2020-08-12 16:43:02 +08:00
|
|
|
mod criterion;
|
2020-06-22 23:56:07 +08:00
|
|
|
mod heed_codec;
|
2020-08-15 22:16:00 +08:00
|
|
|
mod node;
|
2020-06-05 15:48:46 +08:00
|
|
|
mod query_tokens;
|
2020-08-13 20:15:05 +08:00
|
|
|
mod search;
|
2020-08-10 19:47:19 +08:00
|
|
|
mod transitive_arc;
|
2020-08-16 02:37:13 +08:00
|
|
|
pub mod lexer;
|
2020-06-05 02:25:51 +08:00
|
|
|
|
2020-08-13 20:15:05 +08:00
|
|
|
use std::collections::HashMap;
|
2020-08-10 19:47:19 +08:00
|
|
|
use std::fs::{File, OpenOptions};
|
2020-05-31 22:09:34 +08:00
|
|
|
use std::hash::BuildHasherDefault;
|
2020-08-10 19:47:19 +08:00
|
|
|
use std::path::{Path, PathBuf};
|
|
|
|
use std::sync::Arc;
|
2020-05-31 22:09:34 +08:00
|
|
|
|
2020-08-10 19:47:19 +08:00
|
|
|
use anyhow::Context;
|
2020-06-30 04:25:59 +08:00
|
|
|
use fxhash::{FxHasher32, FxHasher64};
|
2020-05-31 23:48:13 +08:00
|
|
|
use heed::types::*;
|
|
|
|
use heed::{PolyDatabase, Database};
|
2020-08-10 19:47:19 +08:00
|
|
|
use memmap::Mmap;
|
|
|
|
use oxidized_mtbl as omtbl;
|
2020-06-05 02:25:51 +08:00
|
|
|
|
2020-08-13 20:15:05 +08:00
|
|
|
pub use self::search::{Search, SearchResult};
|
|
|
|
pub use self::criterion::{Criterion, default_criteria};
|
2020-06-22 23:56:07 +08:00
|
|
|
use self::heed_codec::RoaringBitmapCodec;
|
2020-08-10 19:47:19 +08:00
|
|
|
use self::transitive_arc::TransitiveArc;
|
2020-05-31 22:09:34 +08:00
|
|
|
|
|
|
|
pub type FastMap4<K, V> = HashMap<K, V, BuildHasherDefault<FxHasher32>>;
|
2020-06-30 04:25:59 +08:00
|
|
|
pub type FastMap8<K, V> = HashMap<K, V, BuildHasherDefault<FxHasher64>>;
|
2020-05-31 22:09:34 +08:00
|
|
|
pub type SmallString32 = smallstr::SmallString<[u8; 32]>;
|
2020-06-11 17:55:03 +08:00
|
|
|
pub type SmallVec32<T> = smallvec::SmallVec<[T; 32]>;
|
|
|
|
pub type SmallVec16<T> = smallvec::SmallVec<[T; 16]>;
|
2020-05-31 22:09:34 +08:00
|
|
|
pub type BEU32 = heed::zerocopy::U32<heed::byteorder::BE>;
|
|
|
|
pub type DocumentId = u32;
|
2020-08-06 17:08:24 +08:00
|
|
|
pub type Attribute = u32;
|
2020-07-07 18:21:22 +08:00
|
|
|
pub type Position = u32;
|
2020-05-31 22:09:34 +08:00
|
|
|
|
2020-05-31 23:48:13 +08:00
|
|
|
#[derive(Clone)]
|
|
|
|
pub struct Index {
|
2020-08-10 19:47:19 +08:00
|
|
|
// The database path, where the LMDB and MTBL files are.
|
|
|
|
path: PathBuf,
|
2020-06-23 00:02:22 +08:00
|
|
|
/// Contains many different types (e.g. the documents CSV headers).
|
2020-05-31 23:48:13 +08:00
|
|
|
pub main: PolyDatabase,
|
2020-06-23 00:02:22 +08:00
|
|
|
/// A word and all the positions where it appears in the whole dataset.
|
|
|
|
pub word_positions: Database<Str, RoaringBitmapCodec>,
|
|
|
|
pub prefix_word_positions: Database<Str, RoaringBitmapCodec>,
|
|
|
|
/// Maps a word at a position (u32) and all the documents ids where it appears.
|
|
|
|
pub word_position_docids: Database<ByteSlice, RoaringBitmapCodec>,
|
|
|
|
pub prefix_word_position_docids: Database<ByteSlice, RoaringBitmapCodec>,
|
2020-06-23 01:04:10 +08:00
|
|
|
/// Maps a word and an attribute (u32) to all the documents ids that it appears in.
|
|
|
|
pub word_attribute_docids: Database<ByteSlice, RoaringBitmapCodec>,
|
2020-08-10 19:47:19 +08:00
|
|
|
/// The MTBL store that contains the documents content.
|
|
|
|
documents: omtbl::Reader<TransitiveArc<Mmap>>,
|
2020-05-31 23:48:13 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
impl Index {
|
2020-08-10 19:47:19 +08:00
|
|
|
pub fn new<P: AsRef<Path>>(env: &heed::Env, path: P) -> anyhow::Result<Index> {
|
|
|
|
let documents_path = path.as_ref().join("documents.mtbl");
|
|
|
|
let mut documents = OpenOptions::new().create(true).write(true).read(true).open(documents_path)?;
|
|
|
|
// If the file is empty we must initialize it like an empty MTBL database.
