-
Notifications
You must be signed in to change notification settings - Fork 12.7k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
resolve: Modularize crate-local #[macro_export] macro_rules
#52234
Conversation
r? @cramertj (rust_highfive has picked a reviewer for you, use r? to override) |
@bors try |
[Experiment] resolve: Modularize crate-local `#[macro_export] macro_rules` Based on #50911, cc #50911 (comment) `#[macro_export] macro_rules` items are collected from the whole crate and are planted into the root module as items, so the external view of the crate is symmetric with its internal view and something like `$crate::my_macro` where `my_macro` is `#[macro_export] macro_rules` works both locally and from other crates.
ping @pietroalbini |
The job Click to expand the log.
I'm a bot! I can only do what humans tell me to, so if this was not helpful or you have suggestions for improvements, please ping or otherwise contact |
💔 Test failed - status-travis |
The job Click to expand the log.
I'm a bot! I can only do what humans tell me to, so if this was not helpful or you have suggestions for improvements, please ping or otherwise contact |
1 similar comment
The job Click to expand the log.
I'm a bot! I can only do what humans tell me to, so if this was not helpful or you have suggestions for improvements, please ping or otherwise contact |
Is it necessary to fix the std doc generation mode ( |
I don't think std docs needs to be working to get a crater run, but a successful try build is needed. |
@bors try |
⌛ Trying commit 1f844a847d62ded4322840a4ddbb499e0228cd48 with merge d39a6f7eddafd45dcc60e99001b8cd08c2bd3255... |
The job Click to expand the log.
I'm a bot! I can only do what humans tell me to, so if this was not helpful or you have suggestions for improvements, please ping or otherwise contact |
☀️ Test successful - status-travis |
@craterbot run start=master#ae5b629efd79de78e6ba7ef493c32857bd7f9cf9 end=try#d39a6f7eddafd45dcc60e99001b8cd08c2bd3255 mode=check-only |
👌 Experiment ℹ️ Crater is a tool to run experiments across the whole Rust ecosystem. Learn more |
🚧 Experiment ℹ️ Crater is a tool to run experiments across the whole Rust ecosystem. Learn more |
🎉 Experiment
|
|
@CAD97 The issue is that any Indeterminacies like this cause import resolution to stuck, like it happened in #53144 and similar cases. Perhaps the error can be relaxed a bit / made more fine-grained, but I haven't thought how to do that in detail. |
Hello, I've got a bit of a question, as I've just run into this error during implementation. Since referral by absolute paths is an error, how is the external use of a macro-generated macro supposed to occur? |
@Kixiron |
The error mentions "by absolute paths", but if in the (failing) test case 3:
one changes to a definitely-relative path like Furthermore, if one encloses the macros inside a module:
Then the compiler suggests using |
I don't think I'm understanding how I'm meant to work around this. Do I really need to |
Indeed, this seems like a huge problem for hygine. I was writing a macro that needed to expand a helper macro and I found it incredibly problematic, as the fcw triggered when adding |
Regarding Regarding a more general workaround, the trick is to "append a
@petrochenkov IIUC, if the issue with |
Using the re-exporting trick from rust-lang/rust#52234 (comment) to reference the macro recursively at a known path. This way any external callers don't have to import the macro or its dependencies.
Using the re-exporting trick from rust-lang/rust#52234 (comment) to reference the macro recursively at a known path. This way any external callers don't have to import the macro or its dependencies.
Using the re-exporting trick from rust-lang/rust#52234 (comment) to reference the macro recursively at a known path. This way any external callers don't have to import the macro or its dependencies.
