Use Tantivy for querying story contents

This commit is contained in:
Joakim Soderlund 2020-03-08 00:54:56 +01:00
parent 36ade958bb
commit 723b19b035
8 changed files with 1079 additions and 23 deletions

1
.gitignore vendored
View file

@ -1,2 +1,3 @@
/cache
/target /target
**/*.rs.bk **/*.rs.bk

857
Cargo.lock generated

File diff suppressed because it is too large Load diff

View file

@ -8,6 +8,7 @@ edition = "2024"
members = [ members = [
"cli", "cli",
"query", "query",
"search",
] ]
default-members = [ default-members = [

View file

@ -7,8 +7,8 @@ edition = "2024"
[dependencies.fimfareader] [dependencies.fimfareader]
path = ".." path = ".."
[dependencies.fimfareader-query] [dependencies.fimfareader-search]
path = "../query" path = "../search"
[dependencies.rustyline] [dependencies.rustyline]
default-features = false default-features = false

View file

@ -5,10 +5,11 @@ use std::error::Error;
use std::result::Result; use std::result::Result;
use std::time::Instant; use std::time::Instant;
use fimfareader::archive::Fetcher;
use fimfareader_query::parse;
use rustyline::DefaultEditor; use rustyline::DefaultEditor;
use fimfareader::archive::Fetcher;
use fimfareader_search::Searcher;
fn main() -> Result<(), Box<dyn Error>> { fn main() -> Result<(), Box<dyn Error>> {
let argv = args().collect::<Vec<String>>(); let argv = args().collect::<Vec<String>>();
let mut editor = DefaultEditor::new()?; let mut editor = DefaultEditor::new()?;
@ -28,33 +29,41 @@ fn main() -> Result<(), Box<dyn Error>> {
println!("Finished loading in {finish:?}."); println!("Finished loading in {finish:?}.");
println!("The archive contains {count} stories."); println!("The archive contains {count} stories.");
let searcher = Searcher::new(&fetcher);
while let Ok(line) = editor.readline(">>> ") { while let Ok(line) = editor.readline(">>> ") {
editor.add_history_entry(&line)?; editor.add_history_entry(&line)?;
let filter = match parse(&line) {
Ok(filter) => filter,
Err(error) => {
println!("{}", error);
continue;
}
};
let start = Instant::now(); let start = Instant::now();
let stories = fetcher.filter(&filter);
let result = searcher
.search(&line)
.into_iter()
.filter(|(_sid, score)| *score > 10f32)
.map(|(sid, score)| (i32::try_from(sid).unwrap(), score))
.filter_map(|(sid, score)| Some((fetcher.fetch(sid)?, score)))
.collect::<Vec<_>>();
let finish = (Instant::now() - start).as_millis(); let finish = (Instant::now() - start).as_millis();
let count = stories.len(); let count = result.len();
println!("Found {} stories in {} milliseconds!", count, finish); println!("Found {} stories in {} milliseconds!", count, finish);
if count > 32 { for (story, score) in result {
continue;
}
for story in stories.iter() {
let key = &story.id; let key = &story.id;
let title = &story.title; let title = &story.title;
println!("[{}] {}", key, title); let tags = story
.tags
.iter()
.map(|tag| tag.name.to_string())
.collect::<Vec<_>>()
.join(", ");
println!("{:02.02}% [{:>6}] {}", score, key, title);
println!("{}", tags);
println!("{}", story.short_description);
println!("{}", story.url);
println!();
} }
} }

16
search/Cargo.toml Normal file
View file

@ -0,0 +1,16 @@
[package]
name = "fimfareader-search"
version = "0.1.0"
authors = ["Joakim Soderlund <joakim.soderlund@gmail.com>"]
edition = "2024"
[dependencies.fimfareader]
path = ".."
[dependencies.tantivy]
git = "https://github.com/quickwit-oss/tantivy.git"
[dependencies.zip]
version = "=0.6.6"
features = ["deflate"]
default-features = false

