Merge pull request #293 from github/change-storage-serde

Improve SQLite storage

Author: Hendrik van Antwerpen

languages/tree-sitter-stack-graphs-typescript/Cargo.toml

@@ -32,7 +32,7 @@ clap = { version = "4", optional = true }
 glob = "0.3"
 serde = { version = "1.0", features = ["derive"] }
 serde_json = "1.0"
-stack-graphs = { version = "0.11", path = "../../stack-graphs" }
+stack-graphs = { version = ">=0.11, <=0.12", path = "../../stack-graphs" }
 tree-sitter-stack-graphs = { version = "0.7", path = "../../tree-sitter-stack-graphs" }
 tree-sitter-typescript = "0.20.2"
 tsconfig = "0.1.0"

stack-graphs/CHANGELOG.md

@@ -5,13 +5,19 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
-## Unreleased
+## v0.12.0 -- 2023-07-27
+
+### Added
+
+- New `SQLiteReader::clear` and `SQLiteReader::clear_paths` methods that make it easier to reuse instances.
+- The method `SQLiteReader::load_graph_for_file` now returns the file handle for the loaded file.
 
 ### Changed
 
 - The `Appendable` trait has been simplified. Its `Ctx` type parameter is gone, in favor of a separate trait `ToAppendable` that is used to find appendables for a handle. The type itself moved from the `cycles` to the `stitching` module.
 - The `ForwardPartialPathStitcher` has been generalized so that it can be used to build paths from a database or from graph edges. It now takes a type parameter indicating the type of candidates it uses. Instead of a `Database` instance, it expects a value that implements the `Candidates` and `ToAppendable` traits. The `ForwardPartialPathStitcher::process_next_phase` expects an additional `extend_until` closure that controls whether the extended paths are considered for further extension or not (using `|_,_,_| true` retains old behavior).
 - The SQLite database implementation is using a new schema which stores binary instead of JSON values, resulting in faster write times and smaller databases.
+- Renamed method `SQLiteReader::load_graph_for_file_or_directory` to `SQLiteReader::load_graphs_for_file_or_directory`.
 
 ### Fixed
 

stack-graphs/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "stack-graphs"
-version = "0.11.0"
+version = "0.12.0"
 description = "Name binding for arbitrary programming languages"
 homepage = "https://github.com/github/stack-graphs/tree/main/stack-graphs"
 repository = "https://github.com/github/stack-graphs/"
@@ -15,7 +15,7 @@ edition = "2018"
 [features]
 copious-debugging = []
 serde = ["dep:serde", "lsp-positions/serde"]
-storage = ["rusqlite", "serde", "rmp-serde"]
+storage = ["postcard", "rusqlite", "serde"]
 visualization = ["serde", "serde_json"]
 
 [lib]
@@ -31,7 +31,7 @@ fxhash = "0.2"
 itertools = "0.10"
 libc = "0.2"
 lsp-positions = { version = "0.3", path = "../lsp-positions" }
-rmp-serde = { version = "1.1", optional = true }
+postcard = { version = "1", optional = true, features = ["use-std"] }
 rusqlite = { version = "0.28", optional = true, features = ["bundled", "functions"] }
 serde = { version = "1.0", optional = true, features = ["derive"] }
 serde_json = { version = "1.0", optional = true }

stack-graphs/README.md

@@ -9,7 +9,7 @@ To use this library, add the following to your `Cargo.toml`:
 
 ``` toml
 [dependencies]
-stack-graphs = "0.11"
+stack-graphs = "0.12"
 ```
 
 Check out our [documentation](https://docs.rs/stack-graphs/) for more details on

stack-graphs/src/arena.rs

@@ -181,6 +181,14 @@ impl<T> Arena<T> {
         }
     }
 
+    /// Clear the arena, keeping underlying allocated capacity.  After this, all previous handles into
+    /// the arena are invalid.
+    #[cfg_attr(not(feature = "storage"), allow(dead_code))]
+    #[inline(always)]
+    pub(crate) fn clear(&mut self) {
+        self.items.clear();
+    }
+
     /// Adds a new instance to this arena, returning a stable handle to it.
     ///
     /// Note that we do not deduplicate instances of `T` in any way.  If you add two instances that
@@ -280,6 +288,14 @@ impl<H, T> SupplementalArena<H, T> {
         }
     }
 
