fmt::Display for UserError {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
match self {
Self::AttributeLimitReached => f.write_str("maximum number of attributes reached"),
Self::Csv(error) => error.fmt(f),
Self::DocumentLimitReached => f.write_str("maximum number of documents reached"),
Self::InvalidFilter(error) => error.fmt(f),
Self::InvalidCriterionName { name } => write!(f, "invalid criterion {}", name),
Self::InvalidDocumentId { document_id } => {
let json = serde_json::to_string(document_id).unwrap();
write!(f, "document identifier is invalid {}", json)
},
Self::InvalidFilterAttribute(error) => error.fmt(f),
Self::MissingDocumentId { document } => {
let json = serde_json::to_string(document).unwrap();
write!(f, "document doesn't have an identifier {}", json)
},
Self::MissingPrimaryKey => f.write_str("missing primary key"),
Self::MaxDatabaseSizeReached => f.write_str("maximum database size reached"),
// TODO where can we find it instead of writing the text ourselves?
Self::NoSpaceLeftOnDevice => f.write_str("no space left on device"),
Self::InvalidStoreFile => f.write_str("store file is not a valid database file"),
Self::PrimaryKeyCannotBeChanged => {
f.write_str("primary key cannot be changed if the database contains documents")
},
Self::PrimaryKeyCannotBeReset => {
f.write_str("primary key cannot be reset if the database contains documents")
},
Self::SerdeJson(error) => error.fmt(f),
Self::UnknownInternalDocumentId { document_id } => {
write!(f, "an unknown internal document id have been used ({})", document_id)
},
}
}
}
impl StdError for UserError {}
impl fmt::Display for FieldIdMapMissingEntry {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
match self {
Self::FieldId { field_id, process } => {
write!(f, "unknown field id {} coming from the {} process", field_id, process)
},
Self::FieldName { field_name, process } => {
write!(f, "unknown field name {} coming from the {} process", field_name, process)
},
}
}
}
impl StdError for FieldIdMapMissingEntry {}
impl fmt::Display for SerializationError {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
match self {
Self::Decoding { db_name: Some(name) } => {
write!(f, "decoding from the {} database failed", name)
},
Self::Decoding { db_name: None } => f.write_str("decoding failed"),
Self::Encoding { db_name: Some(name) } => {
write!(f, "encoding into the {} database failed", name)
},
Self::Encoding { db_name: None } => f.write_str("encoding failed"),
Self::InvalidNumberSerialization => f.write_str("number is not a valid finite number"),
}
}
}
impl StdError for SerializationError {}

13
milli/src/heed_codec/facet/facet_value_string_codec.rs
View File

@ -5,6 +5,14 @@ use crate::FieldId;
pub struct FacetValueStringCodec;
impl FacetValueStringCodec {
pub fn serialize_into(field_id: FieldId, value: &str, out: &mut Vec<u8>) {
out.reserve(value.len() + 1);
out.push(field_id);
out.extend_from_slice(value.as_bytes());
}
}
impl<'a> heed::BytesDecode<'a> for FacetValueStringCodec {
type DItem = (FieldId, &'a str);
@ -19,9 +27,8 @@ impl<'a> heed::BytesEncode<'a> for FacetValueStringCodec {
type EItem = (FieldId, &'a str);
fn bytes_encode((field_id, value): &Self::EItem) -> Option<Cow<[u8]>> {
let mut bytes = Vec::with_capacity(value.len() + 1);
bytes.push(*field_id);
bytes.extend_from_slice(value.as_bytes());
let mut bytes = Vec::new();
FacetValueStringCodec::serialize_into(*field_id, value, &mut bytes);
Some(Cow::Owned(bytes))
}
}

15
milli/src/heed_codec/facet/field_doc_id_facet_string_codec.rs
View File

@ -6,6 +6,15 @@ use crate::{FieldId, DocumentId};
pub struct FieldDocIdFacetStringCodec;
impl FieldDocIdFacetStringCodec {
pub fn serialize_into(field_id: FieldId, document_id: DocumentId, value: &str, out: &mut Vec<u8>) {
out.reserve(1 + 4 + value.len());
out.push(field_id);
out.extend_from_slice(&document_id.to_be_bytes());
out.extend_from_slice(value.as_bytes());
}
}
impl<'a> heed::BytesDecode<'a> for FieldDocIdFacetStringCodec {
type DItem = (FieldId, DocumentId, &'a str);
@ -22,10 +31,8 @@ impl<'a> heed::BytesEncode<'a> for FieldDocIdFacetStringCodec {
type EItem = (FieldId, DocumentId, &'a str);
fn bytes_encode((field_id, document_id, value): &Self::EItem) -> Option<Cow<[u8]>> {
let mut bytes = Vec::with_capacity(1 + 4 + value.len());
bytes.push(*field_id);
bytes.extend_from_slice(&document_id.to_be_bytes());
bytes.extend_from_slice(value.as_bytes());
let mut bytes = Vec::new();
FieldDocIdFacetStringCodec::serialize_into(*field_id, *document_id, value, &mut bytes);
Some(Cow::Owned(bytes))
}
}

15
milli/src/heed_codec/roaring_bitmap/bo_roaring_bitmap_codec.rs
View File

@ -6,6 +6,13 @@ use roaring::RoaringBitmap;
pub struct BoRoaringBitmapCodec;
impl BoRoaringBitmapCodec {
pub fn serialize_into(bitmap: &RoaringBitmap, out: &mut Vec<u8>) {
out.reserve(bitmap.len() as usize * size_of::<u32>());
bitmap.iter().map(u32::to_ne_bytes).for_each(|bytes| out.extend_from_slice(&bytes));
}
}
impl heed::BytesDecode<'_> for BoRoaringBitmapCodec {
type DItem = RoaringBitmap;
@ -25,12 +32,8 @@ impl heed::BytesEncode<'_> for BoRoaringBitmapCodec {
type EItem = RoaringBitmap;
fn bytes_encode(item: &Self::EItem) -> Option<Cow<[u8]>> {
let mut out = Vec::with_capacity(item.len() as usize * size_of::<u32>());
item.iter()
.map(|i| i.to_ne_bytes())
.for_each(|bytes| out.extend_from_slice(&bytes));
let mut out = Vec::new();
BoRoaringBitmapCodec::serialize_into(item, &mut out);
Some(Cow::Owned(out))
}
}

9
milli/src/heed_codec/roaring_bitmap/cbo_roaring_bitmap_codec.rs
View File

@ -23,18 +23,17 @@ impl CboRoaringBitmapCodec {
}
}
pub fn serialize_into(roaring: &RoaringBitmap, vec: &mut Vec<u8>) -> io::Result<()> {
pub fn serialize_into(roaring: &RoaringBitmap, vec: &mut Vec<u8>) {
if roaring.len() <= THRESHOLD as u64 {
// If the number of items (u32s) to encode is less than or equal to the threshold
// it means that it would weigh the same or less than the RoaringBitmap
// header, so we directly encode them using ByteOrder instead.
for integer in roaring {
vec.write_u32::<NativeEndian>(integer)?;
vec.write_u32::<NativeEndian>(integer).unwrap();
}
Ok(())
} else {
// Otherwise, we use the classic RoaringBitmapCodec that writes a header.
roaring.serialize_into(vec)
roaring.serialize_into(vec).unwrap();
}
}
@ -68,7 +67,7 @@ impl heed::BytesEncode<'_> for CboRoaringBitmapCodec {
fn bytes_encode(item: &Self::EItem) -> Option<Cow<[u8]>> {
let mut vec = Vec::with_capacity(Self::serialized_size(item));
Self::serialize_into(item, &mut vec).ok()?;
Self::serialize_into(item, &mut vec);
Some(Cow::Owned(vec))
}
}

