mirror of https://github.com/google/comprehensive-rust.git synced 2025-04-03 01:56:12 +02:00

Format all Markdown files with dprint (#1157 )

This is the result of running `dprint fmt` after removing `src/` from
the list of excluded directories.

This also reformats the Rust code: we might want to tweak this a bit in
the future since some of the changes removes the hand-formatting. Of
course, this formatting can be seen as a mis-feature, so maybe this is
good overall.

Thanks to mdbook-i18n-helpers 0.2, the POT file is nearly unchanged
after this, meaning that all existing translations remain valid! A few
messages were changed because of stray whitespace characters:

     msgid ""
     "Slices always borrow from another object. In this example, `a` has to remain "
    -"'alive' (in scope) for at least as long as our slice. "
    +"'alive' (in scope) for at least as long as our slice."
     msgstr ""

The formatting is enforced in CI and we will have to see how annoying
this is in practice for the many contributors. If it becomes annoying,
we should look into fixing dprint/check#11 so that `dprint` can annotate
the lines that need fixing directly, then I think we can consider more
strict formatting checks.

I added more customization to `rustfmt.toml`. This is to better emulate
the dense style used in the course:

- `max_width = 85` allows lines to take up the full width available in
our code blocks (when taking margins and the line numbers into account).
- `wrap_comments = true` ensures that we don't show very long comments
in the code examples. I edited some comments to shorten them and avoid
unnecessary line breaks — please trim other unnecessarily long comments
when you see them! Remember we're writing code for slides 😄
- `use_small_heuristics = "Max"` allows for things like struct literals
and if-statements to take up the full line width configured above.

The formatting settings apply to all our Rust code right now — I think
we could improve this with https://github.com/dprint/dprint/issues/711
which lets us add per-directory `dprint` configuration files. However,
the `inherit: true` setting is not yet implemented (as far as I can
tell), so a nested configuration file will have to copy most or all of
the top-level file.

2023-12-31 00:15:07 +01:00

1.6 KiB

Raw Blame History

CXX Error Handling: PNG Example

A prototype of a PNG decoder illustrates what can be done when the successful result cannot be passed across the FFI boundary:

#[cxx::bridge(namespace = "gfx::rust_bindings")]
mod ffi {
    extern "Rust" {
        /// This returns an FFI-friendly equivalent of `Result<PngReader<'a>,
        /// ()>`.
        fn new_png_reader<'a>(input: &'a [u8]) -> Box<ResultOfPngReader<'a>>;

        /// C++ bindings for the `crate::png::ResultOfPngReader` type.
        type ResultOfPngReader<'a>;
        fn is_err(self: &ResultOfPngReader) -> bool;
        fn unwrap_as_mut<'a, 'b>(
            self: &'b mut ResultOfPngReader<'a>,
        ) -> &'b mut PngReader<'a>;

        /// C++ bindings for the `crate::png::PngReader` type.
        type PngReader<'a>;
        fn height(self: &PngReader) -> u32;
        fn width(self: &PngReader) -> u32;
        fn read_rgba8(self: &mut PngReader, output: &mut [u8]) -> bool;
    }
}

PngReader and ResultOfPngReader are Rust types --- objects of these types cannot cross the FFI boundary without indirection of a Box<T>. We can't have an out_parameter: &mut PngReader, because CXX doesn't allow C++ to store Rust objects by value.

This example illustrates that even though CXX doesn't support arbitrary generics nor templates, we can still pass them across the FFI boundary by manually specializing / monomorphizing them into a non-generic type. In the example ResultOfPngReader is a non-generic type that forwards into appropriate methods of Result<T, E> (e.g. into is_err, unwrap, and/or as_mut).

1.6 KiB Raw Blame History

CXX Error Handling: PNG Example

1.6 KiB

Raw Blame History