+    /// Clear the supplemantal arena, keeping underlying allocated capacity.  After this,
+    /// all previous handles into the arena are invalid.
+    #[cfg_attr(not(feature = "storage"), allow(dead_code))]
+    #[inline(always)]
+    pub(crate) fn clear(&mut self) {
+        self.items.clear();
+    }
+
     /// Creates a new, empty supplemental arena, preallocating enough space to store supplemental
     /// data for all of the instances that have already been allocated in a (regular) arena.
     pub fn with_capacity(arena: &Arena<H>) -> SupplementalArena<H, T> {

stack-graphs/src/partial.rs

@@ -2624,4 +2624,11 @@ impl PartialPaths {
             partial_path_edges: Deque::new_arena(),
         }
     }
+
+    #[cfg_attr(not(feature = "storage"), allow(dead_code))]
+    pub(crate) fn clear(&mut self) {
+        self.partial_symbol_stacks.clear();
+        self.partial_scope_stacks.clear();
+        self.partial_path_edges.clear();
+    }
 }

stack-graphs/src/stitching.rs

@@ -249,6 +249,18 @@ impl Database {
         }
     }
 
+    /// Clear the database.  After this, all previous handles into the database are
+    /// invalid.
+    #[cfg_attr(not(feature = "storage"), allow(dead_code))]
+    pub(crate) fn clear(&mut self) {
+        self.partial_paths.clear();
+        self.local_nodes.clear();
+        self.symbol_stack_keys.clear();
+        self.symbol_stack_key_cache.clear();
+        self.paths_by_start_node.clear();
+        self.root_paths_by_precondition.clear();
+    }
+
     /// Adds a partial path to this database.  We do not deduplicate partial paths in any way; it's
     /// your responsibility to only add each partial path once.
     pub fn add_partial_path(

stack-graphs/src/storage.rs

@@ -30,7 +30,7 @@ use crate::stitching::ForwardPartialPathStitcher;
 use crate::CancellationError;
 use crate::CancellationFlag;
 
-const VERSION: usize = 3;
+const VERSION: usize = 4;
 
 const SCHEMA: &str = r#"
         CREATE TABLE metadata (
@@ -75,9 +75,7 @@ pub enum StorageError {
     #[error(transparent)]
     Serde(#[from] serde::Error),
     #[error(transparent)]
-    RmpSerdeDecode(#[from] rmp_serde::decode::Error),
-    #[error(transparent)]
-    RmpSerdeEncode(#[from] rmp_serde::encode::Error),
+    PostcardError(#[from] postcard::Error),
 }
 
 pub type Result<T> = std::result::Result<T, StorageError>;
@@ -283,7 +281,7 @@ impl SQLiteWriter {
             &file.to_string_lossy(),
             tag,
             error,
-            &rmp_serde::to_vec(&graph)?,
+            &postcard::to_stdvec(&graph)?,
         ))?;
         Ok(())
     }
@@ -323,7 +321,7 @@ impl SQLiteWriter {
         let mut stmt =
             conn.prepare_cached("INSERT INTO graphs (file, tag, value) VALUES (?, ?, ?)")?;
         let graph = serde::StackGraph::from_graph_filter(graph, &FileFilter(file));
-        stmt.execute((file_str, tag, &rmp_serde::to_vec(&graph)?))?;
+        stmt.execute((file_str, tag, &postcard::to_stdvec(&graph)?))?;
         Ok(())
     }
 
@@ -364,7 +362,7 @@ impl SQLiteWriter {
                 );
                 let symbol_stack = path.symbol_stack_precondition.storage_key(graph, partials);
                 let path = serde::PartialPath::from_partial_path(graph, partials, path);
-                root_stmt.execute((file_str, symbol_stack, &rmp_serde::to_vec(&path)?))?;
+                root_stmt.execute((file_str, symbol_stack, &postcard::to_stdvec(&path)?))?;
                 root_path_count += 1;
             } else if start_node.is_in_file(file) {
                 copious_debugging!(
@@ -375,7 +373,7 @@ impl SQLiteWriter {
                 node_stmt.execute((
                     file_str,
                     path.start_node.local_id,
-                    &rmp_serde::to_vec(&path)?,
+                    &postcard::to_stdvec(&path)?,
                 ))?;
                 node_path_count += 1;
             } else {
@@ -447,6 +445,28 @@ impl SQLiteReader {
         })
     }
 