409
milli/src/index.rs
View File

@ -2,14 +2,14 @@ use std::borrow::Cow;
use std::collections::{HashMap, HashSet};
use std::path::Path;
use anyhow::Context;
use chrono::{DateTime, Utc};
use heed::{Database, PolyDatabase, RoTxn, RwTxn};
use heed::types::*;
use roaring::RoaringBitmap;
use crate::error::{UserError, FieldIdMapMissingEntry, InternalError};
use crate::{Criterion, default_criteria, FacetDistribution, FieldsDistribution, Search};
use crate::{BEU32, DocumentId, ExternalDocumentsIds, FieldId};
use crate::{BEU32, DocumentId, ExternalDocumentsIds, FieldId, Result};
use crate::{
BEU32StrCodec, BoRoaringBitmapCodec, CboRoaringBitmapCodec,
ObkvCodec, RoaringBitmapCodec, RoaringBitmapLenCodec, StrLevelPositionCodec, StrStrU8Codec,
@ -21,25 +21,44 @@ use crate::heed_codec::facet::{
};
use crate::fields_ids_map::FieldsIdsMap;
pub const CRITERIA_KEY: &str = "criteria";
pub const DISPLAYED_FIELDS_KEY: &str = "displayed-fields";
pub const DISTINCT_FIELD_KEY: &str = "distinct-field-key";
pub const DOCUMENTS_IDS_KEY: &str = "documents-ids";
pub const FILTERABLE_FIELDS_KEY: &str = "filterable-fields";
pub const FIELDS_DISTRIBUTION_KEY: &str = "fields-distribution";
pub const FIELDS_IDS_MAP_KEY: &str = "fields-ids-map";
pub const HARD_EXTERNAL_DOCUMENTS_IDS_KEY: &str = "hard-external-documents-ids";
pub const NUMBER_FACETED_DOCUMENTS_IDS_PREFIX: &str = "number-faceted-documents-ids";
pub const PRIMARY_KEY_KEY: &str = "primary-key";
pub const SEARCHABLE_FIELDS_KEY: &str = "searchable-fields";
pub const SOFT_EXTERNAL_DOCUMENTS_IDS_KEY: &str = "soft-external-documents-ids";
pub const STOP_WORDS_KEY: &str = "stop-words";
pub const STRING_FACETED_DOCUMENTS_IDS_PREFIX: &str = "string-faceted-documents-ids";
pub const SYNONYMS_KEY: &str = "synonyms";
pub const WORDS_FST_KEY: &str = "words-fst";
pub const WORDS_PREFIXES_FST_KEY: &str = "words-prefixes-fst";
const CREATED_AT_KEY: &str = "created-at";
const UPDATED_AT_KEY: &str = "updated-at";
pub mod main_key {
pub const CRITERIA_KEY: &str = "criteria";
pub const DISPLAYED_FIELDS_KEY: &str = "displayed-fields";
pub const DISTINCT_FIELD_KEY: &str = "distinct-field-key";
pub const DOCUMENTS_IDS_KEY: &str = "documents-ids";
pub const FILTERABLE_FIELDS_KEY: &str = "filterable-fields";
pub const FIELDS_DISTRIBUTION_KEY: &str = "fields-distribution";
pub const FIELDS_IDS_MAP_KEY: &str = "fields-ids-map";
pub const HARD_EXTERNAL_DOCUMENTS_IDS_KEY: &str = "hard-external-documents-ids";
pub const NUMBER_FACETED_DOCUMENTS_IDS_PREFIX: &str = "number-faceted-documents-ids";
pub const PRIMARY_KEY_KEY: &str = "primary-key";
pub const SEARCHABLE_FIELDS_KEY: &str = "searchable-fields";
pub const SOFT_EXTERNAL_DOCUMENTS_IDS_KEY: &str = "soft-external-documents-ids";
pub const STOP_WORDS_KEY: &str = "stop-words";
pub const STRING_FACETED_DOCUMENTS_IDS_PREFIX: &str = "string-faceted-documents-ids";
pub const SYNONYMS_KEY: &str = "synonyms";
pub const WORDS_FST_KEY: &str = "words-fst";
pub const WORDS_PREFIXES_FST_KEY: &str = "words-prefixes-fst";
pub const CREATED_AT_KEY: &str = "created-at";
pub const UPDATED_AT_KEY: &str = "updated-at";
}
pub mod db_name {
pub const MAIN: &str = "main";
pub const WORD_DOCIDS: &str = "word-docids";
pub const WORD_PREFIX_DOCIDS: &str = "word-prefix-docids";
pub const DOCID_WORD_POSITIONS: &str = "docid-word-positions";
pub const WORD_PAIR_PROXIMITY_DOCIDS: &str = "word-pair-proximity-docids";
pub const WORD_PREFIX_PAIR_PROXIMITY_DOCIDS: &str = "word-prefix-pair-proximity-docids";
pub const WORD_LEVEL_POSITION_DOCIDS: &str = "word-level-position-docids";
pub const WORD_PREFIX_LEVEL_POSITION_DOCIDS: &str = "word-prefix-level-position-docids";
pub const FIELD_ID_WORD_COUNT_DOCIDS: &str = "field-id-word-count-docids";
pub const FACET_ID_F64_DOCIDS: &str = "facet-id-f64-docids";
pub const FACET_ID_STRING_DOCIDS: &str = "facet-id-string-docids";
pub const FIELD_ID_DOCID_FACET_F64S: &str = "field-id-docid-facet-f64s";
pub const FIELD_ID_DOCID_FACET_STRINGS: &str = "field-id-docid-facet-strings";
pub const DOCUMENTS: &str = "documents";
}
#[derive(Clone)]
pub struct Index {
@ -84,24 +103,26 @@ pub struct Index {
}
impl Index {
pub fn new<P: AsRef<Path>>(mut options: heed::EnvOpenOptions, path: P) -> anyhow::Result<Index> {
pub fn new<P: AsRef<Path>>(mut options: heed::EnvOpenOptions, path: P) -> Result<Index> {
use db_name::*;
options.max_dbs(14);
let env = options.open(path)?;
let main = env.create_poly_database(Some("main"))?;
let word_docids = env.create_database(Some("word-docids"))?;
let word_prefix_docids = env.create_database(Some("word-prefix-docids"))?;
let docid_word_positions = env.create_database(Some("docid-word-positions"))?;
let word_pair_proximity_docids = env.create_database(Some("word-pair-proximity-docids"))?;
let word_prefix_pair_proximity_docids = env.create_database(Some("word-prefix-pair-proximity-docids"))?;
let word_level_position_docids = env.create_database(Some("word-level-position-docids"))?;
let field_id_word_count_docids = env.create_database(Some("field-id-word-count-docids"))?;
let word_prefix_level_position_docids = env.create_database(Some("word-prefix-level-position-docids"))?;
let facet_id_f64_docids = env.create_database(Some("facet-id-f64-docids"))?;
let facet_id_string_docids = env.create_database(Some("facet-id-string-docids"))?;
let field_id_docid_facet_f64s = env.create_database(Some("field-id-docid-facet-f64s"))?;
let field_id_docid_facet_strings = env.create_database(Some("field-id-docid-facet-strings"))?;
let documents = env.create_database(Some("documents"))?;
let main = env.create_poly_database(Some(MAIN))?;
let word_docids = env.create_database(Some(WORD_DOCIDS))?;
let word_prefix_docids = env.create_database(Some(WORD_PREFIX_DOCIDS))?;
let docid_word_positions = env.create_database(Some(DOCID_WORD_POSITIONS))?;
let word_pair_proximity_docids = env.create_database(Some(WORD_PAIR_PROXIMITY_DOCIDS))?;
let word_prefix_pair_proximity_docids = env.create_database(Some(WORD_PREFIX_PAIR_PROXIMITY_DOCIDS))?;
let word_level_position_docids = env.create_database(Some(WORD_LEVEL_POSITION_DOCIDS))?;
let field_id_word_count_docids = env.create_database(Some(FIELD_ID_WORD_COUNT_DOCIDS))?;
let word_prefix_level_position_docids = env.create_database(Some(WORD_PREFIX_LEVEL_POSITION_DOCIDS))?;
let facet_id_f64_docids = env.create_database(Some(FACET_ID_F64_DOCIDS))?;
let facet_id_string_docids = env.create_database(Some(FACET_ID_STRING_DOCIDS))?;
let field_id_docid_facet_f64s = env.create_database(Some(FIELD_ID_DOCID_FACET_F64S))?;
let field_id_docid_facet_strings = env.create_database(Some(FIELD_ID_DOCID_FACET_STRINGS))?;
let documents = env.create_database(Some(DOCUMENTS))?;
Index::initialize_creation_dates(&env, main)?;
@ -127,10 +148,10 @@ impl Index {
fn initialize_creation_dates(env: &heed::Env, main: PolyDatabase) -> heed::Result<()> {
let mut txn = env.write_txn()?;
// The db was just created, we update its metadata with the relevant information.
if main.get::<_, Str, SerdeJson<DateTime<Utc>>>(&txn, CREATED_AT_KEY)?.is_none() {
if main.get::<_, Str, SerdeJson<DateTime<Utc>>>(&txn, main_key::CREATED_AT_KEY)?.is_none() {
let now = Utc::now();
main.put::<_, Str, SerdeJson<DateTime<Utc>>>(&mut txn, UPDATED_AT_KEY, &now)?;
main.put::<_, Str, SerdeJson<DateTime<Utc>>>(&mut txn, CREATED_AT_KEY, &now)?;
main.put::<_, Str, SerdeJson<DateTime<Utc>>>(&mut txn, main_key::UPDATED_AT_KEY, &now)?;
main.put::<_, Str, SerdeJson<DateTime<Utc>>>(&mut txn, main_key::CREATED_AT_KEY, &now)?;
txn.commit()?;
}
Ok(())
@ -163,43 +184,43 @@ impl Index {
/* documents ids */
/// Writes the documents ids that corresponds to the user-ids-documents-ids FST.
pub fn put_documents_ids(&self, wtxn: &mut RwTxn, docids: &RoaringBitmap) -> heed::Result<()> {
self.main.put::<_, Str, RoaringBitmapCodec>(wtxn, DOCUMENTS_IDS_KEY, docids)
pub(crate) fn put_documents_ids(&self, wtxn: &mut RwTxn, docids: &RoaringBitmap) -> heed::Result<()> {
self.main.put::<_, Str, RoaringBitmapCodec>(wtxn, main_key::DOCUMENTS_IDS_KEY, docids)
}
/// Returns the internal documents ids.
pub fn documents_ids(&self, rtxn: &RoTxn) -> heed::Result<RoaringBitmap> {
Ok(self.main.get::<_, Str, RoaringBitmapCodec>(rtxn, DOCUMENTS_IDS_KEY)?.unwrap_or_default())
Ok(self.main.get::<_, Str, RoaringBitmapCodec>(rtxn, main_key::DOCUMENTS_IDS_KEY)?.unwrap_or_default())
}
/// Returns the number of documents indexed in the database.
pub fn number_of_documents(&self, rtxn: &RoTxn) -> anyhow::Result<u64> {
let count = self.main.get::<_, Str, RoaringBitmapLenCodec>(rtxn, DOCUMENTS_IDS_KEY)?;
pub fn number_of_documents(&self, rtxn: &RoTxn) -> Result<u64> {
let count = self.main.get::<_, Str, RoaringBitmapLenCodec>(rtxn, main_key::DOCUMENTS_IDS_KEY)?;
Ok(count.unwrap_or_default())
}
/* primary key */
/// Writes the documents primary key, this is the field name that is used to store the id.
pub fn put_primary_key(&self, wtxn: &mut RwTxn, primary_key: &str) -> heed::Result<()> {
pub(crate) fn put_primary_key(&self, wtxn: &mut RwTxn, primary_key: &str) -> heed::Result<()> {
self.set_updated_at(wtxn, &Utc::now())?;
self.main.put::<_, Str, Str>(wtxn, PRIMARY_KEY_KEY, &primary_key)
self.main.put::<_, Str, Str>(wtxn, main_key::PRIMARY_KEY_KEY, &primary_key)
}
/// Deletes the primary key of the documents, this can be done to reset indexes settings.
pub fn delete_primary_key(&self, wtxn: &mut RwTxn) -> heed::Result<bool> {
self.main.delete::<_, Str>(wtxn, PRIMARY_KEY_KEY)
self.main.delete::<_, Str>(wtxn, main_key::PRIMARY_KEY_KEY)
}
/// Returns the documents primary key, `None` if it hasn't been defined.
pub fn primary_key<'t>(&self, rtxn: &'t RoTxn) -> heed::Result<Option<&'t str>> {
self.main.get::<_, Str, Str>(rtxn, PRIMARY_KEY_KEY)
self.main.get::<_, Str, Str>(rtxn, main_key::PRIMARY_KEY_KEY)
}
/* external documents ids */
/// Writes the external documents ids and internal ids (i.e. `u32`).
pub fn put_external_documents_ids<'a>(
pub(crate) fn put_external_documents_ids<'a>(
&self,
wtxn: &mut RwTxn,
external_documents_ids: &ExternalDocumentsIds<'a>,
@ -208,16 +229,16 @@ impl Index {
let ExternalDocumentsIds { hard, soft } = external_documents_ids;
let hard = hard.as_fst().as_bytes();
let soft = soft.as_fst().as_bytes();
self.main.put::<_, Str, ByteSlice>(wtxn, HARD_EXTERNAL_DOCUMENTS_IDS_KEY, hard)?;
self.main.put::<_, Str, ByteSlice>(wtxn, SOFT_EXTERNAL_DOCUMENTS_IDS_KEY, soft)?;
self.main.put::<_, Str, ByteSlice>(wtxn, main_key::HARD_EXTERNAL_DOCUMENTS_IDS_KEY, hard)?;
self.main.put::<_, Str, ByteSlice>(wtxn, main_key::SOFT_EXTERNAL_DOCUMENTS_IDS_KEY, soft)?;
Ok(())
}
/// Returns the external documents ids map which associate the external ids
/// with the internal ids (i.e. `u32`).
pub fn external_documents_ids<'t>(&self, rtxn: &'t RoTxn) -> anyhow::Result<ExternalDocumentsIds<'t>> {
let hard = self.main.get::<_, Str, ByteSlice>(rtxn, HARD_EXTERNAL_DOCUMENTS_IDS_KEY)?;
let soft = self.main.get::<_, Str, ByteSlice>(rtxn, SOFT_EXTERNAL_DOCUMENTS_IDS_KEY)?;
pub fn external_documents_ids<'t>(&self, rtxn: &'t RoTxn) -> Result<ExternalDocumentsIds<'t>> {
let hard = self.main.get::<_, Str, ByteSlice>(rtxn, main_key::HARD_EXTERNAL_DOCUMENTS_IDS_KEY)?;
let soft = self.main.get::<_, Str, ByteSlice>(rtxn, main_key::SOFT_EXTERNAL_DOCUMENTS_IDS_KEY)?;
let hard = match hard {
Some(hard) => fst::Map::new(hard)?.map_data(Cow::Borrowed)?,
None => fst::Map::default().map_data(Cow::Owned)?,
@ -233,93 +254,112 @@ impl Index {
/// Writes the fields ids map which associate the documents keys with an internal field id
/// (i.e. `u8`), this field id is used to identify fields in the obkv documents.
pub fn put_fields_ids_map(&self, wtxn: &mut RwTxn, map: &FieldsIdsMap) -> heed::Result<()> {
self.main.put::<_, Str, SerdeJson<FieldsIdsMap>>(wtxn, FIELDS_IDS_MAP_KEY, map)
pub(crate) fn put_fields_ids_map(&self, wtxn: &mut RwTxn, map: &FieldsIdsMap) -> heed::Result<()> {
self.main.put::<_, Str, SerdeJson<FieldsIdsMap>>(wtxn, main_key::FIELDS_IDS_MAP_KEY, map)
}
/// Returns the fields ids map which associate the documents keys with an internal field id
/// (i.e. `u8`), this field id is used to identify fields in the obkv documents.
pub fn fields_ids_map(&self, rtxn: &RoTxn) -> heed::Result<FieldsIdsMap> {
Ok(self.main.get::<_, Str, SerdeJson<FieldsIdsMap>>(rtxn, FIELDS_IDS_MAP_KEY)?.unwrap_or_default())
Ok(self.main.get::<_, Str, SerdeJson<FieldsIdsMap>>(
rtxn,
main_key::FIELDS_IDS_MAP_KEY,
)?.unwrap_or_default())
}
/* fields distribution */
/// Writes the fields distribution which associates every field name with
/// the number of times it occurs in the documents.
pub fn put_fields_distribution(&self, wtxn: &mut RwTxn, distribution: &FieldsDistribution) -> heed::Result<()> {
self.main.put::<_, Str, SerdeJson<FieldsDistribution>>(wtxn, FIELDS_DISTRIBUTION_KEY, distribution)
pub(crate) fn put_fields_distribution(&self, wtxn: &mut RwTxn, distribution: &FieldsDistribution) -> heed::Result<()> {
self.main.put::<_, Str, SerdeJson<FieldsDistribution>>(wtxn, main_key::FIELDS_DISTRIBUTION_KEY, distribution)
}
/// Returns the fields distribution which associates every field name with
/// the number of times it occurs in the documents.
pub fn fields_distribution(&self, rtxn: &RoTxn) -> heed::Result<FieldsDistribution> {
Ok(self.main.get::<_, Str, SerdeJson<FieldsDistribution>>(rtxn, FIELDS_DISTRIBUTION_KEY)?.unwrap_or_default())
Ok(self.main.get::<_, Str, SerdeJson<FieldsDistribution>>(
rtxn,
main_key::FIELDS_DISTRIBUTION_KEY,
)?.unwrap_or_default())
}
/* displayed fields */
/// Writes the fields that must be displayed in the defined order.
/// There must be not be any duplicate field id.
pub fn put_displayed_fields(&self, wtxn: &mut RwTxn, fields: &[&str]) -> heed::Result<()> {
self.main.put::<_, Str, SerdeBincode<&[&str]>>(wtxn, DISPLAYED_FIELDS_KEY, &fields)
pub(crate) fn put_displayed_fields(&self, wtxn: &mut RwTxn, fields: &[&str]) -> heed::Result<()> {
self.main.put::<_, Str, SerdeBincode<&[&str]>>(wtxn, main_key::DISPLAYED_FIELDS_KEY, &fields)
}
/// Deletes the displayed fields ids, this will make the engine to display
/// all the documents attributes in the order of the `FieldsIdsMap`.
pub fn delete_displayed_fields(&self, wtxn: &mut RwTxn) -> heed::Result<bool> {
self.main.delete::<_, Str>(wtxn, DISPLAYED_FIELDS_KEY)
self.main.delete::<_, Str>(wtxn, main_key::DISPLAYED_FIELDS_KEY)
}
/// Returns the displayed fields in the order they were set by the user. If it returns
/// `None` it means that all the attributes are set as displayed in the order of the `FieldsIdsMap`.
pub fn displayed_fields<'t>(&self, rtxn: &'t RoTxn) -> heed::Result<Option<Vec<&'t str>>> {
self.main.get::<_, Str, SerdeBincode<Vec<&'t str>>>(rtxn, DISPLAYED_FIELDS_KEY)
self.main.get::<_, Str, SerdeBincode<Vec<&'t str>>>(rtxn, main_key::DISPLAYED_FIELDS_KEY)
}
pub fn displayed_fields_ids(&self, rtxn: &RoTxn) -> heed::Result<Option<Vec<FieldId>>> {
let fields_ids_map = self.fields_ids_map(rtxn)?;
let ids = self.displayed_fields(rtxn)?
.map(|fields| fields
.into_iter()
.map(|name| fields_ids_map.id(name).expect("Field not found"))
.collect::<Vec<_>>());
Ok(ids)
/// Identical to `displayed_fields`, but returns the ids instead.
pub fn displayed_fields_ids(&self, rtxn: &RoTxn) -> Result<Option<Vec<FieldId>>> {
match self.displayed_fields(rtxn)? {
Some(fields) => {
let fields_ids_map = self.fields_ids_map(rtxn)?;
let mut fields_ids = Vec::new();
for name in fields.into_iter() {
match fields_ids_map.id(name) {
Some(field_id) => fields_ids.push(field_id),
None => return Err(FieldIdMapMissingEntry::FieldName {
field_name: name.to_string(),
process: "Index::displayed_fields_ids",
}.into()),
}
}
Ok(Some(fields_ids))
},
None => Ok(None),
}
}
/* searchable fields */
/// Writes the searchable fields, when this list is specified, only these are indexed.
pub fn put_searchable_fields(&self, wtxn: &mut RwTxn, fields: &[&str]) -> heed::Result<()> {
self.main.put::<_, Str, SerdeBincode<&[&str]>>(wtxn, SEARCHABLE_FIELDS_KEY, &fields)
pub(crate) fn put_searchable_fields(&self, wtxn: &mut RwTxn, fields: &[&str]) -> heed::Result<()> {
self.main.put::<_, Str, SerdeBincode<&[&str]>>(wtxn, main_key::SEARCHABLE_FIELDS_KEY, &fields)
}
/// Deletes the searchable fields, when no fields are specified, all fields are indexed.
pub fn delete_searchable_fields(&self, wtxn: &mut RwTxn) -> heed::Result<bool> {
self.main.delete::<_, Str>(wtxn, SEARCHABLE_FIELDS_KEY)
self.main.delete::<_, Str>(wtxn, main_key::SEARCHABLE_FIELDS_KEY)
}
/// Returns the searchable fields, those are the fields that are indexed,
/// if the searchable fields aren't there it means that **all** the fields are indexed.
pub fn searchable_fields<'t>(&self, rtxn: &'t RoTxn) -> heed::Result<Option<Vec<&'t str>>> {
self.main.get::<_, Str, SerdeBincode<Vec<&'t str>>>(rtxn, SEARCHABLE_FIELDS_KEY)
self.main.get::<_, Str, SerdeBincode<Vec<&'t str>>>(rtxn, main_key::SEARCHABLE_FIELDS_KEY)
}
/// Identical to `searchable_fields`, but returns the ids instead.
pub fn searchable_fields_ids(&self, rtxn: &RoTxn) -> heed::Result<Option<Vec<FieldId>>> {
pub fn searchable_fields_ids(&self, rtxn: &RoTxn) -> Result<Option<Vec<FieldId>>> {
match self.searchable_fields(rtxn)? {
Some(names) => {
let fields_map = self.fields_ids_map(rtxn)?;
let mut ids = Vec::new();
for name in names {
let id = fields_map
.id(name)
.ok_or_else(|| format!("field id map must contain {:?}", name))
.expect("corrupted data: ");
ids.push(id);
Some(fields) => {
let fields_ids_map = self.fields_ids_map(rtxn)?;
let mut fields_ids = Vec::new();
for name in fields {
match fields_ids_map.id(name) {
Some(field_id) => fields_ids.push(field_id),
None => return Err(FieldIdMapMissingEntry::FieldName {
field_name: name.to_string(),
process: "Index::searchable_fields_ids",
}.into()),
}
}
Ok(Some(ids))
}
Ok(Some(fields_ids))
},
None => Ok(None),
}
}
@ -327,35 +367,42 @@ impl Index {
/* filterable fields */
/// Writes the filterable fields names in the database.
pub fn put_filterable_fields(&self, wtxn: &mut RwTxn, fields: &HashSet<String>) -> heed::Result<()> {
self.main.put::<_, Str, SerdeJson<_>>(wtxn, FILTERABLE_FIELDS_KEY, fields)
pub(crate) fn put_filterable_fields(&self, wtxn: &mut RwTxn, fields: &HashSet<String>) -> heed::Result<()> {
self.main.put::<_, Str, SerdeJson<_>>(wtxn, main_key::FILTERABLE_FIELDS_KEY, fields)
}
/// Deletes the filterable fields ids in the database.
pub fn delete_filterable_fields(&self, wtxn: &mut RwTxn) -> heed::Result<bool> {
self.main.delete::<_, Str>(wtxn, FILTERABLE_FIELDS_KEY)
self.main.delete::<_, Str>(wtxn, main_key::FILTERABLE_FIELDS_KEY)
}
/// Returns the filterable fields names.
pub fn filterable_fields(&self, rtxn: &RoTxn) -> heed::Result<HashSet<String>> {
Ok(self.main.get::<_, Str, SerdeJson<_>>(rtxn, FILTERABLE_FIELDS_KEY)?.unwrap_or_default())
Ok(self.main.get::<_, Str, SerdeJson<_>>(
rtxn,
main_key::FILTERABLE_FIELDS_KEY,
)?.unwrap_or_default())
}
/// Same as `filterable_fields`, but returns ids instead.
pub fn filterable_fields_ids(&self, rtxn: &RoTxn) -> heed::Result<HashSet<FieldId>> {
let filterable_fields = self.filterable_fields(rtxn)?;
/// Identical to `filterable_fields`, but returns ids instead.
pub fn filterable_fields_ids(&self, rtxn: &RoTxn) -> Result<HashSet<FieldId>> {
let fields = self.filterable_fields(rtxn)?;
let fields_ids_map = self.fields_ids_map(rtxn)?;
let filterable_fields = filterable_fields
.iter()
.map(|k| {
fields_ids_map
.id(k)
.ok_or_else(|| format!("{:?} should be present in the field id map", k))
.expect("corrupted data: ")
})
.collect();
Ok(filterable_fields)
let mut fields_ids = HashSet::new();
for name in fields {
match fields_ids_map.id(&name) {
Some(field_id) => {
fields_ids.insert(field_id);
},
None => return Err(FieldIdMapMissingEntry::FieldName {
field_name: name,
process: "Index::filterable_fields_ids",
}.into()),
}
}
Ok(fields_ids)
}
/* faceted documents ids */
@ -363,7 +410,7 @@ impl Index {
/// Returns the faceted fields names.
///
/// Faceted fields are the union of all the filterable, distinct, and Asc/Desc fields.
pub fn faceted_fields(&self, rtxn: &RoTxn) -> heed::Result<HashSet<String>> {
pub fn faceted_fields(&self, rtxn: &RoTxn) -> Result<HashSet<String>> {
let filterable_fields = self.filterable_fields(rtxn)?;
let distinct_field = self.distinct_field(rtxn)?;
let asc_desc_fields = self.criteria(rtxn)?
@ -382,36 +429,40 @@ impl Index {
Ok(faceted_fields)
}
/// Same as `faceted_fields`, but returns ids instead.
pub fn faceted_fields_ids(&self, rtxn: &RoTxn) -> heed::Result<HashSet<FieldId>> {
let faceted_fields = self.faceted_fields(rtxn)?;
/// Identical to `faceted_fields`, but returns ids instead.
pub fn faceted_fields_ids(&self, rtxn: &RoTxn) -> Result<HashSet<FieldId>> {
let fields = self.faceted_fields(rtxn)?;
let fields_ids_map = self.fields_ids_map(rtxn)?;
let faceted_fields = faceted_fields
.iter()
.map(|k| {
fields_ids_map
.id(k)
.ok_or_else(|| format!("{:?} should be present in the field id map", k))
.expect("corrupted data: ")
})
.collect();
Ok(faceted_fields)
let mut fields_ids = HashSet::new();
for name in fields.into_iter() {
match fields_ids_map.id(&name) {
Some(field_id) => {
fields_ids.insert(field_id);
},
None => return Err(FieldIdMapMissingEntry::FieldName {
field_name: name,
process: "Index::faceted_fields_ids",
}.into()),
}
}
Ok(fields_ids)
}
/* faceted documents ids */
/// Writes the documents ids that are faceted with numbers under this field id.
pub fn put_number_faceted_documents_ids(
pub(crate) fn put_number_faceted_documents_ids(
&self,
wtxn: &mut RwTxn,
field_id: FieldId,
docids: &RoaringBitmap,
) -> heed::Result<()>
{
let mut buffer = [0u8; STRING_FACETED_DOCUMENTS_IDS_PREFIX.len() + 1];
buffer[..STRING_FACETED_DOCUMENTS_IDS_PREFIX.len()]
.copy_from_slice(STRING_FACETED_DOCUMENTS_IDS_PREFIX.as_bytes());
let mut buffer = [0u8; main_key::STRING_FACETED_DOCUMENTS_IDS_PREFIX.len() + 1];
buffer[..main_key::STRING_FACETED_DOCUMENTS_IDS_PREFIX.len()]
.copy_from_slice(main_key::STRING_FACETED_DOCUMENTS_IDS_PREFIX.as_bytes());
*buffer.last_mut().unwrap() = field_id;
self.main.put::<_, ByteSlice, RoaringBitmapCodec>(wtxn, &buffer, docids)
}
@ -423,9 +474,9 @@ impl Index {
field_id: FieldId,
) -> heed::Result<RoaringBitmap>
{
let mut buffer = [0u8; STRING_FACETED_DOCUMENTS_IDS_PREFIX.len() + 1];
buffer[..STRING_FACETED_DOCUMENTS_IDS_PREFIX.len()]
.copy_from_slice(STRING_FACETED_DOCUMENTS_IDS_PREFIX.as_bytes());
let mut buffer = [0u8; main_key::STRING_FACETED_DOCUMENTS_IDS_PREFIX.len() + 1];
buffer[..main_key::STRING_FACETED_DOCUMENTS_IDS_PREFIX.len()]
.copy_from_slice(main_key::STRING_FACETED_DOCUMENTS_IDS_PREFIX.as_bytes());
*buffer.last_mut().unwrap() = field_id;
match self.main.get::<_, ByteSlice, RoaringBitmapCodec>(rtxn, &buffer)? {
Some(docids) => Ok(docids),
@ -434,16 +485,16 @@ impl Index {
}
/// Writes the documents ids that are faceted with strings under this field id.
pub fn put_string_faceted_documents_ids(
pub(crate) fn put_string_faceted_documents_ids(
&self,
wtxn: &mut RwTxn,
field_id: FieldId,
docids: &RoaringBitmap,
) -> heed::Result<()>
{
let mut buffer = [0u8; NUMBER_FACETED_DOCUMENTS_IDS_PREFIX.len() + 1];
buffer[..NUMBER_FACETED_DOCUMENTS_IDS_PREFIX.len()]
.copy_from_slice(NUMBER_FACETED_DOCUMENTS_IDS_PREFIX.as_bytes());
let mut buffer = [0u8; main_key::NUMBER_FACETED_DOCUMENTS_IDS_PREFIX.len() + 1];
buffer[..main_key::NUMBER_FACETED_DOCUMENTS_IDS_PREFIX.len()]
.copy_from_slice(main_key::NUMBER_FACETED_DOCUMENTS_IDS_PREFIX.as_bytes());
*buffer.last_mut().unwrap() = field_id;
self.main.put::<_, ByteSlice, RoaringBitmapCodec>(wtxn, &buffer, docids)
}
@ -455,9 +506,9 @@ impl Index {
field_id: FieldId,
) -> heed::Result<RoaringBitmap>
{
let mut buffer = [0u8; NUMBER_FACETED_DOCUMENTS_IDS_PREFIX.len() + 1];
buffer[..NUMBER_FACETED_DOCUMENTS_IDS_PREFIX.len()]
.copy_from_slice(NUMBER_FACETED_DOCUMENTS_IDS_PREFIX.as_bytes());
let mut buffer = [0u8; main_key::NUMBER_FACETED_DOCUMENTS_IDS_PREFIX.len() + 1];
buffer[..main_key::NUMBER_FACETED_DOCUMENTS_IDS_PREFIX.len()]
.copy_from_slice(main_key::NUMBER_FACETED_DOCUMENTS_IDS_PREFIX.as_bytes());
*buffer.last_mut().unwrap() = field_id;
match self.main.get::<_, ByteSlice, RoaringBitmapCodec>(rtxn, &buffer)? {
Some(docids) => Ok(docids),
@ -468,29 +519,29 @@ impl Index {
/* distinct field */
pub(crate) fn put_distinct_field(&self, wtxn: &mut RwTxn, distinct_field: &str) -> heed::Result<()> {
self.main.put::<_, Str, Str>(wtxn, DISTINCT_FIELD_KEY, distinct_field)
self.main.put::<_, Str, Str>(wtxn, main_key::DISTINCT_FIELD_KEY, distinct_field)
}
pub fn distinct_field<'a>(&self, rtxn: &'a RoTxn) -> heed::Result<Option<&'a str>> {
self.main.get::<_, Str, Str>(rtxn, DISTINCT_FIELD_KEY)
self.main.get::<_, Str, Str>(rtxn, main_key::DISTINCT_FIELD_KEY)
}
pub(crate) fn delete_distinct_field(&self, wtxn: &mut RwTxn) -> heed::Result<bool> {
self.main.delete::<_, Str>(wtxn, DISTINCT_FIELD_KEY)
self.main.delete::<_, Str>(wtxn, main_key::DISTINCT_FIELD_KEY)
}
/* criteria */
pub fn put_criteria(&self, wtxn: &mut RwTxn, criteria: &[Criterion]) -> heed::Result<()> {
self.main.put::<_, Str, SerdeJson<&[Criterion]>>(wtxn, CRITERIA_KEY, &criteria)
pub(crate) fn put_criteria(&self, wtxn: &mut RwTxn, criteria: &[Criterion]) -> heed::Result<()> {
self.main.put::<_, Str, SerdeJson<&[Criterion]>>(wtxn, main_key::CRITERIA_KEY, &criteria)
}
pub fn delete_criteria(&self, wtxn: &mut RwTxn) -> heed::Result<bool> {
self.main.delete::<_, Str>(wtxn, CRITERIA_KEY)
self.main.delete::<_, Str>(wtxn, main_key::CRITERIA_KEY)
}
pub fn criteria(&self, rtxn: &RoTxn) -> heed::Result<Vec<Criterion>> {
match self.main.get::<_, Str, SerdeJson<Vec<Criterion>>>(rtxn, CRITERIA_KEY)? {
match self.main.get::<_, Str, SerdeJson<Vec<Criterion>>>(rtxn, main_key::CRITERIA_KEY)? {
Some(criteria) => Ok(criteria),
None => Ok(default_criteria()),
}
@ -499,13 +550,13 @@ impl Index {
/* words fst */
/// Writes the FST which is the words dictionary of the engine.
pub fn put_words_fst<A: AsRef<[u8]>>(&self, wtxn: &mut RwTxn, fst: &fst::Set<A>) -> heed::Result<()> {
self.main.put::<_, Str, ByteSlice>(wtxn, WORDS_FST_KEY, fst.as_fst().as_bytes())
pub(crate) fn put_words_fst<A: AsRef<[u8]>>(&self, wtxn: &mut RwTxn, fst: &fst::Set<A>) -> heed::Result<()> {
self.main.put::<_, Str, ByteSlice>(wtxn, main_key::WORDS_FST_KEY, fst.as_fst().as_bytes())
}
/// Returns the FST which is the words dictionary of the engine.
pub fn words_fst<'t>(&self, rtxn: &'t RoTxn) -> anyhow::Result<fst::Set<Cow<'t, [u8]>>> {
match self.main.get::<_, Str, ByteSlice>(rtxn, WORDS_FST_KEY)? {
pub fn words_fst<'t>(&self, rtxn: &'t RoTxn) -> Result<fst::Set<Cow<'t, [u8]>>> {
match self.main.get::<_, Str, ByteSlice>(rtxn, main_key::WORDS_FST_KEY)? {
Some(bytes) => Ok(fst::Set::new(bytes)?.map_data(Cow::Borrowed)?),
None => Ok(fst::Set::default().map_data(Cow::Owned)?),
}
@ -513,16 +564,16 @@ impl Index {
/* stop words */
pub fn put_stop_words<A: AsRef<[u8]>>(&self, wtxn: &mut RwTxn, fst: &fst::Set<A>) -> heed::Result<()> {
self.main.put::<_, Str, ByteSlice>(wtxn, STOP_WORDS_KEY, fst.as_fst().as_bytes())
pub(crate) fn put_stop_words<A: AsRef<[u8]>>(&self, wtxn: &mut RwTxn, fst: &fst::Set<A>) -> heed::Result<()> {
self.main.put::<_, Str, ByteSlice>(wtxn, main_key::STOP_WORDS_KEY, fst.as_fst().as_bytes())
}
pub fn delete_stop_words(&self, wtxn: &mut RwTxn) -> heed::Result<bool> {
self.main.delete::<_, Str>(wtxn, STOP_WORDS_KEY)
self.main.delete::<_, Str>(wtxn, main_key::STOP_WORDS_KEY)
}
pub fn stop_words<'t>(&self, rtxn: &'t RoTxn) -> anyhow::Result<Option<fst::Set<&'t [u8]>>> {
match self.main.get::<_, Str, ByteSlice>(rtxn, STOP_WORDS_KEY)? {
pub fn stop_words<'t>(&self, rtxn: &'t RoTxn) -> Result<Option<fst::Set<&'t [u8]>>> {
match self.main.get::<_, Str, ByteSlice>(rtxn, main_key::STOP_WORDS_KEY)? {
Some(bytes) => Ok(Some(fst::Set::new(bytes)?)),
None => Ok(None),
}
@ -530,19 +581,29 @@ impl Index {
/* synonyms */
pub fn put_synonyms(&self, wtxn: &mut RwTxn, synonyms: &HashMap<Vec<String>, Vec<Vec<String>>>) -> heed::Result<()> {
self.main.put::<_, Str, SerdeBincode<_>>(wtxn, SYNONYMS_KEY, synonyms)
pub(crate) fn put_synonyms(
&self,
wtxn: &mut RwTxn,
synonyms: &HashMap<Vec<String>, Vec<Vec<String>>>,
) -> heed::Result<()>
{
self.main.put::<_, Str, SerdeBincode<_>>(wtxn, main_key::SYNONYMS_KEY, synonyms)
}
pub fn delete_synonyms(&self, wtxn: &mut RwTxn) -> heed::Result<bool> {
self.main.delete::<_, Str>(wtxn, SYNONYMS_KEY)
self.main.delete::<_, Str>(wtxn, main_key::SYNONYMS_KEY)
}
pub fn synonyms(&self, rtxn: &RoTxn) -> heed::Result<HashMap<Vec<String>, Vec<Vec<String>>>> {
Ok(self.main.get::<_, Str, SerdeBincode<_>>(rtxn, SYNONYMS_KEY)?.unwrap_or_default())
Ok(self.main.get::<_, Str, SerdeBincode<_>>(rtxn, main_key::SYNONYMS_KEY)?.unwrap_or_default())
}
pub fn words_synonyms<S: AsRef<str>>(&self, rtxn: &RoTxn, words: &[S]) -> heed::Result<Option<Vec<Vec<String>>>> {
pub fn words_synonyms<S: AsRef<str>>(
&self,
rtxn: &RoTxn,
words: &[S],
) -> heed::Result<Option<Vec<Vec<String>>>>
{
let words: Vec<_> = words.iter().map(|s| s.as_ref().to_owned()).collect();
Ok(self.synonyms(rtxn)?.remove(&words))
}
@ -550,13 +611,13 @@ impl Index {
/* words prefixes fst */
/// Writes the FST which is the words prefixes dictionnary of the engine.
pub fn put_words_prefixes_fst<A: AsRef<[u8]>>(&self, wtxn: &mut RwTxn, fst: &fst::Set<A>) -> heed::Result<()> {
self.main.put::<_, Str, ByteSlice>(wtxn, WORDS_PREFIXES_FST_KEY, fst.as_fst().as_bytes())
pub(crate) fn put_words_prefixes_fst<A: AsRef<[u8]>>(&self, wtxn: &mut RwTxn, fst: &fst::Set<A>) -> heed::Result<()> {
self.main.put::<_, Str, ByteSlice>(wtxn, main_key::WORDS_PREFIXES_FST_KEY, fst.as_fst().as_bytes())
}
/// Returns the FST which is the words prefixes dictionnary of the engine.
pub fn words_prefixes_fst<'t>(&self, rtxn: &'t RoTxn) -> anyhow::Result<fst::Set<Cow<'t, [u8]>>> {
match self.main.get::<_, Str, ByteSlice>(rtxn, WORDS_PREFIXES_FST_KEY)? {
pub fn words_prefixes_fst<'t>(&self, rtxn: &'t RoTxn) -> Result<fst::Set<Cow<'t, [u8]>>> {
match self.main.get::<_, Str, ByteSlice>(rtxn, main_key::WORDS_PREFIXES_FST_KEY)? {
Some(bytes) => Ok(fst::Set::new(bytes)?.map_data(Cow::Borrowed)?),
None => Ok(fst::Set::default().map_data(Cow::Owned)?),
}
@ -577,13 +638,13 @@ impl Index {
&self,
rtxn: &'t RoTxn,
ids: impl IntoIterator<Item=DocumentId>,
) -> anyhow::Result<Vec<(DocumentId, obkv::KvReader<'t>)>>
) -> Result<Vec<(DocumentId, obkv::KvReader<'t>)>>
{
let mut documents = Vec::new();
for id in ids {
let kv = self.documents.get(rtxn, &BEU32::new(id))?
.with_context(|| format!("Could not find document {}", id))?;
.ok_or_else(|| UserError::UnknownInternalDocumentId { document_id: id })?;
documents.push((id, kv));
}
@ -594,7 +655,7 @@ impl Index {
pub fn all_documents<'t>(
&self,
rtxn: &'t RoTxn,
) -> anyhow::Result<impl Iterator<Item = heed::Result<(DocumentId, obkv::KvReader<'t>)>>> {
) -> Result<impl Iterator<Item = heed::Result<(DocumentId, obkv::KvReader<'t>)>>> {
Ok(self
.documents
.iter(rtxn)?
@ -611,23 +672,27 @@ impl Index {
}
/// Returns the index creation time.
pub fn created_at(&self, rtxn: &RoTxn) -> heed::Result<DateTime<Utc>> {
let time = self.main
.get::<_, Str, SerdeJson<DateTime<Utc>>>(rtxn, CREATED_AT_KEY)?
.expect("Index without creation time");
Ok(time)
pub fn created_at(&self, rtxn: &RoTxn) -> Result<DateTime<Utc>> {
Ok(self.main
.get::<_, Str, SerdeJson<DateTime<Utc>>>(rtxn, main_key::CREATED_AT_KEY)?
.ok_or(InternalError::DatabaseMissingEntry {
db_name: db_name::MAIN,
key: Some(main_key::CREATED_AT_KEY),
})?)
}
/// Returns the index last updated time.
pub fn updated_at(&self, rtxn: &RoTxn) -> heed::Result<DateTime<Utc>> {
let time = self.main
.get::<_, Str, SerdeJson<DateTime<Utc>>>(rtxn, UPDATED_AT_KEY)?
.expect("Index without update time");
Ok(time)
pub fn updated_at(&self, rtxn: &RoTxn) -> Result<DateTime<Utc>> {
Ok(self.main
.get::<_, Str, SerdeJson<DateTime<Utc>>>(rtxn, main_key::UPDATED_AT_KEY)?
.ok_or(InternalError::DatabaseMissingEntry {
db_name: db_name::MAIN,
key: Some(main_key::UPDATED_AT_KEY),
})?)
}
pub(crate) fn set_updated_at(&self, wtxn: &mut RwTxn, time: &DateTime<Utc>) -> heed::Result<()> {
self.main.put::<_, Str, SerdeJson<DateTime<Utc>>>(wtxn, UPDATED_AT_KEY, &time)
self.main.put::<_, Str, SerdeJson<DateTime<Utc>>>(wtxn, main_key::UPDATED_AT_KEY, &time)
}
}

