hermes_core/
lib.rs

1//! Hermes - A minimal async search engine library
2//!
3//! Features:
4//! - Fully async IO with Directory abstraction for network/local/memory storage
5//! - SSTable-based term dictionary with hot cache and lazy loading
6//! - Bitpacked posting lists with block-level skip info
7//! - Document store with Zstd compression
8//! - Multiple segments with merge support
9//! - Text and numeric field support
10//! - Term, boolean, and boost queries
11//! - MaxScore / block-max pruning query optimizations
12
13pub mod compression;
14pub mod directories;
15pub mod dsl;
16pub mod error;
17pub mod index;
18pub mod merge;
19pub mod query;
20pub mod segment;
21pub mod structures;
22pub mod tokenizer;
23
24// Re-exports from dsl
25pub use dsl::{
26    Document, Field, FieldDef, FieldEntry, FieldType, FieldValue, IndexDef, QueryLanguageParser,
27    Schema, SchemaBuilder, SdlParser, parse_sdl, parse_single_index,
28};
29
30// Re-exports from structures
31pub use structures::{
32    AsyncSSTableReader, BlockPostingList, HorizontalBP128Iterator, HorizontalBP128PostingList,
33    PostingList, PostingListIterator, SSTableValue, TERMINATED, TermInfo,
34};
35
36// Re-exports from directories
37#[cfg(feature = "native")]
38pub use directories::FsDirectory;
39#[cfg(feature = "http")]
40pub use directories::HttpDirectory;
41#[cfg(feature = "native")]
42pub use directories::MmapDirectory;
43pub use directories::{
44    CachingDirectory, Directory, DirectoryWriter, FileHandle, OwnedBytes, RamDirectory,
45    SliceCacheStats, SliceCachingDirectory,
46};
47
48/// Default directory type for native builds - uses memory-mapped files for efficient access
49#[cfg(feature = "native")]
50pub type DefaultDirectory = MmapDirectory;
51
52// Re-exports from segment
53pub use segment::{
54    AsyncSegmentReader, AsyncStoreReader, FieldStats, SegmentId, SegmentMeta, SegmentReader,
55};
56#[cfg(feature = "native")]
57pub use segment::{SegmentBuilder, SegmentBuilderConfig, SegmentBuilderStats};
58
59// Re-exports from query
60pub use query::{
61    Bm25Params, BooleanQuery, BoostQuery, MaxScoreExecutor, Query, ScoredDoc, Scorer, SearchHit,
62    SearchResponse, SearchResult, SparseMaxScoreExecutor, TermQuery, TopKCollector, search_segment,
63};
64
65// Re-exports from tokenizer
66pub use tokenizer::{
67    BoxedTokenizer, Language, LanguageAwareTokenizer, LowercaseTokenizer, MultiLanguageStemmer,
68    SimpleTokenizer, StemmerTokenizer, Token, Tokenizer, TokenizerRegistry, parse_language,
69};
70
71// Re-exports from other modules
72pub use directories::SLICE_CACHE_EXTENSION;
73pub use error::{Error, Result};
74pub use index::Searcher;
75#[cfg(feature = "native")]
76pub use index::{Index, IndexReader, IndexWriter};
77pub use index::{IndexConfig, IndexMetadata, SLICE_CACHE_FILENAME};
78#[cfg(feature = "native")]
79pub use index::{
80    IndexingStats, SchemaConfig, SchemaFieldConfig, create_index_at_path, create_index_from_sdl,
81    index_documents_from_reader, index_json_document, parse_schema,
82};
83
84// Re-exports from merge
85#[cfg(feature = "native")]
86pub use merge::SegmentManager;
87pub use merge::{MergeCandidate, MergePolicy, NoMergePolicy, SegmentInfo, TieredMergePolicy};
88
89pub type DocId = u32;
90pub type TermFreq = u32;
91pub type Score = f32;
92
93/// Default number of indexing threads (cpu / 4, minimum 1).
94/// Centralized so all configs share one definition.
95#[cfg(feature = "native")]
96pub fn default_indexing_threads() -> usize {
97    (num_cpus::get() / 4).max(1)
98}
99
100/// Default number of compression threads (cpu / 4, minimum 1).
101/// Centralized so all configs share one definition.
102#[cfg(feature = "native")]
103pub fn default_compression_threads() -> usize {
104    (num_cpus::get() / 4).max(1)
105}
hermes_core/lib.rs

hermes_core/
lib.rs
