Re: Extending whitespace checks

["Bence Ferdinandy" <bence@ferdinandy.com>]

On Sun Nov 24, 2024 at 03:25, Junio C Hamano <gitster@pobox.com> wrote:
> We have, via the attributes subsystem, a way to choose from a set of
> predefined whitespace rules so that "git diff" can notice that you
> are adding trailing whitespaces to your newly written lines, or you
> are indenting a newly introduced line in a Python script with a HT.
> This can be used, for example, in pre-commit hook to reject an
> attempt to introduce whitespace-damaging changes to the codebase.
>
> Which is great.
>
> I am wondering what we can do to add a different kind of checks to
> help file types with fixed format by extending the same mechanism,
> or the checks I have in mind are too different from the whitespace
> checks and shoehorning it into the existing mechanism does not make
> sense.  The particular check I have an immediate need for is for a
> filetype with lines, each has exactly 4 fields separated with HT in
> between, so the check would ask "does each line have exactly 3 HT on
> it?"  It would be extended to verify CSV files with fixed number of
> fields (but the validator needs to be aware of the quoting rules for
> comma in a value in fields).
>
> I guess the best I could do (outside Git) is
>
>  - write such a validator that can take one line of input and say
>    "this line comforms to the rule".
>
>  - add, via .gitattribute, my own attribute to allow me to mark
>    the files that these rules apply.  Git does not do anything
>    special for this attribute (remember, I said "outside Git").
>
>  - in pre-commit hook, run "git diff ':(attr:myattr)'" to grab
>    changes in these files with special formats, and have the
>    line-by-line validator (above) check the new lines.
>
> to make sure bad lines would not slip into the history, but it would
> be really nice if I can trigger the check as part of "git diff --check",
> which means it would be more ideal if we can do this "inside" Git.
>
> Perhaps we could introduce a mechansim that allows me to do the
> following:
>
>  - An attribute, like whitespace=..., specifies what line-validation
>    function to use to vet each new line introduced to a file with
>    the attribute.
>
>  - A line-validation function can be dynamically loaded/linked
>    (here, we'd need ".gitattribute specifies the logical meaning,
>    while .git/config and friends maps the 'logical meaning' to a
>    specific implementation suitable for the platform" separation,
>    similar to what we use for smudge/clean filters).  Perhaps this
>    would be a good testbed for use of dll, written even in a foreign
>    language like Rust?
>
>  - In the diff machinery, where a '+' line is checked for whitespace
>    anomalies in the existing code, add code to call the dynamically
>    loaded line-validation function when applicable.
>
>  - Profit?
>
> Hmm?

This might be a tangent, but since enhancing whitespace checking was mentioned,
I'd thought I note here:  `git log --check` running in the CI did not catch the
white space errors in this patch (see the last hunk):

https://lore.kernel.org/git/20241121225757.3877852-4-bence@ferdinandy.com/

although it would have been certainly nice. I'm not sure if --check could
already catch this actually, or if it would be easy/possible to have something
general enough that does catch it.


Best,
Bence