feat: add rustc-hash feature (#42)

* feat: add rustc-hash feature

Changes in preparation of [rust-lang/rust#131936][1]:

- Introduce `rustc-hash` dependency and feature.
- Modify the `update.sh` script accordingly.

[1]: rust-lang/rust#131936

* chore: run ./update.sh

* feat(ci): also test with the `rustc-hash` feature

* README reword

---------

Co-authored-by: Alona Enraght-Moony <[email protected]>

Author: jalil-salame

Diff for: .github/workflows/CI.yml

@@ -12,8 +12,13 @@ env:
 jobs:
   cargo-test:
     runs-on: ubuntu-latest
+    strategy:
+      matrix:
+        features:
+          - default
+          - rustc-hash
     steps:
     - uses: actions/checkout@v4
     - uses: dtolnay/rust-toolchain@stable
     - uses: Swatinem/rust-cache@v2
-    - run: cargo test
+    - run: cargo test --no-default-features --features '${{ matrix.features }}'

Diff for: COMMIT.txt

@@ -1 +1 @@
-2e6f3bd1d32455e535de1d9ee154253c333aec73
+d1fa49b2e66c343210c413b68ed57f150b7b89d8

Diff for: Cargo.toml

@@ -9,6 +9,16 @@ repository = "https://github.com/rust-lang/rustdoc-types"
 
 [dependencies]
 serde = {version="1", features=["derive"]}
+rustc-hash = {version="2", optional=true}
+
+[features]
+default = []
+
+# Switch the hashmaps used in rustdoc-types to the FxHashMap from rustc-hash.
+#
+# This might improve performace if your are reading the rustdoc JSON from large
+# crates like aws_sdk_ec2
+rustc-hash = ["dep:rustc-hash"]
 
 [dev-dependencies]
 bincode = "1.3.3"

Diff for: README.md

@@ -13,6 +13,18 @@ let krate: rustdoc_types::Crate = serde_json::from_str(&json_string)?;
 println!("the index has {} items", krate.index.len());
 ```
 
+For performance sensitive crates, consider turning on the `rustc-hash`
+feature. This switches all data structures from `std::collections::HashMap` to
+`rustc-hash::FxHashMap` which improves performance when reading big JSON files
+(like `aws_sdk_rs`'s).
+
+`cargo-semver-checks` benchmarked this change with `aws_sdk_ec2`'s JSON and
+[observed a -3% improvement to the runtime][csc benchmarks]. The performance
+here depends on how much time you spend querying the `HashMap`s, so as always,
+measure first.
+
+[csc benchmarks]: https://rust-lang.zulipchat.com/#narrow/channel/266220-t-rustdoc/topic/rustc-hash.20and.20performance.20of.20rustdoc-types/near/474855731
+
 ## Contributing
 
 This repo is a reexport of

Diff for: src/lib.rs

@@ -3,11 +3,15 @@
 //! These types are the public API exposed through the `--output-format json` flag. The [`Crate`]
 //! struct is the root of the JSON blob and all other items are contained within.
 
+#[cfg(not(feature = "rustc-hash"))]
+use std::collections::HashMap;
 use std::path::PathBuf;
 
-use std::collections::HashMap;
+#[cfg(feature = "rustc-hash")]
+use rustc_hash::FxHashMap as HashMap;
 use serde::{Deserialize, Serialize};
 
+
 /// The version of JSON output that this crate represents.
 ///
 /// This integer is incremented with every breaking change to the API,

Diff for: update.sh

@@ -9,9 +9,7 @@ repo="rust"
 branch="master"
 
 curl -# https://raw.githubusercontent.com/${user}/${repo}/${branch}/src/rustdoc-json-types/lib.rs \
-    | sed 's/rustc_hash::/std::collections::/g' \
-    | sed 's/FxHashMap/HashMap/g' \
-    | sed 's/^pub use /use /' \
+    | sed '/^pub type FxHashMap.*$/d' \
     > src/lib.rs
 
 curl -# https://raw.githubusercontent.com/${user}/${repo}/${branch}/src/rustdoc-json-types/tests.rs > src/tests.rs