feat: sync with tantivy-py 0.26.2

.husky/pre-commit

@@ -1,4 +1 @@
-#!/bin/sh
-. "$(dirname "$0")/_/husky.sh"
-
-yarn lint-staged
+pnpm lint-staged

.oxlintrc.json

@@ -0,0 +1,4 @@
+{
+  "$schema": "./node_modules/oxlint/configuration_schema.json",
+  "ignorePatterns": ["tantivy-py/**"]
+}

.prettierignore

@@ -5,4 +5,6 @@ package-template.wasi-browser.js
 package-template.wasi.cjs
 wasi-worker-browser.mjs
 wasi-worker.mjs
-.yarnrc.yml
+.yarnrc.yml
+tantivy-py
+pnpm-lock.yaml

.taplo.toml

@@ -1,4 +1,4 @@
-exclude = ["node_modules/**/*.toml"]
+exclude = ["node_modules/**/*.toml", "tantivy-py/**/*.toml"]
 
 # https://taplo.tamasfe.dev/configuration/formatter-options.html
 [formatting]

Cargo.toml

@@ -10,14 +10,16 @@ crate-type = ["cdylib"]
 
 [dependencies]
 chrono = { version = "0.4", features = ["serde"] }
+futures = "0.3"
 napi = { version = "3.12", default-features = false, features = [
   "napi8",
   "serde-json",
 ] }
 napi-derive = "3.6"
 serde = { version = "1.0", features = ["derive"] }
 serde_json = { version = "1.0", default-features = false, features = ["alloc"] }
-tantivy = "0.25.0"
+tantivy = "0.26.2"
+tantivy-common = "0.11"
 
 [build-dependencies]
 napi-build = "2.4"

README.md

@@ -6,8 +6,6 @@ Node.js bindings for [Tantivy](https://github.com/quickwit-oss/tantivy), the ful
 
 This project is a Node.js port of [tantivy-py](https://github.com/quickwit-inc/tantivy-py), providing JavaScript/TypeScript bindings for the Tantivy search engine. The implementation closely follows the Python API to maintain consistency across language bindings.
 
-
-
 # Installation
 
 The bindings can be installed using npm:
@@ -60,8 +58,11 @@ This Node.js binding provides access to most of Tantivy's functionality:
 - **Snippet generation** for search result highlighting
 - **Query explanation** for debugging relevance scoring
 - **Multiple field types**: text, integers, floats, dates, facets
-- **Flexible tokenization** and text analysis
-- **JSON document support**
+- **Flexible tokenization** and text analysis, including fast-field tokenizers
+- **JSON document support**, with optional dot-path expansion
+- **Aggregations**, including cardinality
+- **Fast-field reads** (`fastFieldValues`) and prefix term lookup (`termsWithPrefix`)
+- **Schema-free query parsing** via `parseQuery` / `parseQueryLenient`
 
 ## API Compatibility
 
@@ -115,18 +116,21 @@ The Node.js implementation differs from the Python version in several ways:
 
 #### 🔴 Critical Validation Issues
 
-##### Numeric Field Validation (Too Lenient)
+##### Multi-valued Single Fields (Too Lenient)
 
-**Current behavior**: Node.js version accepts invalid values that Python rejects
-**TODO**: Implement strict validation to match Python behavior
+**Current behavior**: an array is accepted wherever a single value is expected
+**TODO**: decide whether to match Python, which rejects it
 
 ```javascript
-// ❌ These currently PASS in Node.js but should FAIL:
-Document.fromDict({ unsigned: -50 }, schema) // Should reject negative for unsigned
-Document.fromDict({ signed: 50.4 }, schema) // Should reject float for integer
-Document.fromDict({ unsigned: [1000, -50] }, schema) // Should reject arrays for single fields
+// ❌ This currently PASSES in Node.js but should FAIL:
+Document.fromDict({ unsigned: [1000, 50] }, schema) // Should reject arrays for single fields
 ```
 
+Numeric values themselves are validated as in tantivy-py: a negative value for an
+unsigned field, or a fractional value for an integer field, throws rather than
+being silently coerced. Values beyond `Number.MAX_SAFE_INTEGER` can be passed as
+a `BigInt`.
+
 ##### Bytes Field Validation (Too Restrictive)
 
