Comments (9)
@Kobzol syn
does parse self
as an Expr
:
The problem here is that it is parsing self { ... }
as an expression as a whole (e.g., self { 0: hi }
is, syntactically, a valid expression: Demo), and so it commits to start parsing a struct
-like expression and fails when it encounters the token fn
instead of a field name or a tuple index.
In general, you cannot really parse a braced group right after arbitrary expressions. Try adding some separator (such as ,
or =>
) to disambiguate. For instance:
impl Trait for Struct {
delegate! {
{
fn foo ...
} => self,
}
Otherwise you would have to do the parsing of the part before the braces manually (e.g., with Punctuated<IdentOrInteger, Token![.]>::parse_separated_non_empty
or something like that, given some enum IdentOrInteger
that impl
s Parse
).
@Boscop in the meantime, you can use (self)
or {self}
to disambiguate 🙂
from rust-delegate.
@lulivi Published rust-delegate
0.5.0
with #26.
from rust-delegate.
Indeed, the self
that the macro introduces is internal / hidden, so the caller of some_macro!
has no way to refer to it / "cannot know what that self
refers to" (the whole point of hygiene was to avoid such confusing situations).
A possible fix is to use:
macro_rules! some_macro {
(|$self:ident| $delegate_to:expr) => {
impl Trait for Struct2 {
delegate! {
// '$delegate' will expand to 'self.0' before ´delegate!' is expanded.
to $delegate_to {
fn method_to_delegate(&$self) -> bool;
}
}
}
};
}
some_macro! { |self| self.0 }
- (another option, since I agree that
self
is kind of special (e.g., within the context ofdelegate!
it is not surprising at all), is to use aVisitMut
fromsyn
to map anyIdent
equal toself
to the same ident, but withSpan::call_site()
(removing any hygiene interactions when dealing withself
s)
from rust-delegate.
Hmm, for some reason syn
refuses to parse self
as an Expr
. This is pretty tricky, I'll probably have to do through the tokens one by one and build the target expression manually if I can't just parse it as an expression.
from rust-delegate.
Ooh, now it actually makes sense, thank you! I'll probably parse it manually, if it won't be too problematic, because I don't want to introduce another syntactic layer.
from rust-delegate.
So I decided to special case this, because self
is pretty much the only bare token that makes sense here (delegating to a global variable or parameters of the delegated method is not currently supported). It works fine for normal cases, however there is some problem with macros generating the delegation tokens.
It seems that rustc inserts None
-delimited groups around macro tokens (media-io/yaserde#79), so this test generates something like this:
Group {
delimiter: None,
stream: TokenStream [
Ident {
ident: "self",
span: #3 bytes(659..671),
},
Punct {
ch: '.',
spacing: Alone,
span: #3 bytes(659..671),
},
Literal {
kind: Integer,
symbol: "0",
suffix: None,
span: #3 bytes(659..671),
},
],
span: #3 bytes(659..671),
},
... rest of input
Before, when I was parsing this as syn::Expr
, it worked fine, so I tried to parse it as an Expr
and if it doesn't work, switch to self
, something like this:
impl syn::parse::Parse for DelegationTarget {
fn parse(input: ParseStream) -> Result<Self, Error> {
let cursor = input.cursor().token_stream();
Ok(if syn::parse2::<syn::Expr>(cursor).is_ok() {
DelegationTarget::Expr(input.parse()?)
} else {
DelegationTarget::SelfTarget(input.parse()?)
})
}
}
The problem is that the first condition doesn't succeed. I tracked it down to this:
syn::parse2::<syn::Expr>(input.cursor().token_stream()).is_ok() // false
input.parse::<syn::Expr>.is_ok() // true
I printed the contents of the two token streams and they seem to be the same, so it seems like there is something wrong with my usage of the parse2
function. This is hitting the limits of my syn
-fu, as I have no idea why should this return a different result.