+    /// Clear all data that has been loaded into this reader instance.
+    /// After this call, all existing handles from this reader are invalid.
+    pub fn clear(&mut self) {
+        self.loaded_graphs.clear();
+        self.graph = StackGraph::new();
+
+        self.loaded_node_paths.clear();
+        self.loaded_root_paths.clear();
+        self.partials.clear();
+        self.db.clear();
+    }
+
+    /// Clear path data that has been loaded into this reader instance.
+    /// After this call, all node handles remain valid, but all path data
+    /// is invalid.
+    pub fn clear_paths(&mut self) {
+        self.loaded_node_paths.clear();
+        self.loaded_root_paths.clear();
+        self.partials.clear();
+        self.db.clear();
+    }
+
     /// Get the file's status in the database. If a tag is provided, it must match or the file
     /// is reported missing.
     pub fn status_for_file<T: AsRef<str>>(
@@ -485,7 +505,7 @@ impl SQLiteReader {
     }
 
     /// Ensure the graph for the given file is loaded.
-    pub fn load_graph_for_file(&mut self, file: &str) -> Result<()> {
+    pub fn load_graph_for_file(&mut self, file: &str) -> Result<Handle<File>> {
         Self::load_graph_for_file_inner(file, &mut self.graph, &mut self.loaded_graphs, &self.conn)
     }
 
@@ -494,21 +514,21 @@ impl SQLiteReader {
         graph: &mut StackGraph,
         loaded_graphs: &mut HashSet<String>,
         conn: &Connection,
-    ) -> Result<()> {
+    ) -> Result<Handle<File>> {
         copious_debugging!("--> Load graph for {}", file);
         if !loaded_graphs.insert(file.to_string()) {
             copious_debugging!(" * Already loaded");
-            return Ok(());
+            return Ok(graph.get_file(file).expect("loaded file to exist"));
         }
         copious_debugging!(" * Load from database");
-        let mut stmt = conn.prepare_cached("SELECT json FROM graphs WHERE file = ?")?;
+        let mut stmt = conn.prepare_cached("SELECT value FROM graphs WHERE file = ?")?;
         let value = stmt.query_row([file], |row| row.get::<_, Vec<u8>>(0))?;
-        let file_graph = rmp_serde::from_slice::<serde::StackGraph>(&value)?;
+        let file_graph = postcard::from_bytes::<serde::StackGraph>(&value)?;
         file_graph.load_into(graph)?;
-        Ok(())
+        Ok(graph.get_file(file).expect("loaded file to exist"))
     }
 
-    pub fn load_graph_for_file_or_directory(
+    pub fn load_graphs_for_file_or_directory(
         &mut self,
         file_or_directory: &Path,
         cancellation_flag: &dyn CancellationFlag,
@@ -559,7 +579,7 @@ impl SQLiteReader {
                 &mut self.loaded_graphs,
                 &self.conn,
             )?;
-            let path = rmp_serde::from_slice::<serde::PartialPath>(&value)?;
+            let path = postcard::from_bytes::<serde::PartialPath>(&value)?;
             let path = path.to_partial_path(&mut self.graph, &mut self.partials)?;
             copious_debugging!(
                 "   > Loaded {}",
@@ -613,7 +633,7 @@ impl SQLiteReader {
                     &mut self.loaded_graphs,
                     &self.conn,
                 )?;
-                let path = rmp_serde::from_slice::<serde::PartialPath>(&value)?;
+                let path = postcard::from_bytes::<serde::PartialPath>(&value)?;
                 let path = path.to_partial_path(&mut self.graph, &mut self.partials)?;
                 copious_debugging!(
                     "   > Loaded {}",

tree-sitter-stack-graphs/CHANGELOG.md

@@ -5,6 +5,10 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## v0.7.1 -- 2023-07-27
+
+Support `stack-graphs` version `0.12`.
+
 ## v0.7.0 -- 2023-06-08
 
 ### Library

tree-sitter-stack-graphs/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "tree-sitter-stack-graphs"
-version = "0.7.0"
+version = "0.7.1"
 description = "Create stack graphs using tree-sitter parsers"
 homepage = "https://github.com/github/stack-graphs/tree/main/tree-sitter-stack-graphs"
 repository = "https://github.com/github/stack-graphs/"
@@ -69,7 +69,7 @@ regex = "1"
 rust-ini = "0.18"
 serde_json = { version="1.0", optional=true }
 sha1 = { version="0.10", optional=true }
-stack-graphs = { version="0.11", path="../stack-graphs" }
+stack-graphs = { version=">=0.11, <=0.12", path="../stack-graphs" }
 thiserror = "1.0"
 time = { version = "0.3", optional = true }
 tokio = { version = "1.26", optional = true, features = ["io-std", "rt", "rt-multi-thread"] }