Use Tantivy to query story content

.gitignore

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

Cargo.toml

@@ -8,6 +8,7 @@ edition = "2021"
 members = [
     "cli",
     "query",
+    "search",
 ]
 
 default-members = [

cli/Cargo.toml

@@ -7,9 +7,14 @@ edition = "2021"
 [dependencies.fimfareader]
 path = ".."
 
-[dependencies.fimfareader-query]
+[dependencies.fimfareader-search]
-path = "../query"
+path = "../search"
 
 [dependencies.rustyline]
 default-features = false
 version = "14"
+
+[dependencies.zip]
+version = "*"
+features = ["deflate"]
+default-features = false

cli/src/main.rs

@@ -1,13 +1,16 @@
 //! Main module.
 
 use std::env::args;
+use std::fs::File;
+use std::io::BufReader;
 use std::time::Instant;
 
 use rustyline::DefaultEditor;
 use rustyline::Result;
+use zip::ZipArchive;
 
 use fimfareader::prelude::*;
-use fimfareader_query::parse;
+use fimfareader_search::Searcher;
 
 fn exit(error: Error) -> ! {
     eprintln!("{}", error);
@@ -34,33 +37,47 @@ fn main() -> Result<()> {
     println!("Finished loading in {} milliseconds.", finish);
     println!("The archive contains {} stories.", count);
 
+    let opener = || {
+        let file = File::open(&argv[1]).unwrap();
+        let reader = BufReader::new(file);
+
+        ZipArchive::new(reader).unwrap()
+    };
+
+    let searcher = Searcher::new(&fetcher, &opener);
+
     while let Ok(line) = editor.readline(">>> ") {
         editor.add_history_entry(&line)?;
-
-        let filter = match parse(&line) {
-            Ok(filter) => filter,
-            Err(error) => {
-                println!("{}", error);
-                continue;
-            }
-        };
-
         let start = Instant::now();
-        let stories = fetcher.filter(&filter);
+
+        let result = searcher
+            .search(&line)
+            .into_iter()
+            .filter(|(_sid, score)| *score > 10f32)
+            .filter_map(|(sid, score)| Some((fetcher.fetch(sid)?, score)))
+            .collect::<Vec<_>>();
+
         let finish = (Instant::now() - start).as_millis();
-        let count = stories.len();
+        let count = result.len();
 
         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 title = &story.title;
 
-            println!("[{}] {}", key, title);
+            let tags = story
+                .tags
+                .iter()
+                .map(|tag| String::from(&tag.name))
+                .collect::<Vec<_>>()
+                .join(", ");
+
+            println!("{:02.02}% [{:>6}] {}", score, key, title);
+            println!("{}", tags);
+            println!("{}", story.short_description);
+            println!("{}", story.url);
+            println!();
         }
     }
 

search/Cargo.toml

@@ -0,0 +1,22 @@
+[package]
+name = "fimfareader-search"
+version = "0.1.0"
+authors = ["Joakim Soderlund <joakim.soderlund@gmail.com>"]
+edition = "2021"
+
+[dependencies.fimfareader]
+path = ".."
+
+[dependencies.rayon]
+version = "*"
+
+[dependencies.tantivy]
+git = "https://github.com/quickwit-oss/tantivy.git"
+
+[dependencies.thread_local]
+version = "*"
+
+[dependencies.zip]
+version = "*"
+features = ["deflate"]
+default-features = false