17
milli/src/lib.rs
View File

@ -1,6 +1,7 @@
#[macro_use] extern crate pest_derive;
mod criterion;
mod error;
mod external_documents_ids;
mod fields_ids_map;
mod search;
@ -14,12 +15,13 @@ pub mod update;
use std::borrow::Cow;
use std::collections::HashMap;
use std::hash::BuildHasherDefault;
use std::result::Result as StdResult;
use anyhow::Context;
use fxhash::{FxHasher32, FxHasher64};
use serde_json::{Map, Value};
pub use self::criterion::{Criterion, default_criteria};
pub use self::error::Error;
pub use self::external_documents_ids::ExternalDocumentsIds;
pub use self::fields_ids_map::FieldsIdsMap;
pub use self::heed_codec::{BEU32StrCodec, StrStrU8Codec, StrLevelPositionCodec, ObkvCodec, FieldIdWordCountCodec};
@ -29,6 +31,8 @@ pub use self::index::Index;
pub use self::search::{Search, FacetDistribution, FilterCondition, SearchResult, MatchingWords};
pub use self::tree_level::TreeLevel;
pub type Result<T> = std::result::Result<T, error::Error>;
pub type FastMap4<K, V> = HashMap<K, V, BuildHasherDefault<FxHasher32>>;
pub type FastMap8<K, V> = HashMap<K, V, BuildHasherDefault<FxHasher64>>;
pub type SmallString32 = smallstr::SmallString<[u8; 32]>;
@ -43,21 +47,24 @@ pub type FieldId = u8;
pub type Position = u32;
pub type FieldsDistribution = HashMap<String, u64>;
type MergeFn = for<'a> fn(&[u8], &[Cow<'a, [u8]>]) -> anyhow::Result<Vec<u8>>;
type MergeFn<E> = for<'a> fn(&[u8], &[Cow<'a, [u8]>]) -> StdResult<Vec<u8>, E>;
/// Transform a raw obkv store into a JSON Object.
pub fn obkv_to_json(
displayed_fields: &[FieldId],
fields_ids_map: &FieldsIdsMap,
obkv: obkv::KvReader,
) -> anyhow::Result<Map<String, Value>>
) -> Result<Map<String, Value>>
{
displayed_fields.iter()
.copied()
.flat_map(|id| obkv.get(id).map(|value| (id, value)))
.map(|(id, value)| {
let name = fields_ids_map.name(id).context("unknown obkv field id")?;
let value = serde_json::from_slice(value)?;
let name = fields_ids_map.name(id).ok_or(error::FieldIdMapMissingEntry::FieldId {
field_id: id,
process: "obkv_to_json",
})?;
let value = serde_json::from_slice(value).map_err(error::InternalError::SerdeJson)?;
Ok((name.to_owned(), value))
})
.collect()

21
milli/src/search/criteria/asc_desc.rs
View File

@ -1,15 +1,15 @@
use std::mem::take;
use anyhow::Context;
use itertools::Itertools;
use log::debug;
use ordered_float::OrderedFloat;
use roaring::RoaringBitmap;
use crate::error::FieldIdMapMissingEntry;
use crate::search::criteria::{resolve_query_tree, CriteriaBuilder};
use crate::search::facet::FacetIter;
use crate::search::query_tree::Operation;
use crate::{FieldId, Index};
use crate::{FieldId, Index, Result};
use super::{Criterion, CriterionParameters, CriterionResult};
/// Threshold on the number of candidates that will make
@ -36,7 +36,7 @@ impl<'t> AscDesc<'t> {
rtxn: &'t heed::RoTxn,
parent: Box<dyn Criterion + 't>,
field_name: String,
) -> anyhow::Result<Self> {
) -> Result<Self> {
Self::new(index, rtxn, parent, field_name, true)
}
@ -45,7 +45,7 @@ impl<'t> AscDesc<'t> {
rtxn: &'t heed::RoTxn,
parent: Box<dyn Criterion + 't>,
field_name: String,
) -> anyhow::Result<Self> {
) -> Result<Self> {
Self::new(index, rtxn, parent, field_name, false)
}
@ -55,11 +55,14 @@ impl<'t> AscDesc<'t> {
parent: Box<dyn Criterion + 't>,
field_name: String,
ascending: bool,
) -> anyhow::Result<Self> {
) -> Result<Self> {
let fields_ids_map = index.fields_ids_map(rtxn)?;
let field_id = fields_ids_map
.id(&field_name)
.with_context(|| format!("field {:?} isn't registered", field_name))?;
.ok_or_else(|| FieldIdMapMissingEntry::FieldName {
field_name: field_name.clone(),
process: "AscDesc::new",
})?;
Ok(AscDesc {
index,
@ -79,7 +82,7 @@ impl<'t> AscDesc<'t> {
impl<'t> Criterion for AscDesc<'t> {
#[logging_timer::time("AscDesc::{}")]
fn next(&mut self, params: &mut CriterionParameters) -> anyhow::Result<Option<CriterionResult>> {
fn next(&mut self, params: &mut CriterionParameters) -> Result<Option<CriterionResult>> {
// remove excluded candidates when next is called, instead of doing it in the loop.
self.allowed_candidates -= params.excluded_candidates;
@ -162,7 +165,7 @@ fn facet_ordered<'t>(
field_id: FieldId,
ascending: bool,
candidates: RoaringBitmap,
) -> anyhow::Result<Box<dyn Iterator<Item = heed::Result<RoaringBitmap>> + 't>> {
) -> Result<Box<dyn Iterator<Item = heed::Result<RoaringBitmap>> + 't>> {
if candidates.len() <= CANDIDATES_THRESHOLD {
let iter = iterative_facet_ordered_iter(index, rtxn, field_id, ascending, candidates)?;
Ok(Box::new(iter.map(Ok)) as Box<dyn Iterator<Item = _>>)
@ -186,7 +189,7 @@ fn iterative_facet_ordered_iter<'t>(
field_id: FieldId,
ascending: bool,
candidates: RoaringBitmap,
) -> anyhow::Result<impl Iterator<Item = RoaringBitmap> + 't> {
) -> Result<impl Iterator<Item = RoaringBitmap> + 't> {
let mut docids_values = Vec::with_capacity(candidates.len() as usize);
for docid in candidates.iter() {
let left = (field_id, docid, f64::MIN);

17
milli/src/search/criteria/attribute.rs
View File

@ -5,7 +5,7 @@ use std::mem::take;
use roaring::RoaringBitmap;
use crate::{TreeLevel, search::build_dfa};
use crate::{TreeLevel, Result, search::build_dfa};
use crate::search::criteria::Query;
use crate::search::query_tree::{Operation, QueryKind};
use crate::search::{word_derivations, WordDerivationsCache};
@ -48,7 +48,7 @@ impl<'t> Attribute<'t> {
impl<'t> Criterion for Attribute<'t> {
#[logging_timer::time("Attribute::{}")]
fn next(&mut self, params: &mut CriterionParameters) -> anyhow::Result<Option<CriterionResult>> {
fn next(&mut self, params: &mut CriterionParameters) -> Result<Option<CriterionResult>> {
// remove excluded candidates when next is called, instead of doing it in the loop.
if let Some((_, _, allowed_candidates)) = self.state.as_mut() {
*allowed_candidates -= params.excluded_candidates;
@ -224,7 +224,12 @@ struct QueryLevelIterator<'t, 'q> {
}
impl<'t, 'q> QueryLevelIterator<'t, 'q> {
fn new(ctx: &'t dyn Context<'t>, queries: &'q [Query], wdcache: &mut WordDerivationsCache) -> anyhow::Result<Option<Self>> {
fn new(
ctx: &'t dyn Context<'t>,
queries: &'q [Query],
wdcache: &mut WordDerivationsCache,
) -> Result<Option<Self>>
{
let mut inner = Vec::with_capacity(queries.len());
for query in queries {
match &query.kind {
@ -471,7 +476,7 @@ fn initialize_query_level_iterators<'t, 'q>(
branches: &'q FlattenedQueryTree,
allowed_candidates: &RoaringBitmap,
wdcache: &mut WordDerivationsCache,
) -> anyhow::Result<BinaryHeap<Branch<'t, 'q>>> {
) -> Result<BinaryHeap<Branch<'t, 'q>>> {
let mut positions = BinaryHeap::with_capacity(branches.len());
for branch in branches {
@ -521,7 +526,7 @@ fn set_compute_candidates<'t>(
branches: &FlattenedQueryTree,
allowed_candidates: &RoaringBitmap,
wdcache: &mut WordDerivationsCache,
) -> anyhow::Result<Option<RoaringBitmap>>
) -> Result<Option<RoaringBitmap>>
{
let mut branches_heap = initialize_query_level_iterators(ctx, branches, allowed_candidates, wdcache)?;
let lowest_level = TreeLevel::min_value();
@ -573,7 +578,7 @@ fn linear_compute_candidates(
ctx: &dyn Context,
branches: &FlattenedQueryTree,
allowed_candidates: &RoaringBitmap,
) -> anyhow::Result<BTreeMap<u64, RoaringBitmap>>
) -> Result<BTreeMap<u64, RoaringBitmap>>
{
fn compute_candidate_rank(branches: &FlattenedQueryTree, words_positions: HashMap<String, RoaringBitmap>) -> u64 {
let mut min_rank = u64::max_value();

6
milli/src/search/criteria/exactness.rs
View File

@ -14,7 +14,7 @@ use crate::search::criteria::{
CriterionResult,
resolve_query_tree,
};
use crate::TreeLevel;
use crate::{TreeLevel, Result};
pub struct Exactness<'t> {
ctx: &'t dyn Context<'t>,
@ -45,7 +45,7 @@ impl<'t> Exactness<'t> {
impl<'t> Criterion for Exactness<'t> {
#[logging_timer::time("Exactness::{}")]
fn next(&mut self, params: &mut CriterionParameters) -> anyhow::Result<Option<CriterionResult>> {
fn next(&mut self, params: &mut CriterionParameters) -> Result<Option<CriterionResult>> {
// remove excluded candidates when next is called, instead of doing it in the loop.
if let Some(state) = self.state.as_mut() {
state.difference_with(params.excluded_candidates);
@ -158,7 +158,7 @@ fn resolve_state(
ctx: &dyn Context,
state: State,
query: &[ExactQueryPart],
) -> anyhow::Result<(RoaringBitmap, Option<State>)>
) -> Result<(RoaringBitmap, Option<State>)>
{
use State::*;
match state {

3
milli/src/search/criteria/final.rs
View File

@ -1,6 +1,7 @@
use log::debug;
use roaring::RoaringBitmap;
use crate::Result;
use crate::search::query_tree::Operation;
use crate::search::WordDerivationsCache;
use super::{resolve_query_tree, Criterion, CriterionResult, CriterionParameters, Context};
@ -29,7 +30,7 @@ impl<'t> Final<'t> {
}
#[logging_timer::time("Final::{}")]
pub fn next(&mut self, excluded_candidates: &RoaringBitmap) -> anyhow::Result<Option<FinalResult>> {
pub fn next(&mut self, excluded_candidates: &RoaringBitmap) -> Result<Option<FinalResult>> {
debug!("Final iteration");
let excluded_candidates = &self.returned_candidates | excluded_candidates;
let mut criterion_parameters = CriterionParameters {

4
milli/src/search/criteria/initial.rs
View File

@ -1,7 +1,7 @@
use roaring::RoaringBitmap;
use crate::Result;
use crate::search::query_tree::Operation;
use super::{Criterion, CriterionResult, CriterionParameters};
pub struct Initial {
@ -22,7 +22,7 @@ impl Initial {
impl Criterion for Initial {
#[logging_timer::time("Initial::{}")]
fn next(&mut self, _: &mut CriterionParameters) -> anyhow::Result<Option<CriterionResult>> {
fn next(&mut self, _: &mut CriterionParameters) -> Result<Option<CriterionResult>> {
Ok(self.answer.take())
}
}

33
milli/src/search/criteria/mod.rs
View File

@ -4,7 +4,7 @@ use std::borrow::Cow;
use roaring::RoaringBitmap;
use crate::{FieldId, TreeLevel, search::{word_derivations, WordDerivationsCache}};
use crate::{Index, DocumentId};
use crate::{Index, DocumentId, Result};
use super::query_tree::{Operation, PrimitiveQueryPart, Query, QueryKind};
use self::asc_desc::AscDesc;
@ -26,7 +26,7 @@ mod words;
pub mod r#final;
pub trait Criterion {
fn next(&mut self, params: &mut CriterionParameters) -> anyhow::Result<Option<CriterionResult>>;
fn next(&mut self, params: &mut CriterionParameters) -> Result<Option<CriterionResult>>;
}
/// The result of a call to the parent criterion.
@ -76,10 +76,11 @@ pub trait Context<'c> {
fn word_position_iterator(&self, word: &str, level: TreeLevel, in_prefix_cache: bool, left: Option<u32>, right: Option<u32>) -> heed::Result<Box<dyn Iterator<Item =heed::Result<((&'c str, TreeLevel, u32, u32), RoaringBitmap)>> + 'c>>;
fn word_position_last_level(&self, word: &str, in_prefix_cache: bool) -> heed::Result<Option<TreeLevel>>;
fn synonyms(&self, word: &str) -> heed::Result<Option<Vec<Vec<String>>>>;
fn searchable_fields_ids(&self) -> heed::Result<Vec<FieldId>>;
fn searchable_fields_ids(&self) -> Result<Vec<FieldId>>;
fn field_id_word_count_docids(&self, field_id: FieldId, word_count: u8) -> heed::Result<Option<RoaringBitmap>>;
fn word_level_position_docids(&self, word: &str, level: TreeLevel, left: u32, right: u32) -> Result<Option<RoaringBitmap>, heed::Error>;
fn word_level_position_docids(&self, word: &str, level: TreeLevel, left: u32, right: u32) -> heed::Result<Option<RoaringBitmap>>;
}
pub struct CriteriaBuilder<'t> {
rtxn: &'t heed::RoTxn<'t>,
index: &'t Index,
@ -173,7 +174,7 @@ impl<'c> Context<'c> for CriteriaBuilder<'c> {
self.index.words_synonyms(self.rtxn, &[word])
}
fn searchable_fields_ids(&self) -> heed::Result<Vec<FieldId>> {
fn searchable_fields_ids(&self) -> Result<Vec<FieldId>> {
match self.index.searchable_fields_ids(self.rtxn)? {
Some(searchable_fields_ids) => Ok(searchable_fields_ids),
None => Ok(self.index.fields_ids_map(self.rtxn)?.ids().collect()),
@ -185,14 +186,14 @@ impl<'c> Context<'c> for CriteriaBuilder<'c> {
self.index.field_id_word_count_docids.get(self.rtxn, &key)
}
fn word_level_position_docids(&self, word: &str, level: TreeLevel, left: u32, right: u32) -> Result<Option<RoaringBitmap>, heed::Error> {
fn word_level_position_docids(&self, word: &str, level: TreeLevel, left: u32, right: u32) -> heed::Result<Option<RoaringBitmap>> {
let key = (word, level, left, right);
self.index.word_level_position_docids.get(self.rtxn, &key)
}
}
impl<'t> CriteriaBuilder<'t> {
pub fn new(rtxn: &'t heed::RoTxn<'t>, index: &'t Index) -> anyhow::Result<Self> {
pub fn new(rtxn: &'t heed::RoTxn<'t>, index: &'t Index) -> Result<Self> {
let words_fst = index.words_fst(rtxn)?;
let words_prefixes_fst = index.words_prefixes_fst(rtxn)?;
Ok(Self { rtxn, index, words_fst, words_prefixes_fst })
@ -203,7 +204,7 @@ impl<'t> CriteriaBuilder<'t> {
query_tree: Option<Operation>,
primitive_query: Option<Vec<PrimitiveQueryPart>>,
filtered_candidates: Option<RoaringBitmap>,
) -> anyhow::Result<Final<'t>>
) -> Result<Final<'t>>
{
use crate::criterion::Criterion as Name;
@ -230,13 +231,13 @@ pub fn resolve_query_tree<'t>(
ctx: &'t dyn Context,
query_tree: &Operation,
wdcache: &mut WordDerivationsCache,
) -> anyhow::Result<RoaringBitmap>
) -> Result<RoaringBitmap>
{
fn resolve_operation<'t>(
ctx: &'t dyn Context,
query_tree: &Operation,
wdcache: &mut WordDerivationsCache,
) -> anyhow::Result<RoaringBitmap>
) -> Result<RoaringBitmap>
{
use Operation::{And, Phrase, Or, Query};
@ -244,7 +245,7 @@ pub fn resolve_query_tree<'t>(
And(ops) => {
let mut ops = ops.iter().map(|op| {
resolve_operation(ctx, op, wdcache)
}).collect::<anyhow::Result<Vec<_>>>()?;
}).collect::<Result<Vec<_>>>()?;
ops.sort_unstable_by_key(|cds| cds.len());
@ -302,7 +303,7 @@ fn all_word_pair_proximity_docids<T: AsRef<str>, U: AsRef<str>>(
left_words: &[(T, u8)],
right_words: &[(U, u8)],
proximity: u8
) -> anyhow::Result<RoaringBitmap>
) -> Result<RoaringBitmap>
{
let mut docids = RoaringBitmap::new();
for (left, _l_typo) in left_words {
@ -318,7 +319,7 @@ fn query_docids(
ctx: &dyn Context,
query: &Query,
wdcache: &mut WordDerivationsCache,
) -> anyhow::Result<RoaringBitmap>
) -> Result<RoaringBitmap>
{
match &query.kind {
QueryKind::Exact { word, .. } => {
@ -354,7 +355,7 @@ fn query_pair_proximity_docids(
right: &Query,
proximity: u8,
wdcache: &mut WordDerivationsCache,
) -> anyhow::Result<RoaringBitmap>
) -> Result<RoaringBitmap>
{
if proximity >= 8 {
let mut candidates = query_docids(ctx, left, wdcache)?;
@ -477,11 +478,11 @@ pub mod test {
todo!()
}
fn searchable_fields_ids(&self) -> heed::Result<Vec<FieldId>> {
fn searchable_fields_ids(&self) -> Result<Vec<FieldId>> {
todo!()
}
fn word_level_position_docids(&self, _word: &str, _level: TreeLevel, _left: u32, _right: u32) -> Result<Option<RoaringBitmap>, heed::Error> {
fn word_level_position_docids(&self, _word: &str, _level: TreeLevel, _left: u32, _right: u32) -> heed::Result<Option<RoaringBitmap>> {
todo!()
}

