feat: indexer

README.md

@@ -2,57 +2,93 @@
 
 Opinionated Python bindings for the [tree-sitter-stack-graphs](https://github.com/github/stack-graphs) rust library.
 
-It exposes very few, easy to use functions to index files and query references.
+It exposes a minimal, opinionated API to leverage the stack-graphs library for reference resolution in source code.
 
-This is a proof of concept draft, to test scripting utilities using stack-graphs easily.
+The rust bindings are built using [PyO3](https://pyo3.rs) and [maturin](https://maturin.rs).
 
-It uses pyo3 and maturin to generate the bindings.
+Note that this is a work in progress, and the API is subject to change. This project is not affiliated with GitHub.
 
 ## Installation & Usage
 
 ```bash
-pip install stack-graphs-python-bindings # or poetry, ...
+pip install stack-graphs-python-bindings
 ```
 
+### Example
+
+Given the following directory structure:
+
+```bash
+tests/js_sample
+├── index.js
+└── module.js
+```
+
+`index.js`:
+
+```javascript
+import { foo } from "./module"
+const baz = foo
+```
+
+`module.js`:
+
+```javascript
+export const foo = "bar"
+```
+
+The following Python script:
+
 ```python
 import os
-from stack_graphs_python import index, Querier, Position, Language
+from stack_graphs_python import Indexer, Querier, Position, Language
 
 db_path = os.path.abspath("./db.sqlite")
 dir = os.path.abspath("./tests/js_sample")
 
 # Index the directory (creates stack-graphs database)
-index([dir], db_path, language=Language.JavaScript)
+indexer = Indexer(db_path, [Language.JavaScript])
+indexer.index_all([dir])
 
 # Instantiate a querier
 querier = Querier(db_path)
 
-# Query a reference at a given position (0-indexed line and column): 
+# Query a reference at a given position (0-indexed line and column):
 # foo in: const baz = foo
 source_reference = Position(path=dir + "/index.js", line=2, column=12)
 results = querier.definitions(source_reference)
 
 for r in results:
-    print(f"{r.path}, l:{r.line}, c: {r.column}")
+    print(r)
 ```
 
-Will result in:
+Will output:
 
 ```bash
-[...]/stack-graphs-python-bindings/tests/js_sample/index.js, l:0, c: 9
-[...]/stack-graphs-python-bindings/tests/js_sample/module.js, l:0, c: 13
+Position(path="[...]/tests/js_sample/index.js", line=0, column=9)
+Position(path="[...]/tests/js_sample/module.js", line=0, column=13)
 ```
 
 That translates to:
 
 ```javascript
 // index.js
 import { foo } from "./module"
+      // ^ line 0, column 9
 
 // module.js
 export const foo = "bar"
+          // ^ line 0, column 13
 ```
 
+> **Note**: All the paths are absolute, and line and column numbers are 0-indexed (first line is 0, first column is 0).
+
+## Known stack-graphs / tree-sitter issues
+
+- Python: module resolution / imports seems to be broken: <https://github.com/github/stack-graphs/issues/430>
+- Typescript: module resolution doesn't work with file extensions (eg. `import { foo } from "./module"` is ok, but `import { foo } from "./module.ts"` is not). **An issue should be opened on the stack-graphs repo**. See: `tests/ts_ok_test.py`
+- Typescript: tree-sitter-typescript fails when passing a generic type to a decorator: <https://github.com/tree-sitter/tree-sitter-typescript/issues/283>
+
 ## Development
 
 ### Ressources
@@ -67,7 +103,7 @@ https://pyo3.rs/v0.21.2/getting-started
 ### Setup
 
 ```bash
-# Setup venv and install maturin through pip
+# Setup venv and install dev dependencies
 make setup
 ```
 