160
search/src/lib.rs Normal file
View file

@ -0,0 +1,160 @@
//! Main module.
use std::fs::create_dir_all;
use std::io::Cursor;
use std::io::Read;
use std::io::Seek;
use std::io::Write;
use std::io::stdout;
use std::path::Path;
use std::time::Instant;
use tantivy::Index;
use tantivy::ReloadPolicy;
use tantivy::TantivyDocument;
use tantivy::collector::TopDocs;
use tantivy::query::QueryParser;
use tantivy::schema;
use tantivy::schema::Schema;
use tantivy::schema::Value;
use zip::read::ZipArchive;
use fimfareader::archive::Fetcher;
use fimfareader::archive::Story;
pub struct Searcher {
index: Index,
}
impl Searcher {
pub fn new<T>(fetcher: &Fetcher<T>) -> Self
where
T: Read + Seek + Send,
{
Searcher {
index: Self::load_index(fetcher),
}
}
fn schema() -> Schema {
let mut builder = Schema::builder();
builder.add_i64_field("sid", schema::INDEXED | schema::STORED);
builder.add_text_field("content", schema::TEXT);
builder.build()
}
fn load_index<T>(fetcher: &Fetcher<T>) -> Index
where
T: Read + Seek + Send,
{
let identity = fetcher.identity().unwrap();
let path = Path::new("cache").join(identity);
if path.exists() {
Index::open_in_dir(path).unwrap()
} else {
Self::make_index(&path, fetcher)
}
}
fn make_index<T>(path: &Path, fetcher: &Fetcher<T>) -> Index
where
T: Read + Seek + Send,
{
let start = Instant::now();
print!("\r\rIndexing archive...\r\r");
create_dir_all(path).unwrap();
let schema = Self::schema();
let index = Index::create_in_dir(path, schema).unwrap();
let mut writer = index.writer(1_073_741_824).unwrap();
let mut buffer = String::with_capacity(1_048_576);
let schema = index.schema();
let identifier = schema.get_field("sid").unwrap();
let content = schema.get_field("content").unwrap();
let story_count = fetcher.iter().count() as f64;
for (i, story) in fetcher.iter().enumerate() {
let progress = (i * 100) as f64 / story_count;
print!("\r\rIndexing archive... {progress:.2}%\r\r");
let cursor = Cursor::new(fetcher.read(story).unwrap());
let mut epub = ZipArchive::new(cursor).unwrap();
let mut document = TantivyDocument::default();
document.add_i64(identifier, story.id.into());
for i in 0..epub.len() {
let mut file = epub.by_index(i).unwrap();
if !file.name().ends_with(".html") {
continue;
}
file.read_to_string(&mut buffer).unwrap();
document.add_text(content, &buffer);
buffer.clear();
}
writer.add_document(document).unwrap();
}
print!("\r\rCommitting archive index...\r\r");
stdout().flush().unwrap();
writer.commit().unwrap();
writer.wait_merging_threads().unwrap();
let finish = (Instant::now() - start).as_secs();
println!("Index generated in {finish} seconds.");
index
}
pub fn search(&self, text: &str) -> Vec<(i64, f32)> {
let reader = self
.index
.reader_builder()
.reload_policy(ReloadPolicy::OnCommitWithDelay)
.try_into()
.unwrap();
let schema = self.index.schema();
let identifier = schema.get_field("sid").unwrap();
let content = schema.get_field("content").unwrap();
let parser = QueryParser::for_index(&self.index, vec![content]);
let query = parser.parse_query(text).unwrap();
let searcher = reader.searcher();
let limit = TopDocs::with_limit(32);
let docs = searcher.search(&query, &limit).unwrap();
docs.into_iter()
.map(|(score, address)| {
let doc: TantivyDocument = searcher.doc(address).unwrap();
match doc.get_first(identifier).map(|v| v.as_i64()) {
Some(Some(value)) => (value, score),
_ => panic!("Invalid story key type!"),
}
})
.collect()
}
pub fn parse(&self, text: &str) -> impl Fn(&Story) -> bool + Sync {
let mut sids: Vec<_> = self
.search(text)
.into_iter()
.filter(|(_, score)| *score > 10f32)
.map(|(sid, _)| sid)
.collect();
sids.sort();
move |story| sids.binary_search(&story.id.into()).is_ok()
}
}

View file

@ -104,10 +104,26 @@ impl<T: Read + Seek> Fetcher<T> {
Ok(buf) Ok(buf)
} }
pub fn identity(&self) -> Result<String> {
let Ok(mut archive) = self.archive.lock() else {
return Err(Error::archive("Could not acquire fetcher lock"));
};
let Ok(index) = archive.by_name("index.json") else {
return Err(Error::archive("Could not open archive index"));
};
Ok(format!("{}", index.crc32()))
}
pub fn iter(&self) -> impl Iterator<Item = &Story> { pub fn iter(&self) -> impl Iterator<Item = &Story> {
self.index.iter() self.index.iter()
} }
pub fn par_iter(&self) -> impl ParallelIterator<Item = &Story> {
self.index.par_iter()
}
pub fn filter<F>(&self, function: &F) -> Vec<&Story> pub fn filter<F>(&self, function: &F) -> Vec<&Story>
where where
F: Sync + Fn(&Story) -> bool, F: Sync + Fn(&Story) -> bool,