proc-macro2
A drop-in substitute for Rust's compiler proc_macro API that works anywhere and makes procedural macros unit-testable.
Repository Health
Technical Analysis
proc-macro2 is a wrapper around the Rust compiler’s built-in proc_macro crate. Where proc_macro types can only ever exist inside a running procedural macro, proc-macro2 mirrors the same TokenStream, TokenTree, Ident, Literal, Punct, Group, and Span API in types that work in any context: build scripts, regular binaries, ordinary libraries, and unit tests.
It is the foundation the modern Rust macro ecosystem is built on. Libraries like syn and quote target proc-macro2 rather than proc_macro so that token-based tooling can be developed, reused, and tested outside the narrow confines of an actual macro invocation. Inside a real proc macro it transparently delegates to the compiler; everywhere else it falls back to a self-contained pure-Rust implementation.
What You Get
- A complete mirror of the compiler’s
proc_macrotypes — TokenStream, TokenTree, Group, Ident, Punct, Literal, and Span — usable outside a proc macro. - Zero-cost delegation to the real compiler API when running inside a procedural macro, and a pure-Rust fallback lexer/parser everywhere else.
- The ability to unit-test macro helper logic in isolation, since proc-macro2 values can be constructed and inspected in ordinary #[test] functions.
- An optional
span-locationsfeature exposing line/column source positions for better diagnostics. - A
no_std-compatible core with a single lightweight dependency (unicode-ident).
Common Use Cases
- Writing custom derive, attribute, and function-like procedural macros, typically alongside syn and quote.
- Building reusable token-processing libraries that need to run in build.rs or plain application code, not just inside a macro.
- Unit-testing the internal logic of a procedural macro without a full compiler macro-expansion harness.
- Generating Rust source code programmatically from build scripts and codegen tools.
Under The Hood
Architecture
proc-macro2 is structured as a thin public facade over two interchangeable backends selected at runtime. src/lib.rs (~51KB) defines the public TokenStream, TokenTree, Group, Ident, Punct, Literal, and Span types. src/detection.rs decides, once and cached, whether the real compiler proc_macro API is available; src/wrapper.rs (~31KB) bridges to it when so, while src/fallback.rs (~35KB) provides a complete standalone implementation for build.rs, binaries, and tests. The fallback path parses source text via a hand-written lexer in src/parse.rs (~29KB), with src/rustc_literal_escaper.rs handling literal escape semantics and src/rcvec.rs providing a reference-counted vector for cheap token-stream cloning. A build.rs probe compiles feature-detection snippets to gate unstable compiler APIs.
Tech Stack
Pure Rust, edition 2021, with a minimum supported Rust version of 1.71. The only runtime dependency is unicode-ident for identifier validation; the crate is #![no_std] at its core. Cargo features gate behavior: proc-macro (default) enables the compiler bridge, span-locations exposes line/column data, and a procmacro2_semver_exempt cfg opts into nightly-tracking APIs. Dev-dependencies (quote, flate2, tar, rayon, rustversion) drive the test and benchmark suites.
Code Quality
Maturity shows throughout: deny(unsafe_op_in_unsafe_fn), an extensive curated clippy allow-list, and a broad tests/ suite (test.rs, comments.rs, features.rs, marker.rs, test_fmt.rs, test_size.rs) plus a tests/ui trybuild directory, a fuzz/ target, and benches/. The dual-backend design is carefully partitioned so the two implementations stay behaviorally identical, and the code is authored and maintained primarily by David Tolnay with 1,288 commits over eight years.
API Design
The defining design choice is deliberate one-to-one fidelity with the standard library’s proc_macro API, so migrating code between the two is largely a matter of swapping the import path. Types implement the expected Display, FromStr, Clone, and iterator conversions, making token streams ergonomic to build and inspect. Documentation is thorough via docs.rs with runnable examples, and the two-line Cargo.toml setup gets users started immediately.
Used by 27 apps in this directory
OpenShell
AI Agents · Developer Tools
The safe, private runtime that lets autonomous AI agents operate in sandboxed environments governed by declarative YAML policies — blocking data exfiltration, credential leaks, and unauthorized network activity before they happen.
ParadeDB
Analytics · Databases · Search
Born out of Y Combinator's S2023 batch, ParadeDB is a Postgres extension that delivers Elasticsearch-quality BM25 search and real-time analytics without a separate search cluster to manage.
Penpot
Design Tools
Open-source design platform with GPU-accelerated canvas, native design tokens, and MCP-powered AI workflows for teams that ship fast.
PostgresML
AI Development · Databases
Run ML training and LLM inference natively inside PostgreSQL with GPU acceleration — no data movement required.
Qdrant
AI Development · Databases · Search
Open-source vector database and search engine built in Rust for production-grade AI applications — from semantic search to RAG pipelines and recommendation systems.
Spacedrive
Collaboration · File Storage
One file manager for all your devices and clouds — powered by a Virtual Distributed File System built in Rust.
Stalwart
Collaboration
All-in-one secure mail and collaboration server covering IMAP, JMAP, SMTP, CalDAV, CardDAV, and WebDAV in a single memory-safe Rust binary.
Svix
Automation · Developer Tools
Open source, self-hostable webhook infrastructure that handles delivery, retries, HMAC signing, and multi-tenant event management so you never have to build a webhooks system from scratch.
TensorZero
Ab Testing Experimentation · AI Development · Monitoring
TensorZero unifies the LLM gateway, observability, evaluation, optimization, and experimentation stack behind a single OpenAI-compatible API, built in Rust for sub-millisecond p99 latency.