|
|
|
|
if documents.metadata()?.len() == 0 {
|
|
|
|
omtbl::Writer::new(&mut documents).finish()?;
|
|
|
|
}
|
|
|
|
let documents = unsafe { memmap::Mmap::map(&documents)? };
|
|
|
|
|
2020-06-23 00:02:22 +08:00
|
|
|
Ok(Index {
|
2020-08-10 19:47:19 +08:00
|
|
|
path: path.as_ref().to_path_buf(),
|
2020-06-23 00:02:22 +08:00
|
|
|
main: env.create_poly_database(None)?,
|
|
|
|
word_positions: env.create_database(Some("word-positions"))?,
|
|
|
|
prefix_word_positions: env.create_database(Some("prefix-word-positions"))?,
|
|
|
|
word_position_docids: env.create_database(Some("word-position-docids"))?,
|
|
|
|
prefix_word_position_docids: env.create_database(Some("prefix-word-position-docids"))?,
|
2020-06-23 01:04:10 +08:00
|
|
|
word_attribute_docids: env.create_database(Some("word-attribute-docids"))?,
|
2020-08-10 19:47:19 +08:00
|
|
|
documents: omtbl::Reader::new(TransitiveArc(Arc::new(documents)))?,
|
2020-06-23 00:02:22 +08:00
|
|
|
})
|
2020-05-31 23:48:13 +08:00
|
|
|
}
|
|
|
|
|
2020-08-10 19:47:19 +08:00
|
|
|
pub fn refresh_documents(&mut self) -> anyhow::Result<()> {
|
|
|
|
let documents_path = self.path.join("documents.mtbl");
|
|
|
|
let documents = File::open(&documents_path)?;
|
|
|
|
let documents = unsafe { memmap::Mmap::map(&documents)? };
|
|
|
|
self.documents = omtbl::Reader::new(TransitiveArc(Arc::new(documents)))?;
|
|
|
|
Ok(())
|
|
|
|
}
|
|
|
|
|
2020-06-29 19:54:47 +08:00
|
|
|
pub fn put_headers(&self, wtxn: &mut heed::RwTxn, headers: &[u8]) -> anyhow::Result<()> {
|
|
|
|
Ok(self.main.put::<_, Str, ByteSlice>(wtxn, "headers", headers)?)
|
|
|
|
}
|
|
|
|
|
2020-05-31 23:48:13 +08:00
|
|
|
pub fn headers<'t>(&self, rtxn: &'t heed::RoTxn) -> heed::Result<Option<&'t [u8]>> {
|
|
|
|
self.main.get::<_, Str, ByteSlice>(rtxn, "headers")
|
|
|
|
}
|
|
|
|
|
2020-07-13 23:50:16 +08:00
|
|
|
pub fn number_of_attributes<'t>(&self, rtxn: &'t heed::RoTxn) -> anyhow::Result<Option<usize>> {
|
|
|
|
match self.headers(rtxn)? {
|
|
|
|
Some(headers) => {
|
|
|
|
let mut rdr = csv::Reader::from_reader(headers);
|
|
|
|
let headers = rdr.headers()?;
|
|
|
|
Ok(Some(headers.len()))
|
|
|
|
}
|
|
|
|
None => Ok(None),
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2020-06-29 19:54:47 +08:00
|
|
|
pub fn put_fst<A: AsRef<[u8]>>(&self, wtxn: &mut heed::RwTxn, fst: &fst::Set<A>) -> anyhow::Result<()> {
|
|
|
|
Ok(self.main.put::<_, Str, ByteSlice>(wtxn, "words-fst", fst.as_fst().as_bytes())?)
|
|
|
|
}
|
|
|
|
|
2020-06-23 01:04:10 +08:00
|
|
|
pub fn fst<'t>(&self, rtxn: &'t heed::RoTxn) -> anyhow::Result<Option<fst::Set<&'t [u8]>>> {
|
|
|
|
match self.main.get::<_, Str, ByteSlice>(rtxn, "words-fst")? {
|
|
|
|
Some(bytes) => Ok(Some(fst::Set::new(bytes)?)),
|
|
|
|
None => Ok(None),
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2020-08-10 19:47:19 +08:00
|
|
|
/// Returns a [`Vec`] of the requested documents. Returns an error if a document is missing.
|
|
|
|
pub fn documents<I: IntoIterator<Item=DocumentId>>(&self, iter: I) -> anyhow::Result<Vec<(DocumentId, Vec<u8>)>> {
|
|
|
|
iter.into_iter().map(|id| {
|
|
|
|
let key = id.to_be_bytes();
|
|
|
|
let content = self.documents.clone().get(&key)?.with_context(|| format!("Could not find document {}.", id))?;
|
|
|
|
Ok((id, content.as_ref().to_vec()))
|
|
|
|
})
|
|
|
|
.collect()
|
|
|
|
}
|
|
|
|
|
|
|
|
/// Returns the number of documents indexed in the database.
|
|
|
|
pub fn number_of_documents(&self) -> usize {
|
|
|
|
self.documents.metadata().count_entries as usize
|
|
|
|
}
|
|
|
|
|
2020-08-13 20:15:05 +08:00
|
|
|
pub fn search<'a>(&'a self, rtxn: &'a heed::RoTxn) -> Search<'a> {
|
|
|
|
Search::new(rtxn, self)
|
2020-05-31 23:48:13 +08:00
|
|
|
}
|
|
|
|
}
|