19
milli/src/search/criteria/proximity.rs
View File

@ -5,9 +5,10 @@ use std::mem::take;
use roaring::RoaringBitmap;
use log::debug;
use crate::{DocumentId, Position, search::{query_tree::QueryKind}};
use crate::search::query_tree::{maximum_proximity, Operation, Query};
use crate::search::{build_dfa, WordDerivationsCache};
use crate::search::{query_tree::QueryKind};
use crate::{DocumentId, Position, Result};
use super::{
Context,
Criterion,
@ -55,7 +56,7 @@ impl<'t> Proximity<'t> {
impl<'t> Criterion for Proximity<'t> {
#[logging_timer::time("Proximity::{}")]
fn next(&mut self, params: &mut CriterionParameters) -> anyhow::Result<Option<CriterionResult>> {
fn next(&mut self, params: &mut CriterionParameters) -> Result<Option<CriterionResult>> {
// remove excluded candidates when next is called, instead of doing it in the loop.
if let Some((_, _, allowed_candidates)) = self.state.as_mut() {
*allowed_candidates -= params.excluded_candidates;
@ -161,7 +162,7 @@ fn resolve_candidates<'t>(
proximity: u8,
cache: &mut Cache,
wdcache: &mut WordDerivationsCache,
) -> anyhow::Result<RoaringBitmap>
) -> Result<RoaringBitmap>
{
fn resolve_operation<'t>(
ctx: &'t dyn Context,
@ -169,7 +170,7 @@ fn resolve_candidates<'t>(
proximity: u8,
cache: &mut Cache,
wdcache: &mut WordDerivationsCache,
) -> anyhow::Result<Vec<(Query, Query, RoaringBitmap)>>
) -> Result<Vec<(Query, Query, RoaringBitmap)>>
{
use Operation::{And, Phrase, Or};
@ -227,7 +228,7 @@ fn resolve_candidates<'t>(
proximity: u8,
cache: &mut Cache,
wdcache: &mut WordDerivationsCache,
) -> anyhow::Result<Vec<(Query, Query, RoaringBitmap)>>
) -> Result<Vec<(Query, Query, RoaringBitmap)>>
{
fn pair_combinations(mana: u8, left_max: u8) -> impl Iterator<Item = (u8, u8)> {
(0..=mana.min(left_max)).map(move |m| (m, mana - m))
@ -281,7 +282,7 @@ fn resolve_candidates<'t>(
proximity: u8,
cache: &mut Cache,
wdcache: &mut WordDerivationsCache,
) -> anyhow::Result<Vec<(Query, Query, RoaringBitmap)>>
) -> Result<Vec<(Query, Query, RoaringBitmap)>>
{
// Extract the first two elements but gives the tail
// that is just after the first element.
@ -324,13 +325,13 @@ fn resolve_plane_sweep_candidates(
query_tree: &Operation,
allowed_candidates: &RoaringBitmap,
wdcache: &mut WordDerivationsCache,
) -> anyhow::Result<BTreeMap<u8, RoaringBitmap>>
) -> Result<BTreeMap<u8, RoaringBitmap>>
{
/// FIXME may be buggy with query like "new new york"
fn plane_sweep(
groups_positions: Vec<Vec<(Position, u8, Position)>>,
consecutive: bool,
) -> anyhow::Result<Vec<(Position, u8, Position)>>
) -> Result<Vec<(Position, u8, Position)>>
{
fn compute_groups_proximity(
groups: &[(usize, (Position, u8, Position))],
@ -451,7 +452,7 @@ fn resolve_plane_sweep_candidates(
rocache: &mut HashMap<&'a Operation, Vec<(Position, u8, Position)>>,
words_positions: &HashMap<String, RoaringBitmap>,
wdcache: &mut WordDerivationsCache,
) -> anyhow::Result<Vec<(Position, u8, Position)>>
) -> Result<Vec<(Position, u8, Position)>>
{
use Operation::{And, Phrase, Or};

13
milli/src/search/criteria/typo.rs
View File

@ -5,6 +5,7 @@ use roaring::RoaringBitmap;
use crate::search::query_tree::{maximum_typo, Operation, Query, QueryKind};
use crate::search::{word_derivations, WordDerivationsCache};
use crate::Result;
use super::{
Candidates,
Context,
@ -43,7 +44,7 @@ impl<'t> Typo<'t> {
impl<'t> Criterion for Typo<'t> {
#[logging_timer::time("Typo::{}")]
fn next(&mut self, params: &mut CriterionParameters) -> anyhow::Result<Option<CriterionResult>> {
fn next(&mut self, params: &mut CriterionParameters) -> Result<Option<CriterionResult>> {
use Candidates::{Allowed, Forbidden};
// remove excluded candidates when next is called, instead of doing it in the loop.
match self.state.as_mut() {
@ -163,14 +164,14 @@ fn alterate_query_tree(
mut query_tree: Operation,
number_typos: u8,
wdcache: &mut WordDerivationsCache,
) -> anyhow::Result<Operation>
) -> Result<Operation>
{
fn recurse(
words_fst: &fst::Set<Cow<[u8]>>,
operation: &mut Operation,
number_typos: u8,
wdcache: &mut WordDerivationsCache,
) -> anyhow::Result<()>
) -> Result<()>
{
use Operation::{And, Phrase, Or};
@ -218,7 +219,7 @@ fn resolve_candidates<'t>(
number_typos: u8,
cache: &mut HashMap<(Operation, u8), RoaringBitmap>,
wdcache: &mut WordDerivationsCache,
) -> anyhow::Result<RoaringBitmap>
) -> Result<RoaringBitmap>
{
fn resolve_operation<'t>(
ctx: &'t dyn Context,
@ -226,7 +227,7 @@ fn resolve_candidates<'t>(
number_typos: u8,
cache: &mut HashMap<(Operation, u8), RoaringBitmap>,
wdcache: &mut WordDerivationsCache,
) -> anyhow::Result<RoaringBitmap>
) -> Result<RoaringBitmap>
{
use Operation::{And, Phrase, Or, Query};
@ -277,7 +278,7 @@ fn resolve_candidates<'t>(
mana: u8,
cache: &mut HashMap<(Operation, u8), RoaringBitmap>,
wdcache: &mut WordDerivationsCache,
) -> anyhow::Result<RoaringBitmap>
) -> Result<RoaringBitmap>
{
match branches.split_first() {
Some((head, [])) => {

3
milli/src/search/criteria/words.rs
View File

@ -4,6 +4,7 @@ use log::debug;
use roaring::RoaringBitmap;
use crate::search::query_tree::Operation;
use crate::Result;
use super::{Context, Criterion, CriterionParameters, CriterionResult, resolve_query_tree};
pub struct Words<'t> {
@ -30,7 +31,7 @@ impl<'t> Words<'t> {
impl<'t> Criterion for Words<'t> {
#[logging_timer::time("Words::{}")]
fn next(&mut self, params: &mut CriterionParameters) -> anyhow::Result<Option<CriterionResult>> {
fn next(&mut self, params: &mut CriterionParameters) -> Result<Option<CriterionResult>> {
// remove excluded candidates when next is called, instead of doing it in the loop.
if let Some(candidates) = self.candidates.as_mut() {
*candidates -= params.excluded_candidates;

28
milli/src/search/distinct/facet_distinct.rs
View File

@ -3,9 +3,11 @@ use std::mem::size_of;
use heed::types::ByteSlice;
use roaring::RoaringBitmap;
use super::{Distinct, DocIter};
use crate::error::InternalError;
use crate::heed_codec::facet::*;
use crate::{DocumentId, FieldId, Index};
use crate::index::db_name;
use crate::{DocumentId, FieldId, Index, Result};
use super::{Distinct, DocIter};
const FID_SIZE: usize = size_of::<FieldId>();
const DOCID_SIZE: usize = size_of::<DocumentId>();
@ -57,14 +59,17 @@ impl<'a> FacetDistinctIter<'a> {
.get(self.txn, &(self.distinct, 0, key, key))
}
fn distinct_string(&mut self, id: DocumentId) -> anyhow::Result<()> {
fn distinct_string(&mut self, id: DocumentId) -> Result<()> {
let iter = facet_string_values(id, self.distinct, self.index, self.txn)?;
for item in iter {
let ((_, _, value), _) = item?;
let facet_docids = self
.facet_string_docids(value)?
.expect("Corrupted data: Facet values must exist");
.ok_or(InternalError::DatabaseMissingEntry {
db_name: db_name::FACET_ID_STRING_DOCIDS,
key: None,
})?;
self.excluded.union_with(&facet_docids);
}
@ -73,14 +78,17 @@ impl<'a> FacetDistinctIter<'a> {
Ok(())
}
fn distinct_number(&mut self, id: DocumentId) -> anyhow::Result<()> {
fn distinct_number(&mut self, id: DocumentId) -> Result<()> {
let iter = facet_number_values(id, self.distinct, self.index, self.txn)?;
for item in iter {
let ((_, _, value), _) = item?;
let facet_docids = self
.facet_number_docids(value)?
.expect("Corrupted data: Facet values must exist");
.ok_or(InternalError::DatabaseMissingEntry {
db_name: db_name::FACET_ID_F64_DOCIDS,
key: None,
})?;
self.excluded.union_with(&facet_docids);
}
@ -92,7 +100,7 @@ impl<'a> FacetDistinctIter<'a> {
/// Performs the next iteration of the facet distinct. This is a convenience method that is
/// called by the Iterator::next implementation that transposes the result. It makes error
/// handling easier.
fn next_inner(&mut self) -> anyhow::Result<Option<DocumentId>> {
fn next_inner(&mut self) -> Result<Option<DocumentId>> {
// The first step is to remove all the excluded documents from our candidates
self.candidates.difference_with(&self.excluded);
@ -129,7 +137,7 @@ fn facet_number_values<'a>(
distinct: FieldId,
index: &Index,
txn: &'a heed::RoTxn,
) -> anyhow::Result<heed::RoPrefix<'a, FieldDocIdFacetF64Codec, heed::types::Unit>> {
) -> Result<heed::RoPrefix<'a, FieldDocIdFacetF64Codec, heed::types::Unit>> {
let key = facet_values_prefix_key(distinct, id);
let iter = index
@ -146,7 +154,7 @@ fn facet_string_values<'a>(
distinct: FieldId,
index: &Index,
txn: &'a heed::RoTxn,
) -> anyhow::Result<heed::RoPrefix<'a, FieldDocIdFacetStringCodec, heed::types::Unit>> {
) -> Result<heed::RoPrefix<'a, FieldDocIdFacetStringCodec, heed::types::Unit>> {
let key = facet_values_prefix_key(distinct, id);
let iter = index
@ -159,7 +167,7 @@ fn facet_string_values<'a>(
}
impl Iterator for FacetDistinctIter<'_> {
type Item = anyhow::Result<DocumentId>;
type Item = Result<DocumentId>;
fn next(&mut self) -> Option<Self::Item> {
self.next_inner().transpose()

6
milli/src/search/distinct/mod.rs
View File

@ -3,13 +3,13 @@ mod noop_distinct;
use roaring::RoaringBitmap;
use crate::DocumentId;
use crate::{DocumentId, Result};
pub use facet_distinct::FacetDistinct;
pub use noop_distinct::NoopDistinct;
/// A trait implemented by document interators that are returned by calls to `Distinct::distinct`.
/// It provides a way to get back the ownership to the excluded set.
pub trait DocIter: Iterator<Item = anyhow::Result<DocumentId>> {
pub trait DocIter: Iterator<Item = Result<DocumentId>> {
/// Returns ownership on the internal exluded set.
fn into_excluded(self) -> RoaringBitmap;
}
@ -106,7 +106,7 @@ mod test {
/// Checks that all the candidates are distinct, and returns the candidates number.
pub(crate) fn validate_distinct_candidates(
candidates: impl Iterator<Item=anyhow::Result<DocumentId>>,
candidates: impl Iterator<Item = crate::Result<DocumentId>>,
distinct: FieldId,
index: &Index,
) -> usize {

4
milli/src/search/distinct/noop_distinct.rs
View File

@ -1,6 +1,6 @@
use roaring::{RoaringBitmap, bitmap::IntoIter};
use crate::DocumentId;
use crate::{DocumentId, Result};
use super::{DocIter, Distinct};
/// A distinct implementer that does not perform any distinct,
@ -13,7 +13,7 @@ pub struct NoopDistinctIter {
}
impl Iterator for NoopDistinctIter {
type Item = anyhow::Result<DocumentId>;
type Item = Result<DocumentId>;
fn next(&mut self) -> Option<Self::Item> {
self.candidates.next().map(Ok)

11
milli/src/search/facet/facet_distribution.rs
View File

@ -2,15 +2,15 @@ use std::collections::{HashSet, BTreeMap};
use std::ops::Bound::Unbounded;
use std::{cmp, fmt};
use anyhow::Context;
use heed::{Database, BytesDecode};
use heed::types::{ByteSlice, Unit};
use roaring::RoaringBitmap;
use crate::error::FieldIdMapMissingEntry;
use crate::facet::FacetType;
use crate::heed_codec::facet::FacetValueStringCodec;
use crate::search::facet::{FacetIter, FacetRange};
use crate::{Index, FieldId, DocumentId};
use crate::{Index, FieldId, DocumentId, Result};
/// The default number of values by facets that will
/// be fetched from the key-value store.
@ -195,14 +195,15 @@ impl<'a> FacetDistribution<'a> {
}
}
pub fn execute(&self) -> anyhow::Result<BTreeMap<String, BTreeMap<String, u64>>> {
pub fn execute(&self) -> Result<BTreeMap<String, BTreeMap<String, u64>>> {
let fields_ids_map = self.index.fields_ids_map(self.rtxn)?;
let filterable_fields = self.index.filterable_fields(self.rtxn)?;
let mut distribution = BTreeMap::new();
for name in filterable_fields {
let fid = fields_ids_map.id(&name).with_context(|| {
format!("missing field name {:?} from the fields id map", name)
let fid = fields_ids_map.id(&name).ok_or_else(|| FieldIdMapMissingEntry::FieldName {
field_name: name.clone(),
process: "FacetDistribution::execute",
})?;
let values = self.facet_values(fid)?;
distribution.insert(name, values);

70
milli/src/search/facet/filter_condition.rs
View File

@ -1,6 +1,7 @@
use std::collections::HashSet;
use std::fmt::Debug;
use std::ops::Bound::{self, Included, Excluded};
use std::result::Result as StdResult;
use std::str::FromStr;
use either::Either;
@ -11,8 +12,9 @@ use pest::iterators::{Pair, Pairs};
use pest::Parser;
use roaring::RoaringBitmap;
use crate::error::UserError;
use crate::heed_codec::facet::{FacetValueStringCodec, FacetLevelValueF64Codec};
use crate::{Index, FieldId, FieldsIdsMap, CboRoaringBitmapCodec};
use crate::{Index, FieldId, FieldsIdsMap, CboRoaringBitmapCodec, Result};
use super::FacetRange;
use super::parser::Rule;
@ -60,7 +62,7 @@ impl FilterCondition {
rtxn: &heed::RoTxn,
index: &Index,
array: I,
) -> anyhow::Result<Option<FilterCondition>>
) -> Result<Option<FilterCondition>>
where I: IntoIterator<Item=Either<J, B>>,
J: IntoIterator<Item=A>,
A: AsRef<str>,
@ -104,11 +106,11 @@ impl FilterCondition {
rtxn: &heed::RoTxn,
index: &Index,
expression: &str,
) -> anyhow::Result<FilterCondition>
) -> Result<FilterCondition>
{
let fields_ids_map = index.fields_ids_map(rtxn)?;
let filterable_fields = index.filterable_fields_ids(rtxn)?;
let lexed = FilterParser::parse(Rule::prgm, expression)?;
let lexed = FilterParser::parse(Rule::prgm, expression).map_err(UserError::InvalidFilter)?;
FilterCondition::from_pairs(&fields_ids_map, &filterable_fields, lexed)
}
@ -116,7 +118,7 @@ impl FilterCondition {
fim: &FieldsIdsMap,
ff: &HashSet<FieldId>,
expression: Pairs<Rule>,
) -> anyhow::Result<Self>
) -> Result<Self>
{
PREC_CLIMBER.climb(
expression,
@ -133,7 +135,7 @@ impl FilterCondition {
Rule::term => Self::from_pairs(fim, ff, pair.into_inner()),
_ => unreachable!(),
},
|lhs: anyhow::Result<Self>, op: Pair<Rule>, rhs: anyhow::Result<Self>| {
|lhs: Result<Self>, op: Pair<Rule>, rhs: Result<Self>| {
match op.as_rule() {
Rule::or => Ok(Or(Box::new(lhs?), Box::new(rhs?))),
Rule::and => Ok(And(Box::new(lhs?), Box::new(rhs?))),
@ -158,16 +160,17 @@ impl FilterCondition {
fields_ids_map: &FieldsIdsMap,
filterable_fields: &HashSet<FieldId>,
item: Pair<Rule>,
) -> anyhow::Result<FilterCondition>
) -> Result<FilterCondition>
{
let mut items = item.into_inner();
let fid = field_id(fields_ids_map, filterable_fields, &mut items)?;
let fid = field_id(fields_ids_map, filterable_fields, &mut items)
.map_err(UserError::InvalidFilterAttribute)?;
let (lresult, _) = pest_parse(items.next().unwrap());
let (rresult, _) = pest_parse(items.next().unwrap());
let lvalue = lresult?;
let rvalue = rresult?;
let lvalue = lresult.map_err(UserError::InvalidFilter)?;
let rvalue = rresult.map_err(UserError::InvalidFilter)?;
Ok(Operator(fid, Between(lvalue, rvalue)))
}
@ -176,10 +179,11 @@ impl FilterCondition {
fields_ids_map: &FieldsIdsMap,
filterable_fields: &HashSet<FieldId>,
item: Pair<Rule>,
) -> anyhow::Result<FilterCondition>
) -> Result<FilterCondition>
{
let mut items = item.into_inner();
let fid = field_id(fields_ids_map, filterable_fields, &mut items)?;
let fid = field_id(fields_ids_map, filterable_fields, &mut items)
.map_err(UserError::InvalidFilterAttribute)?;
let value = items.next().unwrap();
let (result, svalue) = pest_parse(value);
@ -192,60 +196,68 @@ impl FilterCondition {
fields_ids_map: &FieldsIdsMap,
filterable_fields: &HashSet<FieldId>,
item: Pair<Rule>,
) -> anyhow::Result<FilterCondition>
) -> Result<FilterCondition>
{
let mut items = item.into_inner();
let fid = field_id(fields_ids_map, filterable_fields, &mut items)?;
let fid = field_id(fields_ids_map, filterable_fields, &mut items)
.map_err(UserError::InvalidFilterAttribute)?;
let value = items.next().unwrap();
let (result, _svalue) = pest_parse(value);
let value = result.map_err(UserError::InvalidFilter)?;
Ok(Operator(fid, GreaterThan(result?)))
Ok(Operator(fid, GreaterThan(value)))
}
fn greater_than_or_equal(
fields_ids_map: &FieldsIdsMap,
filterable_fields: &HashSet<FieldId>,
item: Pair<Rule>,
) -> anyhow::Result<FilterCondition>
) -> Result<FilterCondition>
{
let mut items = item.into_inner();
let fid = field_id(fields_ids_map, filterable_fields, &mut items)?;
let fid = field_id(fields_ids_map, filterable_fields, &mut items)
.map_err(UserError::InvalidFilterAttribute)?;
let value = items.next().unwrap();
let (result, _svalue) = pest_parse(value);
let value = result.map_err(UserError::InvalidFilter)?;
Ok(Operator(fid, GreaterThanOrEqual(result?)))
Ok(Operator(fid, GreaterThanOrEqual(value)))
}
fn lower_than(
fields_ids_map: &FieldsIdsMap,
filterable_fields: &HashSet<FieldId>,
item: Pair<Rule>,
) -> anyhow::Result<FilterCondition>
) -> Result<FilterCondition>
{
let mut items = item.into_inner();
let fid = field_id(fields_ids_map, filterable_fields, &mut items)?;
let fid = field_id(fields_ids_map, filterable_fields, &mut items)
.map_err(UserError::InvalidFilterAttribute)?;
let value = items.next().unwrap();
let (result, _svalue) = pest_parse(value);
let value = result.map_err(UserError::InvalidFilter)?;
Ok(Operator(fid, LowerThan(result?)))
Ok(Operator(fid, LowerThan(value)))
}
fn lower_than_or_equal(
fields_ids_map: &FieldsIdsMap,
filterable_fields: &HashSet<FieldId>,
item: Pair<Rule>,
) -> anyhow::Result<FilterCondition>
) -> Result<FilterCondition>
{
let mut items = item.into_inner();
let fid = field_id(fields_ids_map, filterable_fields, &mut items)?;
let fid = field_id(fields_ids_map, filterable_fields, &mut items)
.map_err(UserError::InvalidFilterAttribute)?;
let value = items.next().unwrap();
let (result, _svalue) = pest_parse(value);
let value = result.map_err(UserError::InvalidFilter)?;
Ok(Operator(fid, LowerThanOrEqual(result?)))
Ok(Operator(fid, LowerThanOrEqual(value)))
}
}
@ -260,7 +272,7 @@ impl FilterCondition {
left: Bound<f64>,
right: Bound<f64>,
output: &mut RoaringBitmap,
) -> anyhow::Result<()>
) -> Result<()>
{
match (left, right) {
// If the request is an exact value we must go directly to the deepest level.
@ -332,7 +344,7 @@ impl FilterCondition {
strings_db: heed::Database<FacetValueStringCodec, CboRoaringBitmapCodec>,
field_id: FieldId,
operator: &Operator,
) -> anyhow::Result<RoaringBitmap>
) -> Result<RoaringBitmap>
{
// Make sure we always bound the ranges with the field id and the level,
// as the facets values are all in the same database and prefixed by the
@ -390,7 +402,7 @@ impl FilterCondition {
&self,
rtxn: &heed::RoTxn,
index: &Index,
) -> anyhow::Result<RoaringBitmap>
) -> Result<RoaringBitmap>
{
let numbers_db = index.facet_id_f64_docids;
let strings_db = index.facet_id_string_docids;
@ -422,7 +434,7 @@ fn field_id(
fields_ids_map: &FieldsIdsMap,
filterable_fields: &HashSet<FieldId>,
items: &mut Pairs<Rule>,
) -> Result<FieldId, PestError<Rule>>
) -> StdResult<FieldId, PestError<Rule>>
{
// lexing ensures that we at least have a key
let key = items.next().unwrap();
@ -463,7 +475,7 @@ fn field_id(
/// the original string that we tried to parse.
///
/// Returns the parsing error associated with the span if the conversion fails.
fn pest_parse<T>(pair: Pair<Rule>) -> (Result<T, pest::error::Error<Rule>>, String)
fn pest_parse<T>(pair: Pair<Rule>) -> (StdResult<T, pest::error::Error<Rule>>, String)
where T: FromStr,
T::Err: ToString,
{

