Repository navigation
feat!: redesign TypedFlake around an attribute macro - #5
Merged
Merged
Conversation
Replace the declarative `typedflake::id!()` macro with `#[derive(TypedFlake)]` on user-defined structs, enabling users to freely add their own derives and attributes alongside TypedFlake. New workspace structure: - typedflake: public facade re-exporting core + derive macro - typedflake-core: runtime logic (config, context, generator, global, state) - typedflake-macros: proc-macro crate implementing #[derive(TypedFlake)] BREAKING CHANGE: `typedflake::id!(Name)` replaced by `#[derive(TypedFlake)] pub struct Name(u64)` and `typedflake::id!(Name, CONFIG)` replaced by `#[derive(TypedFlake)] #[typedflake(config = CONFIG)] pub struct Name(u64)`
…e Config API Remove unverifiable TWITTER/DISCORD/INSTAGRAM presets from Epoch, BitLayout, and Config. Rename Config::new_unchecked to Config::new since it's already const-validated via panic. Add inline #[typedflake(layout, epoch)] attribute parameters to the derive macro for one-off configuration without separate const declarations. BREAKING CHANGE: Config::new_unchecked renamed to Config::new. Removed Config::TWITTER, Config::DISCORD, BitLayout::TWITTER, BitLayout::DISCORD, Epoch::TWITTER, Epoch::DISCORD, Epoch::INSTAGRAM.
Replace verbose free functions (set_defaults, set_default_config, set_default_instance) with a DefaultsBuilder accessed via defaults(). New API: typedflake::defaults().config(...).instance(w, p).init()?
Record the target interface, decisions, legacy ID compatibility recipe, and implementation phases for the attribute-macro redesign.
Replace `#[derive(TypedFlake)]` with an attribute macro on `i64`/`u64`
newtypes, as described in PROPOSAL.md.
- Formats are `Format { epoch, bits }` constants or inline options, with
one node field instead of worker and process, validated at compile time.
- Nodes are a plain `u32` or a `TypedNode` struct with named fields.
- Generation returns errors instead of blocking; `generate_blocking` and
the optional Tokio `generate_async` wait for sequence capacity.
- `typedflake::init` installs one schema-preserving default node.
- Raw, string, Serde, SQLx, and postgres-types inputs share one
validation path that rejects negative values and reserved bits.
- Merge `typedflake-core` into `typedflake`.
The clock is read after each load of the generation state, so a newer
millisecond stored by another thread is not reported as a rollback.
- Rewrite the README around `#[typedflake]`, with examples compiled as doctests. - Add MIGRATION.md, including declarations that keep reading 0.1 IDs. - Record the changes in CHANGELOG.md and the implementation notes in PROPOSAL.md. - Update CLAUDE.md and AGENTS.md for the two-crate layout. - Run CI against PostgreSQL, each feature alone, and rustdoc.
Panic in the declaration's own const instead of inside a called function. The compile error now points at the `#[typedflake]` attribute with only the reason, and no longer quotes standard library sources, which made the pinned diagnostics depend on `rust-src` being installed. Move the two remaining source-dependent rejections to `compile_fail` doctests.
`typedflake::init` can now be called once for each node type, so a process
that mixes plain and typed nodes keeps the static `generate()` for every
ID. Previously only one node type could have a default and the others
needed explicit generators.
An ID still only receives a node of its own type. A missing default is
`GenerateError::NotInitialized { node }`, which replaces
`NodeSchemaMismatch`.
Database tests also treat an empty `TYPEDFLAKE_TEST_POSTGRES_URL` as unset.
Reading the clock between the load and the compare-and-swap left a window of about 30 ns in which other threads changed the state, so the swap failed almost every time under contention. A rollback is now confirmed with a second reading taken after the load before it is reported.
The `alphabet` option of `#[typedflake]` adds `encode()` and `decode()` for a fixed-length text form. Alphabets hold 2 to 94 ASCII characters and are validated at compile time; presets are BASE36, BASE58, BASE62, and BASE64_URL. Decoding accepts only the exact encoding and ends in the same validation as every other constructor.
`typedflake::SerdeEncoded` serializes an ID as its encoded text and reads only that exact text, since a decimal string can also be valid text in the alphabet for a different ID. It needs the `alphabet` option and cannot be combined with `typedflake::Serde` on one type.
It only imported AGENTS.md, which is the single contributor guide.
pokedotdev
marked this pull request as ready for review
October 5, 2026 06:09
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Redesigns TypedFlake for 0.2.0. See CHANGELOG.md for the changes and MIGRATION.md for moving from 0.1.