@@ -76,3 +112,37 @@ make setup
 ```bash
 make test
 ```
+
+### Manual testing
+
+```bash
+# build the package
+make develop
+# activate the venv
+. venv/bin/activate
+```
+
+### Roadmap
+
+Before releasing 0.1.0, which I expect to be a first stable API, the following needs to be done:
+
+- [ ] Add more testing, especially:
+  - [ ] Test all supported languages (Java, ~~Python~~, ~~TypeScript~~, ~~JavaScript~~)
+  - [ ] Test failing cases, eg. files that cannot be indexed
+- [ ] Add options to the classes:
+  - [ ] Verbosity
+  - [ ] Force for the Indexer
+  - [ ] Fail on error for the Indexer, or continue indexing
+- [ ] Handle the storage (database) in a dedicated class, and pass it to the Indexer and Querier
+- [ ] Add methods to query the indexing status (eg. which files have been indexed, which failed, etc.)
+- [ ] Rely on the main branch of stack-graphs, and update the bindings accordingly
+- [ ] Better error handling, return clear errors, test them and add them to the `.pyi` interface
+- [ ] Lint and format the rust code
+- [ ] CI/CD for the rust code
+- [ ] Lint and format the python code
+- [ ] Propper changelog, starting in 0.1.0
+
+I'd also like to add the following features, after 0.1.0:
+
+- [ ] Expose the exact, lower-level API of stack-graphs, for more flexibility, in a separate module (eg. `stack_graphs_python.core`)
+- [ ] Benchmark performance

src/classes.rs

@@ -2,10 +2,11 @@ use std::fmt::Display;
 
 use pyo3::prelude::*;
 
-use stack_graphs::storage::SQLiteReader;
+use stack_graphs::storage::{SQLiteReader, SQLiteWriter};
 use tree_sitter_stack_graphs::cli::util::{SourcePosition, SourceSpan};
+use tree_sitter_stack_graphs::loader::Loader;
 
-use crate::stack_graphs_wrapper::query_definition;
+use crate::stack_graphs_wrapper::{index_all, new_loader, query_definition};
 
 #[pyclass]
 #[derive(Clone)]
@@ -62,7 +63,41 @@ impl Querier {
     }
 }
 
-// TODO(@nohehf): Indexer class
+#[pyclass]
+pub struct Indexer {
+    db_writer: SQLiteWriter,
+    db_path: String,
+    loader: Loader,
+}
+
+#[pymethods]
+impl Indexer {
+    #[new]
+    pub fn new(db_path: String, languages: Vec<Language>) -> Self {
+        Indexer {
+            db_writer: SQLiteWriter::open(db_path.clone()).unwrap(),
+            db_path: db_path,
+            loader: new_loader(languages),
+        }
+    }
+
+    pub fn index_all(&mut self, paths: Vec<String>) -> PyResult<()> {
+        let paths: Vec<std::path::PathBuf> =
+            paths.iter().map(|p| std::path::PathBuf::from(p)).collect();
+
+        match index_all(paths, &mut self.loader, &mut self.db_writer) {
+            Ok(_) => Ok(()),
+            Err(e) => Err(e.into()),
+        }
+    }
+
+    // @TODO: Add a method to retrieve the status of the files (indexed, failed, etc.)
+    // This might be done on a separate class (Database / Storage), as it is tied to the storage, not a specific indexer
+
+    fn __repr__(&self) -> String {
+        format!("Indexer(db_path=\"{}\")", self.db_path)
+    }
+}
 
 #[pymethods]
 impl Position {

src/lib.rs

@@ -3,7 +3,7 @@ use pyo3::prelude::*;
 mod classes;
 mod stack_graphs_wrapper;
 
-use classes::{Language, Position, Querier};
+use classes::{Indexer, Language, Position, Querier};
 
 /// Formats the sum of two numbers as string.
 #[pyfunction]
@@ -20,10 +20,10 @@ fn index(paths: Vec<String>, db_path: String, language: Language) -> PyResult<()
     let paths: Vec<std::path::PathBuf> =
         paths.iter().map(|p| std::path::PathBuf::from(p)).collect();
 
-    Ok(stack_graphs_wrapper::index(
+    Ok(stack_graphs_wrapper::index_legacy(
         paths,
         &db_path,
-        language.into(),
+        &language.into(),
     )?)
 }
 
@@ -35,5 +35,6 @@ fn stack_graphs_python(_py: Python, m: &PyModule) -> PyResult<()> {
     m.add_class::<Position>()?;
     m.add_class::<Language>()?;
     m.add_class::<Querier>()?;
+    m.add_class::<Indexer>()?;
     Ok(())
 }

src/stack_graphs_wrapper/mod.rs

@@ -21,7 +21,7 @@ impl std::convert::From<StackGraphsError> for PyErr {
     }
 }
 
-fn get_langauge_configuration(lang: Language) -> LanguageConfiguration {
+pub fn get_langauge_configuration(lang: &Language) -> LanguageConfiguration {
     match lang {
         Language::Python => {
             tree_sitter_stack_graphs_python::language_configuration(&NoCancellation)
@@ -36,10 +36,10 @@ fn get_langauge_configuration(lang: Language) -> LanguageConfiguration {
     }
 }
 