3
milli/src/search/facet/mod.rs
View File

@ -9,8 +9,9 @@ use crate::heed_codec::CboRoaringBitmapCodec;
use crate::heed_codec::facet::FacetLevelValueF64Codec;
use crate::{Index, FieldId};
pub use self::filter_condition::{FilterCondition, Operator};
pub use self::facet_distribution::FacetDistribution;
pub use self::filter_condition::{FilterCondition, Operator};
pub(crate) use self::parser::Rule as ParserRule;
mod filter_condition;
mod facet_distribution;

19
milli/src/search/mod.rs
View File

@ -2,6 +2,7 @@ use std::borrow::Cow;
use std::collections::hash_map::{Entry, HashMap};
use std::fmt;
use std::mem::take;
use std::result::Result as StdResult;
use std::str::Utf8Error;
use std::time::Instant;
@ -12,14 +13,17 @@ use meilisearch_tokenizer::{Analyzer, AnalyzerConfig};
use once_cell::sync::Lazy;
use roaring::bitmap::RoaringBitmap;
use distinct::{Distinct, DocIter, FacetDistinct, NoopDistinct};
use crate::error::FieldIdMapMissingEntry;
use crate::search::criteria::r#final::{Final, FinalResult};
use crate::{Index, DocumentId};
use crate::{Index, DocumentId, Result};
pub use self::facet::{FilterCondition, FacetDistribution, FacetIter, Operator};
pub use self::matching_words::MatchingWords;
pub(crate) use self::facet::ParserRule;
use self::query_tree::QueryTreeBuilder;
use distinct::{Distinct, DocIter, FacetDistinct, NoopDistinct};
// Building these factories is not free.
static LEVDIST0: Lazy<LevBuilder> = Lazy::new(|| LevBuilder::new(0, true));
static LEVDIST1: Lazy<LevBuilder> = Lazy::new(|| LevBuilder::new(1, true));
@ -93,7 +97,7 @@ impl<'a> Search<'a> {
self
}
pub fn execute(&self) -> anyhow::Result<SearchResult> {
pub fn execute(&self) -> Result<SearchResult> {
// We create the query tree by spliting the query into tokens.
let before = Instant::now();
let (query_tree, primitive_query) = match self.query.as_ref() {
@ -140,7 +144,10 @@ impl<'a> Search<'a> {
None => self.perform_sort(NoopDistinct, matching_words, criteria),
Some(name) => {
let field_ids_map = self.index.fields_ids_map(self.rtxn)?;
let id = field_ids_map.id(name).expect("distinct not present in field map");
let id = field_ids_map.id(name).ok_or_else(|| FieldIdMapMissingEntry::FieldName {
field_name: name.to_string(),
process: "distinct attribute",
})?;
let distinct = FacetDistinct::new(id, self.index, self.rtxn);
self.perform_sort(distinct, matching_words, criteria)
}
@ -152,7 +159,7 @@ impl<'a> Search<'a> {
mut distinct: D,
matching_words: MatchingWords,
mut criteria: Final,
) -> anyhow::Result<SearchResult>
) -> Result<SearchResult>
{
let mut offset = self.offset;
let mut initial_candidates = RoaringBitmap::new();
@ -225,7 +232,7 @@ pub fn word_derivations<'c>(
max_typo: u8,
fst: &fst::Set<Cow<[u8]>>,
cache: &'c mut WordDerivationsCache,
) -> Result<&'c [(String, u8)], Utf8Error> {
) -> StdResult<&'c [(String, u8)], Utf8Error> {
match cache.entry((word.to_string(), is_prefix, max_typo)) {
Entry::Occupied(entry) => Ok(entry.into_mut()),
Entry::Vacant(entry) => {

14
milli/src/search/query_tree.rs
View File

@ -7,7 +7,7 @@ use meilisearch_tokenizer::TokenKind;
use roaring::RoaringBitmap;
use slice_group_by::GroupBy;
use crate::Index;
use crate::{Index, Result};
type IsOptionalWord = bool;
type IsPrefix = bool;
@ -219,7 +219,7 @@ impl<'a> QueryTreeBuilder<'a> {
/// - if `authorize_typos` is set to `false` the query tree will be generated
/// forcing all query words to match documents without any typo
/// (the criterion `typo` will be ignored)
pub fn build(&self, query: TokenStream) -> anyhow::Result<Option<(Operation, PrimitiveQuery)>> {
pub fn build(&self, query: TokenStream) -> Result<Option<(Operation, PrimitiveQuery)>> {
let stop_words = self.index.stop_words(self.rtxn)?;
let primitive_query = create_primitive_query(query, stop_words, self.words_limit);
if !primitive_query.is_empty() {
@ -291,14 +291,14 @@ fn create_query_tree(
optional_words: bool,
authorize_typos: bool,
query: &[PrimitiveQueryPart],
) -> anyhow::Result<Operation>
) -> Result<Operation>
{
/// Matches on the `PrimitiveQueryPart` and create an operation from it.
fn resolve_primitive_part(
ctx: &impl Context,
authorize_typos: bool,
part: PrimitiveQueryPart,
) -> anyhow::Result<Operation>
) -> Result<Operation>
{
match part {
// 1. try to split word in 2
@ -325,7 +325,7 @@ fn create_query_tree(
ctx: &impl Context,
authorize_typos: bool,
query: &[PrimitiveQueryPart],
) -> anyhow::Result<Operation>
) -> Result<Operation>
{
const MAX_NGRAM: usize = 3;
let mut op_children = Vec::new();
@ -379,7 +379,7 @@ fn create_query_tree(
ctx: &impl Context,
authorize_typos: bool,
query: PrimitiveQuery,
) -> anyhow::Result<Operation>
) -> Result<Operation>
{
let number_phrases = query.iter().filter(|p| p.is_phrase()).count();
let mut operation_children = Vec::new();
@ -532,7 +532,7 @@ mod test {
authorize_typos: bool,
words_limit: Option<usize>,
query: TokenStream,
) -> anyhow::Result<Option<(Operation, PrimitiveQuery)>>
) -> Result<Option<(Operation, PrimitiveQuery)>>
{
let primitive_query = create_primitive_query(query, None, words_limit);
if !primitive_query.is_empty() {

5
milli/src/update/clear_documents.rs
View File

@ -1,6 +1,7 @@
use chrono::Utc;
use roaring::RoaringBitmap;
use crate::{ExternalDocumentsIds, Index, FieldsDistribution};
use crate::{ExternalDocumentsIds, Index, FieldsDistribution, Result};
pub struct ClearDocuments<'t, 'u, 'i> {
wtxn: &'t mut heed::RwTxn<'i, 'u>,
@ -18,7 +19,7 @@ impl<'t, 'u, 'i> ClearDocuments<'t, 'u, 'i> {
ClearDocuments { wtxn, index, _update_id: update_id }
}
pub fn execute(self) -> anyhow::Result<u64> {
pub fn execute(self) -> Result<u64> {
self.index.set_updated_at(self.wtxn, &Utc::now())?;
let Index {
env: _env,

25
milli/src/update/delete_documents.rs
View File

@ -1,15 +1,16 @@
use std::collections::HashMap;
use std::collections::hash_map::Entry;
use anyhow::{anyhow, Context};
use chrono::Utc;
use fst::IntoStreamer;
use heed::types::{ByteSlice, Unit};
use roaring::RoaringBitmap;
use serde_json::Value;
use crate::error::{InternalError, FieldIdMapMissingEntry, UserError};
use crate::heed_codec::CboRoaringBitmapCodec;
use crate::{Index, DocumentId, FieldId, BEU32, SmallString32, ExternalDocumentsIds};
use crate::index::{db_name, main_key};
use crate::{Index, DocumentId, FieldId, BEU32, SmallString32, ExternalDocumentsIds, Result};
use super::ClearDocuments;
pub struct DeleteDocuments<'t, 'u, 'i> {
@ -25,7 +26,7 @@ impl<'t, 'u, 'i> DeleteDocuments<'t, 'u, 'i> {
wtxn: &'t mut heed::RwTxn<'i, 'u>,
index: &'i Index,
update_id: u64,
) -> anyhow::Result<DeleteDocuments<'t, 'u, 'i>>
) -> Result<DeleteDocuments<'t, 'u, 'i>>
{
let external_documents_ids = index
.external_documents_ids(wtxn)?
@ -54,7 +55,7 @@ impl<'t, 'u, 'i> DeleteDocuments<'t, 'u, 'i> {
Some(docid)
}
pub fn execute(self) -> anyhow::Result<u64> {
pub fn execute(self) -> Result<u64> {
self.index.set_updated_at(self.wtxn, &Utc::now())?;
// We retrieve the current documents ids that are in the database.
let mut documents_ids = self.index.documents_ids(self.wtxn)?;
@ -77,8 +78,18 @@ impl<'t, 'u, 'i> DeleteDocuments<'t, 'u, 'i> {
}
let fields_ids_map = self.index.fields_ids_map(self.wtxn)?;
let primary_key = self.index.primary_key(self.wtxn)?.context("missing primary key")?;
let id_field = fields_ids_map.id(primary_key).expect(r#"the field "id" to be present"#);
let primary_key = self.index.primary_key(self.wtxn)?.ok_or_else(|| {
InternalError::DatabaseMissingEntry {
db_name: db_name::MAIN,
key: Some(main_key::PRIMARY_KEY_KEY),
}
})?;
let id_field = fields_ids_map.id(primary_key).ok_or_else(|| {
FieldIdMapMissingEntry::FieldName {
field_name: primary_key.to_string(),
process: "DeleteDocuments::execute",
}
})?;
let Index {
env: _env,
@ -119,7 +130,7 @@ impl<'t, 'u, 'i> DeleteDocuments<'t, 'u, 'i> {
let external_id = match serde_json::from_slice(content).unwrap() {
Value::String(string) => SmallString32::from(string.as_str()),
Value::Number(number) => SmallString32::from(number.to_string()),
_ => return Err(anyhow!("documents ids must be either strings or numbers")),
document_id => return Err(UserError::InvalidDocumentId { document_id }.into()),
};
external_ids.push(external_id);
}

13
milli/src/update/facets.rs
View File

@ -9,11 +9,12 @@ use heed::{BytesEncode, Error};
use log::debug;
use roaring::RoaringBitmap;
use crate::error::InternalError;
use crate::heed_codec::CboRoaringBitmapCodec;
use crate::heed_codec::facet::FacetLevelValueF64Codec;
use crate::Index;
use crate::update::index_documents::WriteMethod;
use crate::update::index_documents::{create_writer, writer_into_reader, write_into_lmdb_database};
use crate::{Index, Result};
pub struct Facets<'t, 'u, 'i> {
wtxn: &'t mut heed::RwTxn<'i, 'u>,
@ -55,7 +56,7 @@ impl<'t, 'u, 'i> Facets<'t, 'u, 'i> {
self
}
pub fn execute(self) -> anyhow::Result<()> {
pub fn execute(self) -> Result<()> {
self.index.set_updated_at(self.wtxn, &Utc::now())?;
// We get the faceted fields to be able to create the facet levels.
let faceted_fields = self.index.faceted_fields_ids(self.wtxn)?;
@ -102,7 +103,7 @@ impl<'t, 'u, 'i> Facets<'t, 'u, 'i> {
self.wtxn,
*self.index.facet_id_f64_docids.as_polymorph(),
content,
|_, _| anyhow::bail!("invalid facet number level merging"),
|_, _| Err(InternalError::IndexingMergingKeys { process: "facet number level" }),
WriteMethod::GetMergePut,
)?;
}
@ -132,7 +133,7 @@ fn compute_facet_number_levels<'t>(
level_group_size: NonZeroUsize,
min_level_size: NonZeroUsize,
field_id: u8,
) -> anyhow::Result<Reader<FileFuse>>
) -> Result<Reader<FileFuse>>
{
let first_level_size = db
.remap_key_type::<ByteSlice>()
@ -195,7 +196,7 @@ fn compute_faceted_documents_ids(
rtxn: &heed::RoTxn,
db: heed::Database<ByteSlice, CboRoaringBitmapCodec>,
field_id: u8,
) -> anyhow::Result<RoaringBitmap>
) -> Result<RoaringBitmap>
{
let mut documents_ids = RoaringBitmap::new();
@ -214,7 +215,7 @@ fn write_number_entry(
left: f64,
right: f64,
ids: &RoaringBitmap,
) -> anyhow::Result<()>
) -> Result<()>
{
let key = (field_id, level, left, right);
let key = FacetLevelValueF64Codec::bytes_encode(&key).ok_or(Error::Encoding)?;

92
milli/src/update/index_documents/merge_function.rs
View File

@ -1,75 +1,43 @@
use std::borrow::Cow;
use std::result::Result as StdResult;
use anyhow::{bail, ensure, Context};
use bstr::ByteSlice as _;
use fst::IntoStreamer;
use roaring::RoaringBitmap;
use crate::heed_codec::CboRoaringBitmapCodec;
use crate::Result;
const WORDS_FST_KEY: &[u8] = crate::index::WORDS_FST_KEY.as_bytes();
const FIELDS_IDS_MAP_KEY: &[u8] = crate::index::FIELDS_IDS_MAP_KEY.as_bytes();
const DOCUMENTS_IDS_KEY: &[u8] = crate::index::DOCUMENTS_IDS_KEY.as_bytes();
pub fn main_merge(key: &[u8], values: &[Cow<[u8]>]) -> anyhow::Result<Vec<u8>> {
match key {
WORDS_FST_KEY => {
let fsts: Vec<_> = values.iter().map(|v| fst::Set::new(v).unwrap()).collect();
// Union of the FSTs
let mut op = fst::set::OpBuilder::new();
fsts.iter().for_each(|fst| op.push(fst.into_stream()));
let op = op.r#union();
let mut build = fst::SetBuilder::memory();
build.extend_stream(op.into_stream()).unwrap();
Ok(build.into_inner().unwrap())
},
FIELDS_IDS_MAP_KEY => {
ensure!(values.windows(2).all(|vs| vs[0] == vs[1]), "fields ids map doesn't match");
Ok(values[0].to_vec())
},
DOCUMENTS_IDS_KEY => roaring_bitmap_merge(values),
otherwise => bail!("wut {:?}", otherwise),
}
/// Only the last value associated with an id is kept.
pub fn keep_latest_obkv(_key: &[u8], obkvs: &[Cow<[u8]>]) -> Result<Vec<u8>> {
Ok(obkvs.last().unwrap().clone().into_owned())
}
pub fn word_docids_merge(_key: &[u8], values: &[Cow<[u8]>]) -> anyhow::Result<Vec<u8>> {
roaring_bitmap_merge(values)
/// Merge all the obks in the order we see them.
pub fn merge_obkvs(_key: &[u8], obkvs: &[Cow<[u8]>]) -> Result<Vec<u8>> {
let mut iter = obkvs.iter();
let first = iter.next().map(|b| b.clone().into_owned()).unwrap();
Ok(iter.fold(first, |acc, current| {
let first = obkv::KvReader::new(&acc);
let second = obkv::KvReader::new(current);
let mut buffer = Vec::new();
merge_two_obkvs(first, second, &mut buffer);
buffer
}))
}
pub fn docid_word_positions_merge(key: &[u8], _values: &[Cow<[u8]>]) -> anyhow::Result<Vec<u8>> {
bail!("merging docid word positions is an error ({:?})", key.as_bstr())
// Union of multiple FSTs
pub fn fst_merge(_key: &[u8], values: &[Cow<[u8]>]) -> Result<Vec<u8>> {
let fsts = values.iter().map(fst::Set::new).collect::<StdResult<Vec<_>, _>>()?;
let op_builder: fst::set::OpBuilder = fsts.iter().map(|fst| fst.into_stream()).collect();
let op = op_builder.r#union();
let mut build = fst::SetBuilder::memory();
build.extend_stream(op.into_stream()).unwrap();
Ok(build.into_inner().unwrap())
}
pub fn field_id_docid_facet_values_merge(_key: &[u8], values: &[Cow<[u8]>]) -> anyhow::Result<Vec<u8>> {
let first = values.first().context("no value to merge")?;
ensure!(values.iter().all(|v| v == first), "invalid field id docid facet value merging");
Ok(first.to_vec())
}
pub fn words_pairs_proximities_docids_merge(_key: &[u8], values: &[Cow<[u8]>]) -> anyhow::Result<Vec<u8>> {
cbo_roaring_bitmap_merge(values)
}
pub fn word_prefix_level_positions_docids_merge(_key: &[u8], values: &[Cow<[u8]>]) -> anyhow::Result<Vec<u8>> {
cbo_roaring_bitmap_merge(values)
}
pub fn word_level_position_docids_merge(_key: &[u8], values: &[Cow<[u8]>]) -> anyhow::Result<Vec<u8>> {
cbo_roaring_bitmap_merge(values)
}
pub fn field_id_word_count_docids_merge(_key: &[u8], values: &[Cow<[u8]>]) -> anyhow::Result<Vec<u8>> {
cbo_roaring_bitmap_merge(values)
}
pub fn facet_field_value_docids_merge(_key: &[u8], values: &[Cow<[u8]>]) -> anyhow::Result<Vec<u8>> {
cbo_roaring_bitmap_merge(values)
}
pub fn documents_merge(key: &[u8], _values: &[Cow<[u8]>]) -> anyhow::Result<Vec<u8>> {
bail!("merging documents is an error ({:?})", key.as_bstr())
pub fn keep_first(_key: &[u8], values: &[Cow<[u8]>]) -> Result<Vec<u8>> {
Ok(values.first().unwrap().to_vec())
}
pub fn merge_two_obkvs(base: obkv::KvReader, update: obkv::KvReader, buffer: &mut Vec<u8>) {
@ -88,7 +56,7 @@ pub fn merge_two_obkvs(base: obkv::KvReader, update: obkv::KvReader, buffer: &mu
writer.finish().unwrap();
}
fn roaring_bitmap_merge(values: &[Cow<[u8]>]) -> anyhow::Result<Vec<u8>> {
pub fn roaring_bitmap_merge(_key: &[u8], values: &[Cow<[u8]>]) -> Result<Vec<u8>> {
let (head, tail) = values.split_first().unwrap();
let mut head = RoaringBitmap::deserialize_from(&head[..])?;
@ -102,7 +70,7 @@ fn roaring_bitmap_merge(values: &[Cow<[u8]>]) -> anyhow::Result<Vec<u8>> {
Ok(vec)
}
fn cbo_roaring_bitmap_merge(values: &[Cow<[u8]>]) -> anyhow::Result<Vec<u8>> {
pub fn cbo_roaring_bitmap_merge(_key: &[u8], values: &[Cow<[u8]>]) -> Result<Vec<u8>> {
let (head, tail) = values.split_first().unwrap();
let mut head = CboRoaringBitmapCodec::deserialize_from(&head[..])?;
@ -112,6 +80,6 @@ fn cbo_roaring_bitmap_merge(values: &[Cow<[u8]>]) -> anyhow::Result<Vec<u8>> {
}
let mut vec = Vec::new();
CboRoaringBitmapCodec::serialize_into(&head, &mut vec)?;
CboRoaringBitmapCodec::serialize_into(&head, &mut vec);
Ok(vec)
}