search/src/lib.rs

@@ -0,0 +1,169 @@
+//! Main module.
+
+use std::cell::RefCell;
+use std::io::Cursor;
+use std::io::Read;
+use std::io::Seek;
+use std::path::Path;
+use std::sync::atomic::AtomicUsize;
+use std::sync::atomic::Ordering;
+use std::time::Instant;
+
+use rayon::iter::ParallelIterator;
+use thread_local::ThreadLocal;
+use zip::read::ZipArchive;
+
+use tantivy::collector::TopDocs;
+use tantivy::directory::MmapDirectory;
+use tantivy::query::QueryParser;
+use tantivy::schema;
+use tantivy::schema::Schema;
+use tantivy::schema::Value;
+use tantivy::Index;
+use tantivy::IndexSettings;
+use tantivy::ReloadPolicy;
+use tantivy::TantivyDocument;
+
+use fimfareader::prelude::*;
+
+pub struct Searcher {
+    schema: Schema,
+    index: Index,
+}
+
+impl Searcher {
+    pub fn new<T, F>(fetcher: &Fetcher<T>, f: &F) -> Self
+    where
+        T: Read + Seek + Send,
+        F: Fn() -> ZipArchive<T> + Sync,
+    {
+        let mut builder = Schema::builder();
+        builder.add_i64_field("sid", schema::INDEXED | schema::STORED);
+        builder.add_text_field("content", schema::TEXT);
+        let schema = builder.build();
+
+        let index = Self::load_index(schema.clone(), fetcher, f);
+
+        Searcher { schema, index }
+    }
+
+    fn load_index<T, F>(schema: Schema, fetcher: &Fetcher<T>, f: &F) -> Index
+    where
+        T: Read + Seek + Send,
+        F: Fn() -> ZipArchive<T> + Sync,
+    {
+        let identity = fetcher.identity().unwrap();
+        let directory = Path::new("cache").join(identity);
+
+        if !directory.exists() {
+            Self::make_index(schema.clone(), fetcher, f);
+        }
+
+        let store = MmapDirectory::open(&directory).unwrap();
+        Index::open_or_create(store, schema).unwrap()
+    }
+
+    fn make_index<T, F>(schema: Schema, fetcher: &Fetcher<T>, f: &F)
+    where
+        T: Read + Seek + Send,
+        F: Fn() -> ZipArchive<T> + Sync,
+    {
+        let identity = fetcher.identity().unwrap();
+        let directory = Path::new("cache").join(identity);
+
+        std::fs::create_dir_all(&directory).unwrap();
+        let store = MmapDirectory::open(&directory).unwrap();
+        let settings = IndexSettings::default();
+        let index = Index::create(store, schema, settings).unwrap();
+
+        let schema = index.schema();
+        let sid = schema.get_field("sid").unwrap();
+        let content = schema.get_field("content").unwrap();
+        let mut writer = index.writer(536_870_912).unwrap();
+
+        let counter = AtomicUsize::new(0);
+        let total = fetcher.iter().count();
+        let start = Instant::now();
+
+        let local = ThreadLocal::new();
+
+        fetcher.par_iter().for_each(|story| {
+            let mut doc = TantivyDocument::default();
+
+            let mut arch = local.get_or(|| RefCell::new(f())).borrow_mut();
+            let mut file = arch.by_name(&story.archive.path).unwrap();
+            let mut data = Vec::with_capacity(file.size() as usize);
+            let mut text = String::with_capacity(1_048_576);
+
+            file.read_to_end(&mut data).unwrap();
+            let mut epub = ZipArchive::new(Cursor::new(data)).unwrap();
+
+            let count = counter.fetch_add(1, Ordering::SeqCst);
+            let percentage = (count as f64 / total as f64) * 100f64;
+            print!("\r\rIndexing archive... {:.2}%\r\r", percentage);
+
+            doc.add_i64(sid, story.id);
+
+            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 text).unwrap();
+                doc.add_text(content, &text);
+                text.clear();
+            }
+
+            writer.add_document(doc).unwrap();
+        });
+
+        writer.commit().unwrap();
+
+        let finish = (Instant::now() - start).as_secs();
+        println!("Index generated in {} seconds.", finish);
+    }
+
+    pub fn search(&self, text: &str) -> Vec<(i64, f32)> {
+        let reader = self
+            .index
+            .reader_builder()
+            .reload_policy(ReloadPolicy::OnCommitWithDelay)
+            .try_into()
+            .unwrap();
+
+        let searcher = reader.searcher();
+        let identitfier = self.schema.get_field("sid").unwrap();
+        let content = self.schema.get_field("content").unwrap();
+        let parser = QueryParser::for_index(&self.index, vec![content]);
+
+        let limit = TopDocs::with_limit(32);
+        let query = parser.parse_query(&text).unwrap();
+        let docs = searcher.search(&query, &limit).unwrap();
+
+        docs.into_iter()
+            .map(|(score, address)| {
+                let doc: TantivyDocument = searcher.doc(address).unwrap();
+
+                match doc.get_first(identitfier).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).is_ok()
+    }
+}

src/archive/fetcher.rs

@@ -97,10 +97,26 @@ impl<T: Read + Seek> Fetcher<T> {
         Ok(buf)
     }
 
+    pub fn identity(&self) -> Result<String> {
+        let mut archive = self.archive.lock().map_err(|e| match e {
+            _ => Error::archive("Could not acquire fetcher lock"),
+        })?;
+
+        let index = archive.by_name("index.json").map_err(|e| match e {
+            _ => Error::archive("Could not open archive index"),
+        })?;
+
+        Ok(format!("{}", index.crc32()))
+    }
+
     pub fn iter(&self) -> impl Iterator<Item = &Story> {
         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>
     where
         F: Sync + Fn(&Story) -> bool,