Automatic `cargo fmt` after `cargo fix`

[kpreid]

Problem

In a project which uses rustfmt, running cargo fix sometimes leaves the source files with non-canonical formatting (e.g. around removed imports), which can be surprising (especially if one is using editor format-on-save) and is an extra step to fix.

Proposed Solution

Add a configurable setting which causes cargo fix to also have the effect of cargo fmt after applying fixes.

This should be a setting, not solely a command-line flag, so that it is automatic and convenient for projects which choose to use rustfmt — contributors won't have to remember to type it.

It might make sense to frame the setting as a general “autoformat” rather than specific to cargo fix, in case any future commands also edit source code and should have the same behavior.

[ehuss]

The cargo team discussed this, and felt like it would be nice to explore this idea more, but would need some design work and community feedback to figure out how it would all fit together.

The general idea is that there would be some configuration somewhere that would indicate that the project uses rustfmt. Then commands like cargo fix could know to run rustfmt in that case.

There was some desire to keep formatting independent of cargo fix, and instead just have rustc generate properly formatted code. However, we noted that would likely not be possible without embedding rustfmt in rustc itself. There's also a lot of complexity, for example a fix that adds braces around a large block of code would need to reformat everything in-between.

Some questions:

• What would that configuration look like, and where would it go? Does it need to be more general?
• Are there other commands that could use this information that rustfmt is being used?
• What should the interaction be with --allow-dirty be?
• Should cargo check if the file is already formatted before applying fixes? If it is unformatted, what should it do? Should there be a separate CLI flag like --allow-dirty to force it to apply formatting (or just piggyback on --allow-dirty?).

[epage]

Started an Internals thread