Maybe it's caused by this: https://github.com/dtolnay/syn/blob/master/src/parse.rs#L1173? Could it be avoided somehow?
@danielhenrymantilla any ideas? :)
from rust-delegate.
Sorry for my late answer, I forgot about this thread.
The good thing is the results you are hitting make sense in my head; and I like the idea of using a None
-delimited group as a workaround. I'm gonna be using ⟨
and ⟩
as sigils to represent the delimiters of None
-delimited groups.
The key thing to understand about such ⟨…⟩
groups is:
-
they group the contained tokens as a single one, much like
()
,{}
and[]
do, -
but they don't show up / are invisible to operations such as stringification, hence your:
I printed the contents of the two token streams and they seem to be the same
So, in your case, your sequence of tokens is:
⟨self . 0⟩ { … }
Then, the difference between:
syn::parse2::<syn::Expr>(input.cursor().token_stream()).is_ok() // false input.parse::<syn::Expr>().is_ok() // true
is that the former tries to parse all the given tokens as an expression, whereas the latter only tries to parse a prefix of such tokens as one.
And given the capture, ⟨self . 0⟩
is an expression (self.0
is), but <expr> { ... }
on the other hand, is not a valid expression (as I mentioned, the ambiguity originated from self
being interpreted as a path, which is no longer the case with our ⟨self . 0⟩
captured expression).
The solution
The goot thing about having waited that long to keep this thread updated, is that, in the meantime, I have stumbled upon a bunch of posts from dtolnay where they mention that this is a common case of ambiguity in the language.
Because of that, I have suspected there had to be one function inside syn
to disambiguate. And indeed there, which does contain the examples I had seen in such posts. Behold:
Expr::parse_without_eager_brace()
I'm pretty sure this will neatly solve your issues 😉
from rust-delegate.
Thank you! That is exactly what I needed. Actually if I thought of using Expr::parse(input)
instead of input.parse<syn::Expr>
I would probably find it immediately 🤦♂️
This test broke:
macro_rules! some_macro {
($delegate_to:expr) => {
impl Trait for Struct2 {
delegate! {
// '$delegate' will expand to 'self.0' before ´delegate!' is expanded.
to $delegate_to {
fn method_to_delegate(&self) -> bool;
}
}
}
};
}
some_macro! {self.0}
It produces
error[E0424]: expected value, found module `self`
--> tests/in_macro_expansion.rs:34:14
|
28 | fn method_to_delegate(&self) -> bool;
| -- this function has a `self` parameter, but a macro invocation can only access identifiers it receives from parameters
...
34 | some_macro! {self.0}
| ^^^^ `self` value is a keyword only available in methods with a `self` parameter
Initially I thought that it's a regression of the new self
parsing, but after bisecting it I realized that Rust 1.46 introduced this change (in macro hygiene?), as it also happens with the old code. Probably caused by this: rust-lang/rust#73293. Not sure if we can work around that easily.
@Boscop can you please check that #26 fixes your issue?
from rust-delegate.
Looking forward for the merge of #26 ! 😄
from rust-delegate.
Related Issues (20)
- Arbitrary inner field expressions HOT 2
- is it possible to delegate all methods? HOT 2
- 0.5.0 fails to compile
- Delegate with additional arguments
- Transfer ownership! HOT 3
- Bridging sync to async gap with delegate HOT 5
- Allow #[from] and #[into] annotations, pattern-matching or even arbitrary transformation functions for delegated arguments HOT 10
- allow #[try_from] and #[try_into] HOT 8
- Is delegating to enum in the target range HOT 4
- Compatability with async_trait HOT 2
- Just unwrap HOT 3
- Assign Return Value Of Delegated Calls To Delegated Field HOT 7
- Add attribute to the return value of all methods HOT 4
- Add support support for qualified calls HOT 6
- Delegate to Self HOT 1
- transform the return value of `Result` and `Option` type HOT 8
- New release? HOT 1
- Change default inlining mode
- `no_std` support HOT 3
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from rust-delegate.