134
milli/src/update/index_documents/mod.rs
View File

@ -3,11 +3,11 @@ use std::collections::HashSet;
use std::fs::File;
use std::io::{self, Seek, SeekFrom, BufReader, BufRead};
use std::num::{NonZeroU32, NonZeroUsize};
use std::result::Result as StdResult;
use std::str;
use std::sync::mpsc::sync_channel;
use std::time::Instant;
use anyhow::Context;
use bstr::ByteSlice as _;
use chrono::Utc;
use grenad::{MergerIter, Writer, Sorter, Merger, Reader, FileFuse, CompressionType};
@ -18,18 +18,15 @@ use rayon::prelude::*;
use rayon::ThreadPool;
use serde::{Serialize, Deserialize};
use crate::index::Index;
use crate::error::{Error, InternalError};
use crate::{Index, Result};
use crate::update::{
Facets, WordsLevelPositions, WordPrefixDocids, WordsPrefixesFst, UpdateIndexingStep,
WordPrefixPairProximityDocids,
};
use self::store::{Store, Readers};
pub use self::merge_function::{
main_merge, word_docids_merge, words_pairs_proximities_docids_merge,
docid_word_positions_merge, documents_merge,
word_level_position_docids_merge, word_prefix_level_positions_docids_merge,
facet_field_value_docids_merge, field_id_docid_facet_values_merge,
field_id_word_count_docids_merge,
fst_merge, cbo_roaring_bitmap_merge, roaring_bitmap_merge, keep_first
};
pub use self::transform::{Transform, TransformOutput};
@ -60,14 +57,14 @@ pub fn create_writer(typ: CompressionType, level: Option<u32>, file: File) -> io
builder.build(file)
}
pub fn create_sorter(
merge: MergeFn,
pub fn create_sorter<E>(
merge: MergeFn<E>,
chunk_compression_type: CompressionType,
chunk_compression_level: Option<u32>,
chunk_fusing_shrink_size: Option<u64>,
max_nb_chunks: Option<usize>,
max_memory: Option<usize>,
) -> Sorter<MergeFn>
) -> Sorter<MergeFn<E>>
{
let mut builder = Sorter::builder(merge);
if let Some(shrink_size) = chunk_fusing_shrink_size {
@ -86,7 +83,7 @@ pub fn create_sorter(
builder.build()
}
pub fn writer_into_reader(writer: Writer<File>, shrink_size: Option<u64>) -> anyhow::Result<Reader<FileFuse>> {
pub fn writer_into_reader(writer: Writer<File>, shrink_size: Option<u64>) -> Result<Reader<FileFuse>> {
let mut file = writer.into_inner()?;
file.seek(SeekFrom::Start(0))?;
let file = if let Some(shrink_size) = shrink_size {
@ -97,19 +94,25 @@ pub fn writer_into_reader(writer: Writer<File>, shrink_size: Option<u64>) -> any
Reader::new(file).map_err(Into::into)
}
pub fn merge_readers(sources: Vec<Reader<FileFuse>>, merge: MergeFn) -> Merger<FileFuse, MergeFn> {
pub fn merge_readers<E>(
sources: Vec<Reader<FileFuse>>,
merge: MergeFn<E>,
) -> Merger<FileFuse, MergeFn<E>>
{
let mut builder = Merger::builder(merge);
builder.extend(sources);
builder.build()
}
pub fn merge_into_lmdb_database(
pub fn merge_into_lmdb_database<E>(
wtxn: &mut heed::RwTxn,
database: heed::PolyDatabase,
sources: Vec<Reader<FileFuse>>,
merge: MergeFn,
merge: MergeFn<E>,
method: WriteMethod,
) -> anyhow::Result<()>
) -> Result<()>
where
Error: From<E>,
{
debug!("Merging {} MTBL stores...", sources.len());
let before = Instant::now();
@ -127,13 +130,15 @@ pub fn merge_into_lmdb_database(
Ok(())
}
pub fn write_into_lmdb_database(
pub fn write_into_lmdb_database<E>(
wtxn: &mut heed::RwTxn,
database: heed::PolyDatabase,
mut reader: Reader<FileFuse>,
merge: MergeFn,
merge: MergeFn<E>,
method: WriteMethod,
) -> anyhow::Result<()>
) -> Result<()>
where
Error: From<E>,
{
debug!("Writing MTBL stores...");
let before = Instant::now();
@ -142,9 +147,7 @@ pub fn write_into_lmdb_database(
WriteMethod::Append => {
let mut out_iter = database.iter_mut::<_, ByteSlice, ByteSlice>(wtxn)?;
while let Some((k, v)) = reader.next()? {
out_iter.append(k, v).with_context(|| {
format!("writing {:?} into LMDB", k.as_bstr())
})?;
out_iter.append(k, v)?;
}
},
WriteMethod::GetMergePut => {
@ -152,7 +155,7 @@ pub fn write_into_lmdb_database(
let mut iter = database.prefix_iter_mut::<_, ByteSlice, ByteSlice>(wtxn, k)?;
match iter.next().transpose()? {
Some((key, old_val)) if key == k => {
let vals = vec![Cow::Borrowed(old_val), Cow::Borrowed(v)];
let vals = &[Cow::Borrowed(old_val), Cow::Borrowed(v)][..];
let val = merge(k, &vals)?;
iter.put_current(k, &val)?;
},
@ -169,13 +172,16 @@ pub fn write_into_lmdb_database(
Ok(())
}
pub fn sorter_into_lmdb_database(
pub fn sorter_into_lmdb_database<E>(
wtxn: &mut heed::RwTxn,
database: heed::PolyDatabase,
sorter: Sorter<MergeFn>,
merge: MergeFn,
sorter: Sorter<MergeFn<E>>,
merge: MergeFn<E>,
method: WriteMethod,
) -> anyhow::Result<()>
) -> Result<()>
where
Error: From<E>,
Error: From<grenad::Error<E>>
{
debug!("Writing MTBL sorter...");
let before = Instant::now();
@ -192,21 +198,21 @@ pub fn sorter_into_lmdb_database(
Ok(())
}
fn merger_iter_into_lmdb_database<R: io::Read>(
fn merger_iter_into_lmdb_database<R: io::Read, E>(
wtxn: &mut heed::RwTxn,
database: heed::PolyDatabase,
mut sorter: MergerIter<R, MergeFn>,
merge: MergeFn,
mut sorter: MergerIter<R, MergeFn<E>>,
merge: MergeFn<E>,
method: WriteMethod,
) -> anyhow::Result<()>
) -> Result<()>
where
Error: From<E>,
{
match method {
WriteMethod::Append => {
let mut out_iter = database.iter_mut::<_, ByteSlice, ByteSlice>(wtxn)?;
while let Some((k, v)) = sorter.next()? {
out_iter.append(k, v).with_context(|| {
format!("writing {:?} into LMDB", k.as_bstr())
})?;
out_iter.append(k, v)?;
}
},
WriteMethod::GetMergePut => {
@ -215,7 +221,10 @@ fn merger_iter_into_lmdb_database<R: io::Read>(
match iter.next().transpose()? {
Some((key, old_val)) if key == k => {
let vals = vec![Cow::Borrowed(old_val), Cow::Borrowed(v)];
let val = merge(k, &vals).expect("merge failed");
let val = merge(k, &vals).map_err(|_| {
// TODO just wrap this error?
InternalError::IndexingMergingKeys { process: "get-put-merge" }
})?;
iter.put_current(k, &val)?;
},
_ => {
@ -322,7 +331,7 @@ impl<'t, 'u, 'i, 'a> IndexDocuments<'t, 'u, 'i, 'a> {
self.autogenerate_docids = false;
}
pub fn execute<R, F>(self, reader: R, progress_callback: F) -> anyhow::Result<DocumentAdditionResult>
pub fn execute<R, F>(self, reader: R, progress_callback: F) -> Result<DocumentAdditionResult>
where
R: io::Read,
F: Fn(UpdateIndexingStep, u64) + Sync,
@ -369,7 +378,7 @@ impl<'t, 'u, 'i, 'a> IndexDocuments<'t, 'u, 'i, 'a> {
Ok(DocumentAdditionResult { nb_documents })
}
pub fn execute_raw<F>(self, output: TransformOutput, progress_callback: F) -> anyhow::Result<()>
pub fn execute_raw<F>(self, output: TransformOutput, progress_callback: F) -> Result<()>
where
F: Fn(UpdateIndexingStep) + Sync
{
@ -407,15 +416,12 @@ impl<'t, 'u, 'i, 'a> IndexDocuments<'t, 'u, 'i, 'a> {
debug!("{} documents actually deleted", deleted_documents_count);
}
let mmap;
let bytes = if documents_count == 0 {
&[][..]
} else {
mmap = unsafe { Mmap::map(&documents_file).context("mmaping the transform documents file")? };
&mmap
};
if documents_count == 0 {
return Ok(());
}
let documents = grenad::Reader::new(bytes).unwrap();
let bytes = unsafe { Mmap::map(&documents_file)? };
let documents = grenad::Reader::new(bytes.as_bytes()).unwrap();
// The enum which indicates the type of the readers
// merges that are potentially done on different threads.
@ -481,7 +487,7 @@ impl<'t, 'u, 'i, 'a> IndexDocuments<'t, 'u, 'i, 'a> {
&progress_callback,
)
})
.collect::<Result<Vec<_>, _>>()?;
.collect::<StdResult<Vec<_>, _>>()?;
let mut main_readers = Vec::with_capacity(readers.len());
let mut word_docids_readers = Vec::with_capacity(readers.len());
@ -539,22 +545,22 @@ impl<'t, 'u, 'i, 'a> IndexDocuments<'t, 'u, 'i, 'a> {
debug!("Merging the main, word docids and words pairs proximity docids in parallel...");
rayon::spawn(move || {
vec![
(DatabaseType::Main, main_readers, main_merge as MergeFn),
(DatabaseType::WordDocids, word_docids_readers, word_docids_merge),
(DatabaseType::Main, main_readers, fst_merge as MergeFn<_>),
(DatabaseType::WordDocids, word_docids_readers, roaring_bitmap_merge),
(
DatabaseType::FacetLevel0NumbersDocids,
facet_field_numbers_docids_readers,
facet_field_value_docids_merge,
cbo_roaring_bitmap_merge,
),
(
DatabaseType::WordLevel0PositionDocids,
word_level_position_docids_readers,
word_level_position_docids_merge,
cbo_roaring_bitmap_merge,
),
(
DatabaseType::FieldIdWordCountDocids,
field_id_word_count_docids_readers,
field_id_word_count_docids_merge,
cbo_roaring_bitmap_merge,
),
]
.into_par_iter()
@ -574,7 +580,7 @@ impl<'t, 'u, 'i, 'a> IndexDocuments<'t, 'u, 'i, 'a> {
facet_field_strings_docids_readers,
field_id_docid_facet_numbers_readers,
field_id_docid_facet_strings_readers,
)) as anyhow::Result<_>
)) as Result<_>
})?;
let (
@ -622,12 +628,12 @@ impl<'t, 'u, 'i, 'a> IndexDocuments<'t, 'u, 'i, 'a> {
total_databases,
});
debug!("Writing the docid word positions into LMDB on disk...");
debug!("Inserting the docid word positions into LMDB on disk...");
merge_into_lmdb_database(
self.wtxn,
*self.index.docid_word_positions.as_polymorph(),
docid_word_positions_readers,
docid_word_positions_merge,
keep_first,
write_method
)?;
@ -637,12 +643,12 @@ impl<'t, 'u, 'i, 'a> IndexDocuments<'t, 'u, 'i, 'a> {
total_databases,
});
debug!("Writing the documents into LMDB on disk...");
debug!("Inserting the documents into LMDB on disk...");
merge_into_lmdb_database(
self.wtxn,
*self.index.documents.as_polymorph(),
documents_readers,
documents_merge,
keep_first,
write_method
)?;
@ -657,7 +663,7 @@ impl<'t, 'u, 'i, 'a> IndexDocuments<'t, 'u, 'i, 'a> {
self.wtxn,
*self.index.facet_id_string_docids.as_polymorph(),
facet_field_strings_docids_readers,
facet_field_value_docids_merge,
cbo_roaring_bitmap_merge,
write_method,
)?;
@ -672,7 +678,7 @@ impl<'t, 'u, 'i, 'a> IndexDocuments<'t, 'u, 'i, 'a> {
self.wtxn,
*self.index.field_id_docid_facet_f64s.as_polymorph(),
field_id_docid_facet_numbers_readers,
field_id_docid_facet_values_merge,
keep_first,
write_method,
)?;
@ -687,7 +693,7 @@ impl<'t, 'u, 'i, 'a> IndexDocuments<'t, 'u, 'i, 'a> {
self.wtxn,
*self.index.field_id_docid_facet_strings.as_polymorph(),
field_id_docid_facet_strings_readers,
field_id_docid_facet_values_merge,
keep_first,
write_method,
)?;
@ -702,7 +708,7 @@ impl<'t, 'u, 'i, 'a> IndexDocuments<'t, 'u, 'i, 'a> {
self.wtxn,
*self.index.word_pair_proximity_docids.as_polymorph(),
words_pairs_proximities_docids_readers,
words_pairs_proximities_docids_merge,
cbo_roaring_bitmap_merge,
write_method,
)?;
@ -721,7 +727,7 @@ impl<'t, 'u, 'i, 'a> IndexDocuments<'t, 'u, 'i, 'a> {
self.wtxn,
self.index.main,
content,
main_merge,
fst_merge,
WriteMethod::GetMergePut,
)?;
},
@ -732,7 +738,7 @@ impl<'t, 'u, 'i, 'a> IndexDocuments<'t, 'u, 'i, 'a> {
self.wtxn,
db,
content,
word_docids_merge,
roaring_bitmap_merge,
write_method,
)?;
},
@ -743,7 +749,7 @@ impl<'t, 'u, 'i, 'a> IndexDocuments<'t, 'u, 'i, 'a> {
self.wtxn,
db,
content,
facet_field_value_docids_merge,
cbo_roaring_bitmap_merge,
write_method,
)?;
},
@ -754,7 +760,7 @@ impl<'t, 'u, 'i, 'a> IndexDocuments<'t, 'u, 'i, 'a> {
self.wtxn,
db,
content,
field_id_word_count_docids_merge,
cbo_roaring_bitmap_merge,
write_method,
)?;
},
@ -765,7 +771,7 @@ impl<'t, 'u, 'i, 'a> IndexDocuments<'t, 'u, 'i, 'a> {
self.wtxn,
db,
content,
word_level_position_docids_merge,
cbo_roaring_bitmap_merge,
write_method,
)?;
}

