Fixes #5613 - Generated C++ code issue with keyword "delete"

[Qubi0-0]

• lazy_static crate to slint-compiler added.
  I added lazy static hashset of all cpp_keywords. Now every ident is checked for potential collidations. If they happen, then _ is added.

[ogoffart]

Thanks for this PR.

Some feedback:

• lazy_static is kind of being phased out and being replaced by std::sync::OnceLock. See https://github.com/rust-lang-nursery/lazy-static.rs?tab=readme-ov-file#standard-library
  You can put the CPP_KEYWORDS OnceLock directly in the is_cpp_keyword function
• This adds a cpp_keywords.json, but this is not usefull, right?
• keywrods needs to be checked after normalizing for -. For example there can be an identifier static-assert
• regarding the test, the files in tests/syntax are not being generated into C++. Only the ones in tests/cases
  An example of test that can be extended with C++ keywords: https://github.com/slint-ui/slint/blob/master/tests/cases/types/structs_keyword.slint
• as the CI points out, this currently break if you have a property of function that is a keyword (example, do -> invoke_do) looks like the ident function should be called on the whole name rather than on part of the identifier.

Thanks a lot for your patch. Do you think you can do the changes?

[Qubi0-0]

Hi.
I will gladly try to fix the Code! I will get on that today.

[Qubi0-0]

I see there's still some issues. I'm not sure what should be done to resolve (as the CI points out, this currently break if you have a property of function that is a keyword (example, do -> invoke_do) looks like the ident function should be called on the whole name rather than on part of the identifier.). Could you elaborate on that please?
Also i have a question about testing. Is there a way to do this locally? I don't want to use your resources to much 😅 (unless it's irrelevant)

[ogoffart]

Basically, the problem is that we use the output of the ident function and concatenate.
For example there:

slint/internal/compiler/generator/cpp.rs

Line 760 in 054035d

name: format!("global_{}", ident(&glob.name)),

slint/internal/compiler/generator/cpp.rs

Line 1095 in 054035d

"{0}if constexpr(std::is_same_v<T, {1}>) {{ return *m_globals.global_{1}.get(); }}",

slint/internal/compiler/generator/cpp.rs

Line 2565 in 054035d

name: format!("invoke_{prop_ident}"),

slint/internal/compiler/generator/cpp.rs

Line 2582 in 054035d

name: format!("on_{}", ident(&p.name)),

slint/internal/compiler/generator/cpp.rs

Line 2696 in 054035d

name: format!("set_{}", &prop_ident),

(and maybe other places)

In these cases, we should either have a different function to escape the - without replacing the keyword (eg: escape), or pass the whole production to the indent function.

[ogoffart]

you can also try to add more stuff such as

struct struct {
   delete: image;
}

export global const {
   in-out property<struct> return;
}

[Qubi0-0]

I added some in new commits. Although in case of for example

 export global const{ 
    in-out property<unsigned> atomic_commit;
    in-out property<namespace> return;  
 }

there is

running 1 test
test test_cpp_types_structs_keyword ... FAILED

failures:

---- test_cpp_types_structs_keyword stdout ----
/tmp/.tmpn7hGBG.cpp: In member function ‘const T& char_::global() const’:
/tmp/.tmpn7hGBG.cpp:293:41: error: ISO C++ forbids declaration of ‘type name’ with no type [-fpermissive]
 293 |     else if constexpr(std::is_same_v<T, const>) { return *m_globals.global_const.get(); }
     |                                         ^~~~~
thread 'test_cpp_types_structs_keyword' panicked at /home/qubi/Projects/slint/target/debug/build/test-driver-cpp-815659352fb91da7/out/test_functions.rs:3100:20:
called `Result::unwrap()` on an `Err` value: "C++ Compilation error (see stdout)"
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

I'm not sure if this should be addressed or not.

[Qubi0-0]

A scenario with concatenated names is resolved with a new function that does what the previous Ident did.
However I see new issues with tests/cases/types/structs_keyword.slint. Im not sure how to address that.

[ogoffart]

The problem is that the test was old and relied on the old syntax that generated the last component. But now there are more components added in the test, so the original match component is no longer generated.
The solution is to keep the match component the last component in the test. (so move the new struct and component before it)

[Qubi0-0]

I changed tests/cases/types/structs_keyword.slint. Now it passes the tests.
I see that the last failing test is about formatting. I know of cargo fmt, although it makes a whole long list of all hashtset entries being put one for each line. What should be done here?

[Qubi0-0]

Thanks for all the help from the beginning. It`s my first opensource PR, so it means a lot to me :D

[Qubi0-0]

Seems everything is in order now :D

[ogoffart]

Thanks a lot for this PR and fixing all the issues with it!