-pub fn index(
+pub fn index_legacy(
     paths: Vec<PathBuf>,
     db_path: &str,
-    language: Language,
+    language: &Language,
 ) -> Result<(), StackGraphsError> {
     let configurations = vec![get_langauge_configuration(language)];
 
@@ -81,6 +81,40 @@ pub fn index(
     }
 }
 
+pub fn new_loader(languages: Vec<Language>) -> Loader {
+    let configurations = languages
+        .iter()
+        .map(|l| get_langauge_configuration(l))
+        .collect();
+
+    Loader::from_language_configurations(configurations, None).unwrap()
+}
+
+pub fn index_all(
+    paths: Vec<PathBuf>,
+    loader: &mut Loader,
+    db_writer: &mut SQLiteWriter,
+) -> Result<(), StackGraphsError> {
+    let reporter = ConsoleReporter::none();
+
+    let mut indexer = Indexer::new(db_writer, loader, &reporter);
+
+    // For now, force reindexing
+    indexer.force = true;
+
+    let paths = canonicalize_paths(paths);
+
+    // https://github.com/github/stack-graphs/blob/7db914c01b35ce024f6767e02dd1ad97022a6bc1/tree-sitter-stack-graphs/src/cli/index.rs#L107
+    let continue_from_none: Option<PathBuf> = None;
+
+    match indexer.index_all(paths, continue_from_none, &NoCancellation) {
+        Ok(_) => Ok(()),
+        Err(e) => Err(StackGraphsError {
+            message: format!("Failed to index: {}", e),
+        }),
+    }
+}
+
 pub fn query_definition(
     reference: SourcePosition,
     db_reader: &mut SQLiteReader,

stack_graphs_python.pyi

@@ -7,6 +7,13 @@ class Language(Enum):
     Java = 3
 
 class Position:
+    """
+    A position in a given file:
+    - path: the path to the file
+    - line: the line number (0-indexed)
+    - column: the column number (0-indexed)
+    """
+
     path: str
     line: int
     column: int
@@ -16,8 +23,38 @@ class Position:
     def __repr__(self) -> str: ...
 
 class Querier:
+    """
+    A class to query the stack graphs database
+    - db_path: the path to the database
+
+    Usage: see Querier.definitions
+    """
     def __init__(self, db_path: str) -> None: ...
-    def definitions(self, reference: Position) -> list[Position]: ...
+    def definitions(self, reference: Position) -> list[Position]:
+        """
+        Get the definitions of a given reference
+        - reference: the position of the reference
+        - returns: a list of positions of the definitions
+        """
+        ...
+    def __repr__(self) -> str: ...
+
+class Indexer:
+    """
+    A class to build the stack graphs of a given set of files
+    - db_path: the path to the database
+    - languages: the list of languages to index
+    """
+    def __init__(self, db_path: str, languages: list[Language]) -> None: ...
+    def index_all(self, paths: list[str]) -> None:
+        """
+        Index all the files in the given paths, recursively
+        """
+        ...
     def __repr__(self) -> str: ...
 
-def index(paths: list[str], db_path: str, language: Language) -> None: ...
+def index(paths: list[str], db_path: str, language: Language) -> None:
+    """
+    DeprecationWarning: The 'index' function is deprecated. Use 'Indexer' instead.
+    """
+    ...

tests/helpers/virtual_files.py

@@ -46,7 +46,7 @@ def _get_positions_in_file(file_path: str, contents: str) -> dict[str, Position]
 
 
 @contextlib.contextmanager
-def string_to_virtual_repo(
+def string_to_virtual_files(
     string: str,
 ) -> Iterator[tuple[str, dict[str, Position]]]:
     """
@@ -62,7 +62,7 @@ def string_to_virtual_repo(
     ^{pos2}
     \"""
 
-    with string_to_virtual_repo(string) as (repo_path, positions):
+    with string_to_virtual_files(string) as (repo_path, positions):
         ...
     ```
 
@@ -104,7 +104,7 @@ def string_to_virtual_repo(
 
     When parsed via:
     ```py
-    with string_to_virtual_repo(string) as (repo_path, positions):
+    with string_to_virtual_files(string) as (repo_path, positions):
         ...
     ```