185
milli/src/update/index_documents/store.rs
View File

@ -6,7 +6,6 @@ use std::iter::FromIterator;
use std::time::Instant;
use std::{cmp, iter};
use anyhow::Context;
use bstr::ByteSlice as _;
use fst::Set;
use grenad::{Reader, FileFuse, Writer, Sorter, CompressionType};
@ -19,24 +18,21 @@ use roaring::RoaringBitmap;
use serde_json::Value;
use tempfile::tempfile;
use crate::error::{Error, InternalError, SerializationError};
use crate::heed_codec::facet::{FacetValueStringCodec, FacetLevelValueF64Codec};
use crate::heed_codec::facet::{FieldDocIdFacetStringCodec, FieldDocIdFacetF64Codec};
use crate::heed_codec::{BoRoaringBitmapCodec, CboRoaringBitmapCodec};
use crate::update::UpdateIndexingStep;
use crate::{json_to_string, SmallVec32, Position, DocumentId, FieldId};
use crate::{json_to_string, SmallVec32, Position, DocumentId, FieldId, Result};
use super::{MergeFn, create_writer, create_sorter, writer_into_reader};
use super::merge_function::{
main_merge, word_docids_merge, words_pairs_proximities_docids_merge,
word_level_position_docids_merge, facet_field_value_docids_merge,
field_id_docid_facet_values_merge, field_id_word_count_docids_merge,
};
use super::merge_function::{fst_merge, keep_first, roaring_bitmap_merge, cbo_roaring_bitmap_merge};
const LMDB_MAX_KEY_LENGTH: usize = 511;
const ONE_KILOBYTE: usize = 1024 * 1024;
const MAX_POSITION: usize = 1000;
const WORDS_FST_KEY: &[u8] = crate::index::WORDS_FST_KEY.as_bytes();
const WORDS_FST_KEY: &[u8] = crate::index::main_key::WORDS_FST_KEY.as_bytes();
pub struct Readers {
pub main: Reader<FileFuse>,
@ -70,15 +66,15 @@ pub struct Store<'s, A> {
chunk_compression_level: Option<u32>,
chunk_fusing_shrink_size: Option<u64>,
// MTBL sorters
main_sorter: Sorter<MergeFn>,
word_docids_sorter: Sorter<MergeFn>,
words_pairs_proximities_docids_sorter: Sorter<MergeFn>,
word_level_position_docids_sorter: Sorter<MergeFn>,
field_id_word_count_docids_sorter: Sorter<MergeFn>,
facet_field_numbers_docids_sorter: Sorter<MergeFn>,
facet_field_strings_docids_sorter: Sorter<MergeFn>,
field_id_docid_facet_numbers_sorter: Sorter<MergeFn>,
field_id_docid_facet_strings_sorter: Sorter<MergeFn>,
main_sorter: Sorter<MergeFn<Error>>,
word_docids_sorter: Sorter<MergeFn<Error>>,
words_pairs_proximities_docids_sorter: Sorter<MergeFn<Error>>,
word_level_position_docids_sorter: Sorter<MergeFn<Error>>,
field_id_word_count_docids_sorter: Sorter<MergeFn<Error>>,
facet_field_numbers_docids_sorter: Sorter<MergeFn<Error>>,
facet_field_strings_docids_sorter: Sorter<MergeFn<Error>>,
field_id_docid_facet_numbers_sorter: Sorter<MergeFn<Error>>,
field_id_docid_facet_strings_sorter: Sorter<MergeFn<Error>>,
// MTBL writers
docid_word_positions_writer: Writer<File>,
documents_writer: Writer<File>,
@ -97,14 +93,14 @@ impl<'s, A: AsRef<[u8]>> Store<'s, A> {
chunk_compression_level: Option<u32>,
chunk_fusing_shrink_size: Option<u64>,
stop_words: Option<&'s Set<A>>,
) -> anyhow::Result<Self>
) -> Result<Self>
{
// We divide the max memory by the number of sorter the Store have.
let max_memory = max_memory.map(|mm| cmp::max(ONE_KILOBYTE, mm / 5));
let linked_hash_map_size = linked_hash_map_size.unwrap_or(500);
let main_sorter = create_sorter(
main_merge,
fst_merge,
chunk_compression_type,
chunk_compression_level,
chunk_fusing_shrink_size,
@ -112,7 +108,7 @@ impl<'s, A: AsRef<[u8]>> Store<'s, A> {
max_memory,
);
let word_docids_sorter = create_sorter(
word_docids_merge,
roaring_bitmap_merge,
chunk_compression_type,
chunk_compression_level,
chunk_fusing_shrink_size,
@ -120,7 +116,7 @@ impl<'s, A: AsRef<[u8]>> Store<'s, A> {
max_memory,
);
let words_pairs_proximities_docids_sorter = create_sorter(
words_pairs_proximities_docids_merge,
cbo_roaring_bitmap_merge,
chunk_compression_type,
chunk_compression_level,
chunk_fusing_shrink_size,
@ -128,7 +124,7 @@ impl<'s, A: AsRef<[u8]>> Store<'s, A> {
max_memory,
);
let word_level_position_docids_sorter = create_sorter(
word_level_position_docids_merge,
cbo_roaring_bitmap_merge,
chunk_compression_type,
chunk_compression_level,
chunk_fusing_shrink_size,
@ -136,7 +132,7 @@ impl<'s, A: AsRef<[u8]>> Store<'s, A> {
max_memory,
);
let field_id_word_count_docids_sorter = create_sorter(
field_id_word_count_docids_merge,
cbo_roaring_bitmap_merge,
chunk_compression_type,
chunk_compression_level,
chunk_fusing_shrink_size,
@ -144,7 +140,7 @@ impl<'s, A: AsRef<[u8]>> Store<'s, A> {
max_memory,
);
let facet_field_numbers_docids_sorter = create_sorter(
facet_field_value_docids_merge,
cbo_roaring_bitmap_merge,
chunk_compression_type,
chunk_compression_level,
chunk_fusing_shrink_size,
@ -152,7 +148,7 @@ impl<'s, A: AsRef<[u8]>> Store<'s, A> {
max_memory,
);
let facet_field_strings_docids_sorter = create_sorter(
facet_field_value_docids_merge,
cbo_roaring_bitmap_merge,
chunk_compression_type,
chunk_compression_level,
chunk_fusing_shrink_size,
@ -160,7 +156,7 @@ impl<'s, A: AsRef<[u8]>> Store<'s, A> {
max_memory,
);
let field_id_docid_facet_numbers_sorter = create_sorter(
field_id_docid_facet_values_merge,
keep_first,
chunk_compression_type,
chunk_compression_level,
chunk_fusing_shrink_size,
@ -168,7 +164,7 @@ impl<'s, A: AsRef<[u8]>> Store<'s, A> {
Some(1024 * 1024 * 1024), // 1MB
);
let field_id_docid_facet_strings_sorter = create_sorter(
field_id_docid_facet_values_merge,
keep_first,
chunk_compression_type,
chunk_compression_level,
chunk_fusing_shrink_size,
@ -225,7 +221,7 @@ impl<'s, A: AsRef<[u8]>> Store<'s, A> {
}
// Save the documents ids under the position and word we have seen it.
fn insert_word_docid(&mut self, word: &str, id: DocumentId) -> anyhow::Result<()> {
fn insert_word_docid(&mut self, word: &str, id: DocumentId) -> Result<()> {
// if get_refresh finds the element it is assured to be at the end of the linked hash map.
match self.word_docids.get_refresh(word.as_bytes()) {
Some(old) => { old.insert(id); },
@ -250,7 +246,7 @@ impl<'s, A: AsRef<[u8]>> Store<'s, A> {
field_id: FieldId,
value: OrderedFloat<f64>,
id: DocumentId,
) -> anyhow::Result<()>
) -> Result<()>
{
let sorter = &mut self.field_id_docid_facet_numbers_sorter;
Self::write_field_id_docid_facet_number_value(sorter, field_id, id, value)?;
@ -283,7 +279,7 @@ impl<'s, A: AsRef<[u8]>> Store<'s, A> {
field_id: FieldId,
value: String,
id: DocumentId,
) -> anyhow::Result<()>
) -> Result<()>
{
let sorter = &mut self.field_id_docid_facet_strings_sorter;
Self::write_field_id_docid_facet_string_value(sorter, field_id, id, &value)?;
@ -315,7 +311,7 @@ impl<'s, A: AsRef<[u8]>> Store<'s, A> {
&mut self,
words_pairs_proximities: impl IntoIterator<Item=((&'a str, &'a str), u8)>,
id: DocumentId,
) -> anyhow::Result<()>
) -> Result<()>
{
for ((w1, w2), prox) in words_pairs_proximities {
let w1 = SmallVec32::from(w1.as_bytes());
@ -354,7 +350,7 @@ impl<'s, A: AsRef<[u8]>> Store<'s, A> {
facet_numbers_values: &mut HashMap<FieldId, Vec<f64>>,
facet_strings_values: &mut HashMap<FieldId, Vec<String>>,
record: &[u8],
) -> anyhow::Result<()>
) -> Result<()>
{
// We compute the list of words pairs proximities (self-join) and write it directly to disk.
let words_pair_proximities = compute_words_pair_proximities(&words_positions);
@ -389,10 +385,12 @@ impl<'s, A: AsRef<[u8]>> Store<'s, A> {
Ok(())
}
fn write_words_pairs_proximities(
sorter: &mut Sorter<MergeFn>,
fn write_words_pairs_proximities<E>(
sorter: &mut Sorter<MergeFn<E>>,
iter: impl IntoIterator<Item=((SmallVec32<u8>, SmallVec32<u8>, u8), RoaringBitmap)>,
) -> anyhow::Result<()>
) -> Result<()>
where
Error: From<E>,
{
let mut key = Vec::new();
let mut buffer = Vec::new();
@ -407,7 +405,7 @@ impl<'s, A: AsRef<[u8]>> Store<'s, A> {
// We serialize the document ids into a buffer
buffer.clear();
buffer.reserve(CboRoaringBitmapCodec::serialized_size(&docids));
CboRoaringBitmapCodec::serialize_into(&docids, &mut buffer)?;
CboRoaringBitmapCodec::serialize_into(&docids, &mut buffer);
// that we write under the generated key into MTBL
if lmdb_key_valid_size(&key) {
sorter.insert(&key, &buffer)?;
@ -421,10 +419,11 @@ impl<'s, A: AsRef<[u8]>> Store<'s, A> {
writer: &mut Writer<File>,
id: DocumentId,
words_positions: &HashMap<String, SmallVec32<Position>>,
) -> anyhow::Result<()>
) -> Result<()>
{
// We prefix the words by the document id.
let mut key = id.to_be_bytes().to_vec();
let mut buffer = Vec::new();
let base_size = key.len();
// We order the words lexicographically, this way we avoid passing by a sorter.
@ -433,24 +432,28 @@ impl<'s, A: AsRef<[u8]>> Store<'s, A> {
for (word, positions) in words_positions {
key.truncate(base_size);
key.extend_from_slice(word.as_bytes());
buffer.clear();
// We serialize the positions into a buffer.
let positions = RoaringBitmap::from_iter(positions.iter().cloned());
let bytes = BoRoaringBitmapCodec::bytes_encode(&positions)
.with_context(|| "could not serialize positions")?;
BoRoaringBitmapCodec::serialize_into(&positions, &mut buffer);
// that we write under the generated key into MTBL
if lmdb_key_valid_size(&key) {
writer.insert(&key, &bytes)?;
writer.insert(&key, &buffer)?;
}
}
Ok(())
}
fn write_word_position_docids(
writer: &mut Sorter<MergeFn>,
fn write_word_position_docids<E>(
writer: &mut Sorter<MergeFn<E>>,
document_id: DocumentId,
words_positions: &HashMap<String, SmallVec32<Position>>,
) -> anyhow::Result<()>
) -> Result<()>
where
Error: From<E>,
{
let mut key_buffer = Vec::new();
let mut data_buffer = Vec::new();
@ -469,8 +472,7 @@ impl<'s, A: AsRef<[u8]>> Store<'s, A> {
data_buffer.clear();
let positions = RoaringBitmap::from_iter(Some(document_id));
// We serialize the positions into a buffer.
CboRoaringBitmapCodec::serialize_into(&positions, &mut data_buffer)
.with_context(|| "could not serialize positions")?;
CboRoaringBitmapCodec::serialize_into(&positions, &mut data_buffer);
// that we write under the generated key into MTBL
if lmdb_key_valid_size(&key_buffer) {
@ -482,56 +484,71 @@ impl<'s, A: AsRef<[u8]>> Store<'s, A> {
Ok(())
}
fn write_facet_field_string_docids<I>(
sorter: &mut Sorter<MergeFn>,
fn write_facet_field_string_docids<I, E>(
sorter: &mut Sorter<MergeFn<E>>,
iter: I,
) -> anyhow::Result<()>
where I: IntoIterator<Item=((FieldId, String), RoaringBitmap)>
) -> Result<()>
where
I: IntoIterator<Item=((FieldId, String), RoaringBitmap)>,
Error: From<E>,
{
let mut key_buffer = Vec::new();
let mut data_buffer = Vec::new();
for ((field_id, value), docids) in iter {
let key = FacetValueStringCodec::bytes_encode(&(field_id, &value))
.map(Cow::into_owned)
.context("could not serialize facet key")?;
let bytes = CboRoaringBitmapCodec::bytes_encode(&docids)
.context("could not serialize docids")?;
if lmdb_key_valid_size(&key) {
sorter.insert(&key, &bytes)?;
key_buffer.clear();
data_buffer.clear();
FacetValueStringCodec::serialize_into(field_id, &value, &mut key_buffer);
CboRoaringBitmapCodec::serialize_into(&docids, &mut data_buffer);
if lmdb_key_valid_size(&key_buffer) {
sorter.insert(&key_buffer, &data_buffer)?;
}
}
Ok(())
}
fn write_facet_field_number_docids<I>(
sorter: &mut Sorter<MergeFn>,
fn write_facet_field_number_docids<I, E>(
sorter: &mut Sorter<MergeFn<E>>,
iter: I,
) -> anyhow::Result<()>
where I: IntoIterator<Item=((FieldId, OrderedFloat<f64>), RoaringBitmap)>
) -> Result<()>
where
I: IntoIterator<Item=((FieldId, OrderedFloat<f64>), RoaringBitmap)>,
Error: From<E>,
{
let mut data_buffer = Vec::new();
for ((field_id, value), docids) in iter {
data_buffer.clear();
let key = FacetLevelValueF64Codec::bytes_encode(&(field_id, 0, *value, *value))
.map(Cow::into_owned)
.context("could not serialize facet key")?;
let bytes = CboRoaringBitmapCodec::bytes_encode(&docids)
.context("could not serialize docids")?;
.ok_or(SerializationError::Encoding { db_name: Some("facet level value") })?;
CboRoaringBitmapCodec::serialize_into(&docids, &mut data_buffer);
if lmdb_key_valid_size(&key) {
sorter.insert(&key, &bytes)?;
sorter.insert(&key, &data_buffer)?;
}
}
Ok(())
}
fn write_field_id_docid_facet_number_value(
sorter: &mut Sorter<MergeFn>,
fn write_field_id_docid_facet_number_value<E>(
sorter: &mut Sorter<MergeFn<E>>,
field_id: FieldId,
document_id: DocumentId,
value: OrderedFloat<f64>,
) -> anyhow::Result<()>
) -> Result<()>
where
Error: From<E>,
{
let key = FieldDocIdFacetF64Codec::bytes_encode(&(field_id, document_id, *value))
.map(Cow::into_owned)
.context("could not serialize facet key")?;
.ok_or(SerializationError::Encoding { db_name: Some("facet level value") })?;
if lmdb_key_valid_size(&key) {
sorter.insert(&key, &[])?;
@ -540,26 +557,30 @@ impl<'s, A: AsRef<[u8]>> Store<'s, A> {
Ok(())
}
fn write_field_id_docid_facet_string_value(
sorter: &mut Sorter<MergeFn>,
fn write_field_id_docid_facet_string_value<E>(
sorter: &mut Sorter<MergeFn<E>>,
field_id: FieldId,
document_id: DocumentId,
value: &str,
) -> anyhow::Result<()>
) -> Result<()>
where
Error: From<E>,
{
let key = FieldDocIdFacetStringCodec::bytes_encode(&(field_id, document_id, value))
.map(Cow::into_owned)
.context("could not serialize facet key")?;
let mut buffer = Vec::new();
if lmdb_key_valid_size(&key) {
sorter.insert(&key, &[])?;
FieldDocIdFacetStringCodec::serialize_into(field_id, document_id, value, &mut buffer);
if lmdb_key_valid_size(&buffer) {
sorter.insert(&buffer, &[])?;
}
Ok(())
}
fn write_word_docids<I>(sorter: &mut Sorter<MergeFn>, iter: I) -> anyhow::Result<()>
where I: IntoIterator<Item=(SmallVec32<u8>, RoaringBitmap)>
fn write_word_docids<I, E>(sorter: &mut Sorter<MergeFn<E>>, iter: I) -> Result<()>
where
I: IntoIterator<Item=(SmallVec32<u8>, RoaringBitmap)>,
Error: From<E>,
{
let mut key = Vec::new();
let mut buffer = Vec::new();
@ -589,7 +610,7 @@ impl<'s, A: AsRef<[u8]>> Store<'s, A> {
num_threads: usize,
log_every_n: Option<usize>,
mut progress_callback: F,
) -> anyhow::Result<Readers>
) -> Result<Readers>
where F: FnMut(UpdateIndexingStep),
{
debug!("{:?}: Indexing in a Store...", thread_index);
@ -618,7 +639,7 @@ impl<'s, A: AsRef<[u8]>> Store<'s, A> {
for (attr, content) in document.iter() {
if self.faceted_fields.contains(&attr) || self.searchable_fields.contains(&attr) {
let value = serde_json::from_slice(content)?;
let value = serde_json::from_slice(content).map_err(InternalError::SerdeJson)?;
let (facet_numbers, facet_strings) = extract_facet_values(&value);
facet_numbers_values.entry(attr).or_insert_with(Vec::new).extend(facet_numbers);
@ -672,7 +693,7 @@ impl<'s, A: AsRef<[u8]>> Store<'s, A> {
Ok(readers)
}
fn finish(mut self) -> anyhow::Result<Readers> {
fn finish(mut self) -> Result<Readers> {
let comp_type = self.chunk_compression_type;
let comp_level = self.chunk_compression_level;
let shrink_size = self.chunk_fusing_shrink_size;
@ -706,7 +727,7 @@ impl<'s, A: AsRef<[u8]>> Store<'s, A> {
let mut docids_buffer = Vec::new();
for ((fid, count), docids) in self.field_id_word_count_docids {
docids_buffer.clear();
CboRoaringBitmapCodec::serialize_into(&docids, &mut docids_buffer)?;
CboRoaringBitmapCodec::serialize_into(&docids, &mut docids_buffer);
self.field_id_word_count_docids_sorter.insert([fid, count], &docids_buffer)?;
}

103
milli/src/update/index_documents/transform.rs
View File

@ -2,16 +2,20 @@ use std::borrow::Cow;
use std::fs::File;
use std::io::{Read, Seek, SeekFrom};
use std::iter::Peekable;
use std::result::Result as StdResult;
use std::time::Instant;
use anyhow::{anyhow, Context};
use grenad::CompressionType;
use log::info;
use roaring::RoaringBitmap;
use serde_json::{Map, Value};
use crate::{Index, BEU32, MergeFn, FieldsIdsMap, ExternalDocumentsIds, FieldId, FieldsDistribution};
use crate::error::{Error, UserError, InternalError};
use crate::index::db_name;
use crate::update::index_documents::merge_function::{merge_obkvs, keep_latest_obkv};
use crate::update::{AvailableDocumentsIds, UpdateIndexingStep};
use crate::{BEU32, MergeFn, FieldsIdsMap, ExternalDocumentsIds, FieldId, FieldsDistribution};
use crate::{Index, Result};
use super::merge_function::merge_two_obkvs;
use super::{create_writer, create_sorter, IndexDocumentsMethod};
@ -52,7 +56,7 @@ fn is_primary_key(field: impl AsRef<str>) -> bool {
}
impl Transform<'_, '_> {
pub fn output_from_json<R, F>(self, reader: R, progress_callback: F) -> anyhow::Result<TransformOutput>
pub fn output_from_json<R, F>(self, reader: R, progress_callback: F) -> Result<TransformOutput>
where
R: Read,
F: Fn(UpdateIndexingStep) + Sync,
@ -60,7 +64,7 @@ impl Transform<'_, '_> {
self.output_from_generic_json(reader, false, progress_callback)
}
pub fn output_from_json_stream<R, F>(self, reader: R, progress_callback: F) -> anyhow::Result<TransformOutput>
pub fn output_from_json_stream<R, F>(self, reader: R, progress_callback: F) -> Result<TransformOutput>
where
R: Read,
F: Fn(UpdateIndexingStep) + Sync,
@ -73,7 +77,7 @@ impl Transform<'_, '_> {
reader: R,
is_stream: bool,
progress_callback: F,
) -> anyhow::Result<TransformOutput>
) -> Result<TransformOutput>
where
R: Read,
F: Fn(UpdateIndexingStep) + Sync,
@ -87,7 +91,7 @@ impl Transform<'_, '_> {
let iter = Box::new(iter) as Box<dyn Iterator<Item=_>>;
iter.peekable()
} else {
let vec: Vec<_> = serde_json::from_reader(reader)?;
let vec: Vec<_> = serde_json::from_reader(reader).map_err(UserError::SerdeJson)?;
let iter = vec.into_iter().map(Ok);
let iter = Box::new(iter) as Box<dyn Iterator<Item=_>>;
iter.peekable()
@ -95,9 +99,12 @@ impl Transform<'_, '_> {
// We extract the primary key from the first document in
// the batch if it hasn't already been defined in the index
let first = match documents.peek().map(Result::as_ref).transpose() {
let first = match documents.peek().map(StdResult::as_ref).transpose() {
Ok(first) => first,
Err(_) => return Err(documents.next().unwrap().unwrap_err().into()),
Err(_) => {
let error = documents.next().unwrap().unwrap_err();
return Err(UserError::SerdeJson(error).into());
},
};
let alternative_name = first.and_then(|doc| doc.keys().find(|f| is_primary_key(f)).cloned());
@ -144,7 +151,7 @@ impl Transform<'_, '_> {
let mut documents_count = 0;
for result in documents {
let document = result?;
let document = result.map_err(UserError::SerdeJson)?;
if self.log_every_n.map_or(false, |len| documents_count % len == 0) {
progress_callback(UpdateIndexingStep::TransformFromUserIntoGenericFormat {
@ -157,7 +164,7 @@ impl Transform<'_, '_> {
// We prepare the fields ids map with the documents keys.
for (key, _value) in &document {
fields_ids_map.insert(&key).context("field id limit reached")?;
fields_ids_map.insert(&key).ok_or(UserError::AttributeLimitReached)?;
}
// We retrieve the user id from the document based on the primary key name,
@ -166,11 +173,13 @@ impl Transform<'_, '_> {
Some(value) => match value {
Value::String(string) => Cow::Borrowed(string.as_str()),
Value::Number(number) => Cow::Owned(number.to_string()),
_ => return Err(anyhow!("documents ids must be either strings or numbers")),
content => return Err(UserError::InvalidDocumentId {
document_id: content.clone(),
}.into()),
},
None => {
if !self.autogenerate_docids {
return Err(anyhow!("missing primary key"));
return Err(UserError::MissingPrimaryKey.into());
}
let uuid = uuid::Uuid::new_v4().to_hyphenated().encode_lower(&mut uuid_buffer);
Cow::Borrowed(uuid)
@ -185,13 +194,15 @@ impl Transform<'_, '_> {
// and this should be the document id we return the one we generated.
if let Some(value) = document.get(name) {
// We serialize the attribute values.
serde_json::to_writer(&mut json_buffer, value)?;
serde_json::to_writer(&mut json_buffer, value).map_err(InternalError::SerdeJson)?;
writer.insert(field_id, &json_buffer)?;
}
// We validate the document id [a-zA-Z0-9\-_].
if field_id == primary_key_id && validate_document_id(&external_id).is_none() {
return Err(anyhow!("invalid document id: {:?}", external_id));
return Err(UserError::InvalidDocumentId {
document_id: Value::from(external_id),
}.into());
}
}
@ -216,7 +227,7 @@ impl Transform<'_, '_> {
)
}
pub fn output_from_csv<R, F>(self, reader: R, progress_callback: F) -> anyhow::Result<TransformOutput>
pub fn output_from_csv<R, F>(self, reader: R, progress_callback: F) -> Result<TransformOutput>
where
R: Read,
F: Fn(UpdateIndexingStep) + Sync,
@ -225,12 +236,12 @@ impl Transform<'_, '_> {
let external_documents_ids = self.index.external_documents_ids(self.rtxn).unwrap();
let mut csv = csv::Reader::from_reader(reader);
let headers = csv.headers()?;
let headers = csv.headers().map_err(UserError::Csv)?;
let mut fields_ids = Vec::new();
// Generate the new fields ids based on the current fields ids and this CSV headers.
for (i, header) in headers.iter().enumerate() {
let id = fields_ids_map.insert(header).context("field id limit reached)")?;
let id = fields_ids_map.insert(header).ok_or(UserError::AttributeLimitReached)?;
fields_ids.push((id, i));
}
@ -246,7 +257,7 @@ impl Transform<'_, '_> {
// Returns the field id in the fields ids map, create an "id" field
// in case it is not in the current headers.
let alternative_name = primary_key_pos.map(|pos| headers[pos].to_string());
let (primary_key_id, _) = compute_primary_key_pair(
let (primary_key_id, primary_key_name) = compute_primary_key_pair(
self.index.primary_key(self.rtxn)?,
&mut fields_ids_map,
alternative_name,
@ -280,7 +291,7 @@ impl Transform<'_, '_> {
let mut documents_count = 0;
let mut record = csv::StringRecord::new();
while csv.read_record(&mut record)? {
while csv.read_record(&mut record).map_err(UserError::Csv)? {
obkv_buffer.clear();
let mut writer = obkv::KvWriter::new(&mut obkv_buffer);
@ -297,7 +308,9 @@ impl Transform<'_, '_> {
// We validate the document id [a-zA-Z0-9\-_].
match validate_document_id(&external_id) {
Some(valid) => valid,
None => return Err(anyhow!("invalid document id: {:?}", external_id)),
None => return Err(UserError::InvalidDocumentId {
document_id: Value::from(external_id),
}.into()),
}
},
None => uuid::Uuid::new_v4().to_hyphenated().encode_lower(&mut uuid_buffer),
@ -315,7 +328,7 @@ impl Transform<'_, '_> {
for (field_id, field) in iter {
// We serialize the attribute values as JSON strings.
json_buffer.clear();
serde_json::to_writer(&mut json_buffer, &field)?;
serde_json::to_writer(&mut json_buffer, &field).map_err(InternalError::SerdeJson)?;
writer.insert(*field_id, &json_buffer)?;
}
@ -330,10 +343,6 @@ impl Transform<'_, '_> {
// Now that we have a valid sorter that contains the user id and the obkv we
// give it to the last transforming function which returns the TransformOutput.
let primary_key_name = fields_ids_map
.name(primary_key_id)
.map(String::from)
.expect("Primary key must be present in fields id map");
self.output_from_sorter(
sorter,
primary_key_name,
@ -347,17 +356,18 @@ impl Transform<'_, '_> {
/// Generate the `TransformOutput` based on the given sorter that can be generated from any
/// format like CSV, JSON or JSON stream. This sorter must contain a key that is the document
/// id for the user side and the value must be an obkv where keys are valid fields ids.
fn output_from_sorter<F>(
fn output_from_sorter<F, E>(
self,
sorter: grenad::Sorter<MergeFn>,
sorter: grenad::Sorter<MergeFn<E>>,
primary_key: String,
fields_ids_map: FieldsIdsMap,
approximate_number_of_documents: usize,
mut external_documents_ids: ExternalDocumentsIds<'_>,
progress_callback: F,
) -> anyhow::Result<TransformOutput>
) -> Result<TransformOutput>
where
F: Fn(UpdateIndexingStep) + Sync,
Error: From<E>,
{
let documents_ids = self.index.documents_ids(self.rtxn)?;
let mut fields_distribution = self.index.fields_distribution(self.rtxn)?;
@ -365,7 +375,7 @@ impl Transform<'_, '_> {
// Once we have sort and deduplicated the documents we write them into a final file.
let mut final_sorter = create_sorter(
|_docid, _obkvs| Err(anyhow!("cannot merge two documents")),
|_id, _obkvs| Err(InternalError::IndexingMergingKeys { process: "documents" }),
self.chunk_compression_type,
self.chunk_compression_level,
self.chunk_fusing_shrink_size,
@ -401,7 +411,10 @@ impl Transform<'_, '_> {
IndexDocumentsMethod::UpdateDocuments => {
let key = BEU32::new(docid);
let base_obkv = self.index.documents.get(&self.rtxn, &key)?
.context("document not found")?;
.ok_or(InternalError::DatabaseMissingEntry {
db_name: db_name::DOCUMENTS,
key: None,
})?;
let update_obkv = obkv::KvReader::new(update_obkv);
merge_two_obkvs(base_obkv, update_obkv, &mut obkv_buffer);
(docid, obkv_buffer.as_slice())
@ -412,7 +425,7 @@ impl Transform<'_, '_> {
// If this user id is new we add it to the external documents ids map
// for new ids and into the list of new documents.
let new_docid = available_documents_ids.next()
.context("no more available documents ids")?;
.ok_or(UserError::DocumentLimitReached)?;
new_external_documents_ids_builder.insert(external_id, new_docid as u64)?;
new_documents_ids.insert(new_docid);
(new_docid, update_obkv)
@ -472,7 +485,7 @@ impl Transform<'_, '_> {
primary_key: String,
old_fields_ids_map: FieldsIdsMap,
new_fields_ids_map: FieldsIdsMap,
) -> anyhow::Result<TransformOutput>
) -> Result<TransformOutput>
{
let fields_distribution = self.index.fields_distribution(self.rtxn)?;
let external_documents_ids = self.index.external_documents_ids(self.rtxn)?;
@ -532,10 +545,10 @@ fn compute_primary_key_pair(
fields_ids_map: &mut FieldsIdsMap,
alternative_name: Option<String>,
autogenerate_docids: bool,
) -> anyhow::Result<(FieldId, String)> {
) -> Result<(FieldId, String)> {
match primary_key {
Some(primary_key) => {
let id = fields_ids_map.insert(primary_key).ok_or(anyhow!("Maximum number of fields exceeded"))?;
let id = fields_ids_map.insert(primary_key).ok_or(UserError::AttributeLimitReached)?;
Ok((id, primary_key.to_string()))
}
None => {
@ -545,35 +558,17 @@ fn compute_primary_key_pair(
if !autogenerate_docids {
// If there is no primary key in the current document batch, we must
// return an error and not automatically generate any document id.
anyhow::bail!("missing primary key")
return Err(UserError::MissingPrimaryKey.into());
}
DEFAULT_PRIMARY_KEY_NAME.to_string()
},
};
let id = fields_ids_map.insert(&name).context("field id limit reached")?;
let id = fields_ids_map.insert(&name).ok_or(UserError::AttributeLimitReached)?;
Ok((id, name))
},
}
}
/// Only the last value associated with an id is kept.
fn keep_latest_obkv(_key: &[u8], obkvs: &[Cow<[u8]>]) -> anyhow::Result<Vec<u8>> {
obkvs.last().context("no last value").map(|last| last.clone().into_owned())
}
/// Merge all the obks in the order we see them.
fn merge_obkvs(_key: &[u8], obkvs: &[Cow<[u8]>]) -> anyhow::Result<Vec<u8>> {
let mut iter = obkvs.iter();
let first = iter.next().map(|b| b.clone().into_owned()).context("no first value")?;
Ok(iter.fold(first, |acc, current| {
let first = obkv::KvReader::new(&acc);
let second = obkv::KvReader::new(current);
let mut buffer = Vec::new();
merge_two_obkvs(first, second, &mut buffer);
buffer
}))
}
fn validate_document_id(document_id: &str) -> Option<&str> {
let document_id = document_id.trim();
Some(document_id).filter(|id| {