 **Current behavior**: Only accepts Buffer objects
@@ -157,26 +161,55 @@ Document.fromDict({ json: 123 }, schema) // Should reject numbers
 Document.fromDict({ json: 'hello' }, schema) // Should reject strings
 ```
 
-#### 🟠 Error Handling Differences
+#### 🔵 Type System Differences
 
-##### Fast Field Configuration
+##### Date Handling
 
-**Current**: Throws exception when field not configured as fast
-**Python**: Returns empty results
-**TODO**: Decide on consistent error handling approach
+**Current**: Dates cross the boundary as millisecond timestamps. `Document.fromDict()`
+and the query builders also accept a JavaScript `Date` or an ISO 8601 string, and
+`toDict()` returns milliseconds since the epoch.
+**Python**: Uses `datetime` objects with nanosecond storage.
 
-##### Query Parser Errors
+Sub-millisecond precision is therefore not representable from JavaScript.
 
-**Current**: Different error message formats
-**TODO**: Align error messages with Python version
+#### 🔵 Resource Management
 
-#### 🔵 Type System Differences
+`IndexWriter` has no equivalent of tantivy-py's `with index.writer() as writer:`
+block, because JavaScript has no deterministic destructor. Call
+`writer.waitMergingThreads()` when you are done with a writer — it commits nothing
+on its own, so commit first, and it is what releases the directory lock:
 
-##### Date Handling
+```javascript
+const writer = index.writer()
+try {
+  writer.addDocument(doc)
+  writer.commit()
+} finally {
+  writer.waitMergingThreads()
+}
+```
+
+#### 🔵 Concurrency
+
+tantivy-py releases the GIL around the blocking calls (`addDocument`, `commit`,
+`search`, …). Node has no GIL, but these methods run synchronously on the main
+thread and are not offloaded to the libuv thread pool.
+
+#### 🔵 Naming
+
+Two shapes differ from tantivy-py because napi-rs models them differently:
+
+- Static constructors live on a companion class: `FilterStatic.lowercase()` and
+  `TokenizerStatic.simple()` rather than `Filter.lowercase()` / `Tokenizer.simple()`.
+- `DocAddress`, `SearchResult` and `SearchHit` are plain objects rather than
+  classes, so they have no constructors or getters — read `hit.docAddress`,
+  `hit.score` and `hit.order` directly.
+
+`FieldType` uses tantivy-py's variant names (`Text`, `Unsigned`, `Integer`,
+`Float`, `Boolean`, `Json`), not tantivy's Rust `Type` names.
 
-**Current**: Uses getTime() timestamps
-**Python**: Uses datetime objects
-**TODO**: Consider more intuitive date API
+Python's pickle hooks (`__reduce__`, `__getnewargs__`) have no counterpart;
+`Schema` exposes `toJSON()` / `Schema.fromJson()` instead.
 
 ## Architecture
 

__test__/fixtures.ts

@@ -214,3 +214,75 @@ export const createSpanishIndex = () => {
   index.reload()
   return index
 }
+
+export const schemaOrderFastFields = () => {
+  return new SchemaBuilder()
+    .addTextField('title', { stored: true })
+    .addUnsignedField('u64_field', { fast: true })
+    .addIntegerField('i64_field', { fast: true })
+    .addFloatField('f64_field', { fast: true })
+    .addBooleanField('bool_field', { fast: true })
+    .addDateField('date_field', { fast: true })
+    .addTextField('str_field', { fast: true })
+    .build()
+}
+
+export const createIndexWithOrderFastFields = (dir?: string) => {
+  const index = new Index(schemaOrderFastFields(), dir)
+  const writer = index.writer(15_000_000, 1)
+
+  const low = new Document()
+  low.addText('title', 'low title')
+  low.addUnsigned('u64_field', 0)
+  low.addInteger('i64_field', -10)
+  low.addFloat('f64_field', 1.5)
+  low.addBoolean('bool_field', false)
+  low.addDate('date_field', Date.UTC(2024, 0, 1))
+  low.addText('str_field', 'apple')
+  writer.addDocument(low)
+
+  const high = new Document()
+  high.addText('title', 'high title')
+  high.addUnsigned('u64_field', 2)
+  high.addInteger('i64_field', 5)
+  high.addFloat('f64_field', 3.14)
+  high.addBoolean('bool_field', true)
+  high.addDate('date_field', Date.UTC(2026, 0, 1))
+  high.addText('str_field', 'cherry')
+  writer.addDocument(high)
+
+  writer.commit()
+  writer.waitMergingThreads()
+  index.reload()
+  return index
+}
+
+export const createIndexWithEmptyFastField = () => {
+  const indexSchema = new SchemaBuilder()
+    .addTextField('title', { fast: true, stored: true })
+    .addTextField('body', { fast: true })
+    .build()
+
+  const index = new Index(indexSchema)
+  const writer = index.writer(15_000_000, 1)
+
+  writer.addDocument(
+    Document.fromDict({
+      title:
+        'A record of the statesmanship and political achievements of Gen. Winfield Scott Hancock, ' +
+        'regular Democratic nominee for president of the United States',
+    }),
+  )
+  writer.addDocument(Document.fromDict({ title: 'Political Achievements of the Earl of Dalkeith' }))
+  writer.addDocument(
+    Document.fromDict({
+      title: 'The Old Man and the Sea',
+      body: 'He was an old man who fished alone inthe Gulf Stream and he had gone eighty-four days now without taking a fish.',
+    }),
+  )
+
+  writer.commit()
+  writer.waitMergingThreads()
+  index.reload()
+  return index
+}