For anyone stumbling here, I fixed it using like the following, #[macro_export]
macro_rules! impl_str_id {
($t:ident) => {
::paste::paste! {
#[derive(Clone, Copy, Hash, Deref)]
#[display("{0}")]
pub struct $t(TinyStr16);
impl_new_type_debug_no_wrap!($t);
#[macro_export]
macro_rules! [<_ $t:snake:lower>] {
($s:literal) => {{ $t::new(tinystr::tinystr!(16, $s)) }}
}
// Fixes macro-expanded `macro_export` macros from the current crate cannot be referred to by absolute paths
#[doc(hidden)]
pub use [<_ $t:snake:lower>] as [<$t:snake:lower>]; |
Regarding #[macro_export]
macro_rules! recur {
(()) => {$crate::recur!{}};
() => ()
} However, this strategy requires accessing the macro by an absolute path so if the macro_rules! mk_recur {() => (
#[macro_export]
macro_rules! recur {
(()) => {$crate::recur!{}};
() => ()
}
pub use recur;
)}
pub mod m1 {
mk_recur!{}
}
mod m2 {
crate::m1::recur!{()}
} |
# Breaking changes ## Generated token macros are no longer `#[macro_export]` The `Token` derive macro generates a `macro_rules!` macro for each of the enum variants. E.g., for Lox's `Token` enum in `examples/lox`: ```rs // examples/lox/src/tokens.rs #[derive(DebugLispToken, PartialEq, Token, Lexer)] pub enum Token { #[subset_of(Ident)] #[pattern = "and|class|else|false|for|fun|if|nil|or|print|return|super|this|true|var|while"] Keyword(Substr, Span), #[pattern = "[a-zA-Z_][a-zA-Z0-9_]*"] Ident(Substr, Span), #[pattern = r"[(){}]"] Brace(Substr, Span), #[pattern = "[,.;]"] Punct(Substr, Span), #[pattern = "[=!<>]=?"] #[pattern = "[-+*/]"] Operator(Substr, Span), #[pattern = "[0-9]+"] NumLit(Substr, Span), #[pattern = r#""[^"]*""#] StrLit(Substr, Span), } ``` We get `macro_rules!` macros named `keyword`, `ident`, `brace`, `punct`, `operator`, `num_lit` and `str_lit`. These are mostly useful for the `Parse` implementations, e.g.: ```rs // examples/lox/src/decl.rs impl Parse for ClassDecl { type Stream = ParseStream; fn parse(input: &mut Self::Stream) -> Result<Self> { // ... let superclass = if input.check(operator![<]) { input.consume(operator![<])?; Some(input.consume_kind(TokenKind::Ident)?) } else { None }; // ... } } ``` Previously, those generated macros were declared like this: ```rs #( #[macro_export] macro_rules! #snake_case_variant_ident { // ... } )* ``` That has always [caused some headaches](rust-lang/rust#52234 (comment)) which I was never really sure how to deal with. In the examples here, and in my consuming projects, I had resorted to using `*` imports everywhere to work around the problem (in hindsight though, I think that likely just hides the lint by obscuring what we're doing instead of actually addressing the issue). This PR attempts an alternative solution. `#[macro_export]` is intended for libraries to expose macros to external consumers, but I don't foresee the macros generated by the `Token` derive being actually useful in that context. So instead, the generated macros are now declared like this: ```rs #( macro_rules! #snake_case_variant_ident { // ... } pub(crate) use #snake_case_variant_ident; )* ``` This is a breaking change for two reasons: 1. If you were depending on those `#[macro_export]` attributes to make the generated macros available to external consumers of your library, that is no longer going to work. Again, I don't imagine this was actually a real-world use case for anyone, but I've been wrong before! Let me know if this is a problem for you and I'll see what we can do about it. 2. If you had been importing those macros from the root of your crate, but your `Token` enum is _not_ declared in the crate root, you'll need to update your import paths to instead import them from the module where the enum is declared. E.g.: ```rs mod token { use gramatika::{DebugLisp, Span, Substr, Token as _}; #[derive(DebugLispToken, PartialEq, Token, Lexer)] pub enum Token { #[pattern = ".+"] Word(Substr, Span), } } mod foo { // use crate::word; // 👎 use crate::token::word; // 👍 } ``` On the bright side, tools like `rust-analyzer` should now find and automatically suggest the correct import paths for those macros, so the fastest way to migrate will probably be to just delete your existing `use` statement and invoke your editor's suggestion feature to re-import any unresolved symbols from their correct paths. # Other changes * Updated the Rust edition to 2021 and fixed any resulting errors * Fixed any new `clippy` lints as a result of upgrading my environment to v1.77.2 * Performed some low-hanging-fruit dependency upgrades. `regex-automata` and `syn` are still out of date for now — I attempted to update the former, but couldn't figure out how to migrate after ~10 minutes of poking around, and unfortunately I have other priorities that need to take precedence. Didn't even attempt `syn` because it's a major version bump, and that crate is basically the backbone of this whole project, so it'll have to wait for now.
This seems to also affect the |
So, if I have a macro |
@wdanilo given the usage you describe ( macro_rules! m1 {( ... ) => (
...
- #[macro_export]
macro_rules! m2 { ... }
+ pub(crate) use m2;
)} If you really need it to be
Important If you have any more usage questions, or if anybody else does, it's probably better to ask over the Rust forum (https://users.rust-lang.org), or over the community Discord (https://discord.gg/rust-lang-community), rather than commenting on this PR 🙂 |
Based on #50911, cc #50911 (comment)
#[macro_export] macro_rules
items are collected from the whole crate and are planted into the root module as items, so the external view of the crate is symmetric with its internal view and something like$crate::my_macro
wheremy_macro
is#[macro_export] macro_rules
works both locally and from other crates.Closes #52726