132
milli/src/update/settings.rs
View File

@ -1,6 +1,6 @@
use std::collections::{BTreeSet, HashMap, HashSet};
use std::result::Result as StdResult;
use anyhow::Context;
use chrono::Utc;
use grenad::CompressionType;
use itertools::Itertools;
@ -9,9 +9,10 @@ use rayon::ThreadPool;
use serde::{Deserialize, Deserializer, Serialize, Serializer};
use crate::criterion::Criterion;
use crate::error::UserError;
use crate::update::index_documents::{IndexDocumentsMethod, Transform};
use crate::update::{ClearDocuments, IndexDocuments, UpdateIndexingStep};
use crate::{FieldsIdsMap, Index};
use crate::{FieldsIdsMap, Index, Result};
#[derive(Debug, Clone, PartialEq)]
pub enum Setting<T> {
@ -33,7 +34,7 @@ impl<T> Setting<T> {
}
impl<T: Serialize> Serialize for Setting<T> {
fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error> where S: Serializer {
fn serialize<S>(&self, serializer: S) -> StdResult<S::Ok, S::Error> where S: Serializer {
match self {
Self::Set(value) => Some(value),
// Usually not_set isn't serialized by setting skip_serializing_if field attribute
@ -43,7 +44,7 @@ impl<T: Serialize> Serialize for Setting<T> {
}
impl<'de, T: Deserialize<'de>> Deserialize<'de> for Setting<T> {
fn deserialize<D>(deserializer: D) -> Result<Self, D::Error> where D: Deserializer<'de> {
fn deserialize<D>(deserializer: D) -> StdResult<Self, D::Error> where D: Deserializer<'de> {
Deserialize::deserialize(deserializer).map(|x| match x {
Some(x) => Self::Set(x),
None => Self::Reset, // Reset is forced by sending null value
@ -71,6 +72,7 @@ pub struct Settings<'a, 't, 'u, 'i> {
stop_words: Setting<BTreeSet<String>>,
distinct_field: Setting<String>,
synonyms: Setting<HashMap<String, Vec<String>>>,
primary_key: Setting<String>,
}
impl<'a, 't, 'u, 'i> Settings<'a, 't, 'u, 'i> {
@ -97,6 +99,7 @@ impl<'a, 't, 'u, 'i> Settings<'a, 't, 'u, 'i> {
stop_words: Setting::NotSet,
distinct_field: Setting::NotSet,
synonyms: Setting::NotSet,
primary_key: Setting::NotSet,
update_id,
}
}
@ -165,7 +168,15 @@ impl<'a, 't, 'u, 'i> Settings<'a, 't, 'u, 'i> {
}
}
fn reindex<F>(&mut self, cb: &F, old_fields_ids_map: FieldsIdsMap) -> anyhow::Result<()>
pub fn reset_primary_key(&mut self) {
self.primary_key = Setting::Reset;
}
pub fn set_primary_key(&mut self, primary_key: String) {
self.primary_key = Setting::Set(primary_key);
}
fn reindex<F>(&mut self, cb: &F, old_fields_ids_map: FieldsIdsMap) -> Result<()>
where
F: Fn(UpdateIndexingStep, u64) + Sync
{
@ -192,7 +203,7 @@ impl<'a, 't, 'u, 'i> Settings<'a, 't, 'u, 'i> {
};
// There already has been a document addition, the primary key should be set by now.
let primary_key = self.index.primary_key(&self.wtxn)?.context("Index must have a primary key")?;
let primary_key = self.index.primary_key(&self.wtxn)?.ok_or(UserError::MissingPrimaryKey)?;
// We remap the documents fields based on the new `FieldsIdsMap`.
let output = transform.remap_index_documents(
@ -220,7 +231,7 @@ impl<'a, 't, 'u, 'i> Settings<'a, 't, 'u, 'i> {
Ok(())
}
fn update_displayed(&mut self) -> anyhow::Result<bool> {
fn update_displayed(&mut self) -> Result<bool> {
match self.displayed_fields {
Setting::Set(ref fields) => {
let mut fields_ids_map = self.index.fields_ids_map(self.wtxn)?;
@ -234,7 +245,7 @@ impl<'a, 't, 'u, 'i> Settings<'a, 't, 'u, 'i> {
for name in names.iter() {
fields_ids_map
.insert(name)
.context("field id limit exceeded")?;
.ok_or(UserError::AttributeLimitReached)?;
}
self.index.put_displayed_fields(self.wtxn, &names)?;
self.index.put_fields_ids_map(self.wtxn, &fields_ids_map)?;
@ -245,13 +256,13 @@ impl<'a, 't, 'u, 'i> Settings<'a, 't, 'u, 'i> {
Ok(true)
}
fn update_distinct_field(&mut self) -> anyhow::Result<bool> {
fn update_distinct_field(&mut self) -> Result<bool> {
match self.distinct_field {
Setting::Set(ref attr) => {
let mut fields_ids_map = self.index.fields_ids_map(self.wtxn)?;
fields_ids_map
.insert(attr)
.context("field id limit exceeded")?;
.ok_or(UserError::AttributeLimitReached)?;
self.index.put_distinct_field(self.wtxn, &attr)?;
self.index.put_fields_ids_map(self.wtxn, &fields_ids_map)?;
@ -264,7 +275,7 @@ impl<'a, 't, 'u, 'i> Settings<'a, 't, 'u, 'i> {
/// Updates the index's searchable attributes. This causes the field map to be recomputed to
/// reflect the order of the searchable attributes.
fn update_searchable(&mut self) -> anyhow::Result<bool> {
fn update_searchable(&mut self) -> Result<bool> {
match self.searchable_fields {
Setting::Set(ref fields) => {
// every time the searchable attributes are updated, we need to update the
@ -285,13 +296,13 @@ impl<'a, 't, 'u, 'i> Settings<'a, 't, 'u, 'i> {
for name in names.iter() {
new_fields_ids_map
.insert(&name)
.context("field id limit exceeded")?;
.ok_or(UserError::AttributeLimitReached)?;
}
for (_, name) in old_fields_ids_map.iter() {
new_fields_ids_map
.insert(&name)
.context("field id limit exceeded")?;
.ok_or(UserError::AttributeLimitReached)?;
}
self.index.put_searchable_fields(self.wtxn, &names)?;
@ -303,7 +314,7 @@ impl<'a, 't, 'u, 'i> Settings<'a, 't, 'u, 'i> {
Ok(true)
}
fn update_stop_words(&mut self) -> anyhow::Result<bool> {
fn update_stop_words(&mut self) -> Result<bool> {
match self.stop_words {
Setting::Set(ref stop_words) => {
let current = self.index.stop_words(self.wtxn)?;
@ -325,7 +336,7 @@ impl<'a, 't, 'u, 'i> Settings<'a, 't, 'u, 'i> {
}
}
fn update_synonyms(&mut self) -> anyhow::Result<bool> {
fn update_synonyms(&mut self) -> Result<bool> {
match self.synonyms {
Setting::Set(ref synonyms) => {
fn normalize(analyzer: &Analyzer<&[u8]>, text: &str) -> Vec<String> {
@ -383,13 +394,13 @@ impl<'a, 't, 'u, 'i> Settings<'a, 't, 'u, 'i> {
}
}
fn update_filterable(&mut self) -> anyhow::Result<()> {
fn update_filterable(&mut self) -> Result<()> {
match self.filterable_fields {
Setting::Set(ref fields) => {
let mut fields_ids_map = self.index.fields_ids_map(self.wtxn)?;
let mut new_facets = HashSet::new();
for name in fields {
fields_ids_map.insert(name).context("field id limit exceeded")?;
fields_ids_map.insert(name).ok_or(UserError::AttributeLimitReached)?;
new_facets.insert(name.clone());
}
self.index.put_filterable_fields(self.wtxn, &new_facets)?;
@ -401,7 +412,7 @@ impl<'a, 't, 'u, 'i> Settings<'a, 't, 'u, 'i> {
Ok(())
}
fn update_criteria(&mut self) -> anyhow::Result<()> {
fn update_criteria(&mut self) -> Result<()> {
match self.criteria {
Setting::Set(ref fields) => {
let mut fields_ids_map = self.index.fields_ids_map(self.wtxn)?;
@ -409,7 +420,7 @@ impl<'a, 't, 'u, 'i> Settings<'a, 't, 'u, 'i> {
for name in fields {
let criterion: Criterion = name.parse()?;
if let Some(name) = criterion.field_name() {
fields_ids_map.insert(name).context("field id limit exceeded")?;
fields_ids_map.insert(name).ok_or(UserError::AttributeLimitReached)?;
}
new_criteria.push(criterion);
}
@ -422,7 +433,32 @@ impl<'a, 't, 'u, 'i> Settings<'a, 't, 'u, 'i> {
Ok(())
}
pub fn execute<F>(mut self, progress_callback: F) -> anyhow::Result<()>
fn update_primary_key(&mut self) -> Result<()> {
match self.primary_key {
Setting::Set(ref primary_key) => {
if self.index.number_of_documents(&self.wtxn)? == 0 {
let mut fields_ids_map = self.index.fields_ids_map(self.wtxn)?;
fields_ids_map.insert(primary_key).ok_or(UserError::AttributeLimitReached)?;
self.index.put_fields_ids_map(self.wtxn, &fields_ids_map)?;
self.index.put_primary_key(self.wtxn, primary_key)?;
Ok(())
} else {
Err(UserError::PrimaryKeyCannotBeChanged.into())
}
},
Setting::Reset => {
if self.index.number_of_documents(&self.wtxn)? == 0 {
self.index.delete_primary_key(self.wtxn)?;
Ok(())
} else {
Err(UserError::PrimaryKeyCannotBeReset.into())
}
},
Setting::NotSet => Ok(()),
}
}
pub fn execute<F>(mut self, progress_callback: F) -> Result<()>
where
F: Fn(UpdateIndexingStep, u64) + Sync
{
@ -435,6 +471,7 @@ impl<'a, 't, 'u, 'i> Settings<'a, 't, 'u, 'i> {
self.update_filterable()?;
self.update_distinct_field()?;
self.update_criteria()?;
self.update_primary_key()?;
// If there is new faceted fields we indicate that we must reindex as we must
// index new fields as facets. It means that the distinct attribute,
@ -461,8 +498,9 @@ mod tests {
use maplit::{btreeset, hashmap, hashset};
use big_s::S;
use crate::{Criterion, FilterCondition, SearchResult};
use crate::error::Error;
use crate::update::{IndexDocuments, UpdateFormat};
use crate::{Criterion, FilterCondition, SearchResult};
use super::*;
@ -976,4 +1014,56 @@ mod tests {
let rtxn = index.read_txn().unwrap();
FilterCondition::from_str(&rtxn, &index, "toto = 32").unwrap_err();
}
#[test]
fn setting_primary_key() {
let path = tempfile::tempdir().unwrap();
let mut options = EnvOpenOptions::new();
options.map_size(10 * 1024 * 1024); // 10 MB
let index = Index::new(options, &path).unwrap();
// Set the primary key settings
let mut wtxn = index.write_txn().unwrap();
let mut builder = Settings::new(&mut wtxn, &index, 0);
builder.set_primary_key(S("mykey"));
builder.execute(|_, _| ()).unwrap();
assert_eq!(index.primary_key(&wtxn).unwrap(), Some("mykey"));
// Then index some documents with the "mykey" primary key.
let content = &br#"[
{ "mykey": 1, "name": "kevin", "age": 23 },
{ "mykey": 2, "name": "kevina", "age": 21 },
{ "mykey": 3, "name": "benoit", "age": 34 },
{ "mykey": 4, "name": "bernard", "age": 34 },
{ "mykey": 5, "name": "bertrand", "age": 34 },
{ "mykey": 6, "name": "bernie", "age": 34 },
{ "mykey": 7, "name": "ben", "age": 34 }
]"#[..];
let mut builder = IndexDocuments::new(&mut wtxn, &index, 1);
builder.update_format(UpdateFormat::Json);
builder.disable_autogenerate_docids();
builder.execute(content, |_, _| ()).unwrap();
wtxn.commit().unwrap();
// We now try to reset the primary key
let mut wtxn = index.write_txn().unwrap();
let mut builder = Settings::new(&mut wtxn, &index, 0);
builder.reset_primary_key();
let err = builder.execute(|_, _| ()).unwrap_err();
assert!(matches!(err, Error::UserError(UserError::PrimaryKeyCannotBeReset)));
wtxn.abort().unwrap();
// But if we clear the database...
let mut wtxn = index.write_txn().unwrap();
let builder = ClearDocuments::new(&mut wtxn, &index, 0);
builder.execute().unwrap();
// ...we can change the primary key
let mut builder = Settings::new(&mut wtxn, &index, 0);
builder.set_primary_key(S("myid"));
builder.execute(|_, _| ()).unwrap();
wtxn.commit().unwrap();
}
}

4
milli/src/update/update_builder.rs
View File

@ -1,7 +1,7 @@
use grenad::CompressionType;
use rayon::ThreadPool;
use crate::Index;
use crate::{Index, Result};
use super::{ClearDocuments, DeleteDocuments, IndexDocuments, Settings, Facets};
pub struct UpdateBuilder<'a> {
@ -76,7 +76,7 @@ impl<'a> UpdateBuilder<'a> {
self,
wtxn: &'t mut heed::RwTxn<'i, 'u>,
index: &'i Index,
) -> anyhow::Result<DeleteDocuments<'t, 'u, 'i>>
) -> Result<DeleteDocuments<'t, 'u, 'i>>
{
DeleteDocuments::new(wtxn, index, self.update_id)
}

11
milli/src/update/word_prefix_docids.rs
View File

@ -5,8 +5,11 @@ use fst::Streamer;
use grenad::CompressionType;
use heed::types::ByteSlice;
use crate::Result;
use crate::update::index_documents::WriteMethod;
use crate::update::index_documents::{create_sorter, word_docids_merge, sorter_into_lmdb_database};
use crate::update::index_documents::{
create_sorter, roaring_bitmap_merge, sorter_into_lmdb_database,
};
pub struct WordPrefixDocids<'t, 'u, 'i> {
wtxn: &'t mut heed::RwTxn<'i, 'u>,
@ -31,7 +34,7 @@ impl<'t, 'u, 'i> WordPrefixDocids<'t, 'u, 'i> {
}
}
pub fn execute(self) -> anyhow::Result<()> {
pub fn execute(self) -> Result<()> {
// Clear the word prefix docids database.
self.index.word_prefix_docids.clear(self.wtxn)?;
@ -40,7 +43,7 @@ impl<'t, 'u, 'i> WordPrefixDocids<'t, 'u, 'i> {
// It is forbidden to keep a mutable reference into the database
// and write into it at the same time, therefore we write into another file.
let mut prefix_docids_sorter = create_sorter(
word_docids_merge,
roaring_bitmap_merge,
self.chunk_compression_type,
self.chunk_compression_level,
self.chunk_fusing_shrink_size,
@ -66,7 +69,7 @@ impl<'t, 'u, 'i> WordPrefixDocids<'t, 'u, 'i> {
self.wtxn,
*self.index.word_prefix_docids.as_polymorph(),
prefix_docids_sorter,
word_docids_merge,
roaring_bitmap_merge,
WriteMethod::Append,
)?;

10
milli/src/update/word_prefix_pair_proximity_docids.rs
View File

@ -7,11 +7,11 @@ use heed::BytesEncode;
use heed::types::ByteSlice;
use log::debug;
use crate::Index;
use crate::{Index, Result};
use crate::heed_codec::StrStrU8Codec;
use crate::update::index_documents::{
WriteMethod, create_sorter, sorter_into_lmdb_database,
words_pairs_proximities_docids_merge,
cbo_roaring_bitmap_merge,
};
pub struct WordPrefixPairProximityDocids<'t, 'u, 'i> {
@ -41,7 +41,7 @@ impl<'t, 'u, 'i> WordPrefixPairProximityDocids<'t, 'u, 'i> {
}
}
pub fn execute(self) -> anyhow::Result<()> {
pub fn execute(self) -> Result<()> {
debug!("Computing and writing the word prefix pair proximity docids into LMDB on disk...");
self.index.word_prefix_pair_proximity_docids.clear(self.wtxn)?;
@ -50,7 +50,7 @@ impl<'t, 'u, 'i> WordPrefixPairProximityDocids<'t, 'u, 'i> {
// Here we create a sorter akin to the previous one.
let mut word_prefix_pair_proximity_docids_sorter = create_sorter(
words_pairs_proximities_docids_merge,
cbo_roaring_bitmap_merge,
self.chunk_compression_type,
self.chunk_compression_level,
self.chunk_fusing_shrink_size,
@ -80,7 +80,7 @@ impl<'t, 'u, 'i> WordPrefixPairProximityDocids<'t, 'u, 'i> {
self.wtxn,
*self.index.word_prefix_pair_proximity_docids.as_polymorph(),
word_prefix_pair_proximity_docids_sorter,
words_pairs_proximities_docids_merge,
cbo_roaring_bitmap_merge,
WriteMethod::Append,
)?;

18
milli/src/update/words_level_positions.rs
View File

@ -11,11 +11,13 @@ use heed::{BytesEncode, Error};
use log::debug;
use roaring::RoaringBitmap;
use crate::error::InternalError;
use crate::heed_codec::{StrLevelPositionCodec, CboRoaringBitmapCodec};
use crate::Result;
use crate::update::index_documents::WriteMethod;
use crate::update::index_documents::{
create_writer, create_sorter, writer_into_reader, write_into_lmdb_database,
word_prefix_level_positions_docids_merge, sorter_into_lmdb_database
cbo_roaring_bitmap_merge, sorter_into_lmdb_database
};
use crate::{Index, TreeLevel};
@ -56,7 +58,7 @@ impl<'t, 'u, 'i> WordsLevelPositions<'t, 'u, 'i> {
self
}
pub fn execute(self) -> anyhow::Result<()> {
pub fn execute(self) -> Result<()> {
debug!("Computing and writing the word levels positions docids into LMDB on disk...");
let entries = compute_positions_levels(
@ -78,7 +80,7 @@ impl<'t, 'u, 'i> WordsLevelPositions<'t, 'u, 'i> {
self.wtxn,
*self.index.word_level_position_docids.as_polymorph(),
entries,
|_, _| anyhow::bail!("invalid word level position merging"),
|_, _| Err(InternalError::IndexingMergingKeys { process: "word level position" }),
WriteMethod::Append,
)?;
@ -86,7 +88,7 @@ impl<'t, 'u, 'i> WordsLevelPositions<'t, 'u, 'i> {
self.index.word_prefix_level_position_docids.clear(self.wtxn)?;
let mut word_prefix_level_positions_docids_sorter = create_sorter(
word_prefix_level_positions_docids_merge,
cbo_roaring_bitmap_merge,
self.chunk_compression_type,
self.chunk_compression_level,
self.chunk_fusing_shrink_size,
@ -119,7 +121,7 @@ impl<'t, 'u, 'i> WordsLevelPositions<'t, 'u, 'i> {
self.wtxn,
*self.index.word_prefix_level_position_docids.as_polymorph(),
word_prefix_level_positions_docids_sorter,
word_prefix_level_positions_docids_merge,
cbo_roaring_bitmap_merge,
WriteMethod::Append,
)?;
@ -142,7 +144,7 @@ impl<'t, 'u, 'i> WordsLevelPositions<'t, 'u, 'i> {
self.wtxn,
*self.index.word_prefix_level_position_docids.as_polymorph(),
entries,
|_, _| anyhow::bail!("invalid word prefix level position merging"),
|_, _| Err(InternalError::IndexingMergingKeys { process: "word prefix level position" }),
WriteMethod::Append,
)?;
@ -174,7 +176,7 @@ fn compute_positions_levels(
shrink_size: Option<u64>,
level_group_size: NonZeroU32,
min_level_size: NonZeroU32,
) -> anyhow::Result<Reader<FileFuse>>
) -> Result<Reader<FileFuse>>
{
// It is forbidden to keep a cursor and write in a database at the same time with LMDB
// therefore we write the facet levels entries into a grenad file before transfering them.
@ -251,7 +253,7 @@ fn write_level_entry(
left: u32,
right: u32,
ids: &RoaringBitmap,
) -> anyhow::Result<()>
) -> Result<()>
{
let key = (word, level, left, right);
let key = StrLevelPositionCodec::bytes_encode(&key).ok_or(Error::Encoding)?;

4
milli/src/update/words_prefixes_fst.rs
View File

@ -2,7 +2,7 @@ use std::iter::FromIterator;
use std::str;
use fst::Streamer;
use crate::{Index, SmallString32};
use crate::{Index, SmallString32, Result};
pub struct WordsPrefixesFst<'t, 'u, 'i> {
wtxn: &'t mut heed::RwTxn<'i, 'u>,
@ -48,7 +48,7 @@ impl<'t, 'u, 'i> WordsPrefixesFst<'t, 'u, 'i> {
self
}
pub fn execute(self) -> anyhow::Result<()> {
pub fn execute(self) -> Result<()> {
let words_fst = self.index.words_fst(&self.wtxn)?;
let number_of_words = words_fst.len();
let min_number_of_words = (number_of_words as f64 * self.threshold) as usize;

7
milli/tests/search/query_criteria.rs
View File

@ -1,5 +1,6 @@
use milli::{Search, SearchResult, Criterion};
use big_s::S;
use milli::update::Settings;
use milli::{Search, SearchResult, Criterion};
use crate::search::{self, EXTERNAL_DOCUMENTS_IDS};
use Criterion::*;
@ -189,7 +190,9 @@ fn criteria_mixup() {
eprintln!("Testing with criteria order: {:?}", &criteria);
//update criteria
let mut wtxn = index.write_txn().unwrap();
index.put_criteria(&mut wtxn, &criteria).unwrap();
let mut builder = Settings::new(&mut wtxn, &index, 0);
builder.set_criteria(criteria.iter().map(ToString::to_string).collect());
builder.execute(|_, _| ()).unwrap();
wtxn.commit().unwrap();
let mut rtxn = index.read_txn().unwrap();