Re: [PATCH v2] feat: add Elixir to supported userdiff languages

From: Łukasz Niemier <lukasz@niemier.pl>
Date: Nov 6, 2019, 22:07 UTC
Message-ID: <ED332708-76D5-447F-A2D7-6495CC6A76DC@niemier.pl>
In-Reply-To: <672bdea4-180a-89e0-c420-df2704ed887e@kdbg.org>

Show 5 quoted lines

> Please use "subsystem: short description" in the subject. For example:
> 
>   userdiff: support Elixir
> 
> would be sufficient in this case.

Ok, will fix.

> Please add your sign-off before the three-dash line so that we know that
> you are entitled to publish this patch. See Documentation/SubmittingPatches.

Yeah, I have seen it too late to fix. Will do.

> It would be enlightening to know what Elixir is. (I haven't googled it,
> yet.)

It is language with Ruby-like syntax for BEAM (Erlang virtual machine).

https://elixir-lang.org

> If it were a popular language, I think I would have heard about
> it. But it may well be possible that I have lived under a rock for too
> long... ;)

It is quite popular among few services (Pinterest, Discord, Bleacher Report), but it is still pretty new, and still need to get a little more spotlight.

> This list is sorted, basically, but your addition perturbates the order.

Oh, sorry, I missed dts.

Show 16 quoted lines

> The default hunk header pattern picks up lines that begin with a letter
> without leading whitespace. The tests that you present here do not show
> that the language specific hunk header pattern is better. The default
> would have picked up the correct lines. And, in fact, when I remove the
> pattern from the code, these tests still pass!
> 
> I'm not saying that the pattern is bad; I say that the tests do not show
> its worthiness. More tests are needed. For example:
> 
> --- 8< ---
> defmodule RIGHT do
> end
> #
> #
> # ChangeMe; do not pick up 'end' line
> --- 8< —

Yeah, I will provide such

> BTW, I guess that any def, defmodule, etc. as the first word on a line
> in the docstring would be picked up incorrectly. Is that a problem?

No, as this would (almost?) always be the definition of the module/function, so it is not a problem at all.

>> +	 "|:'a-zA-Z0-9@_]+'"
> 
> The opening bracket is missing here.

Whoops.

Show 7 quoted lines

> Would it be an option to collapse all but the first pattern (because I
> do not want to start the pattern with an optional part) to
> 
> 	"[:@]['\"]?[a-zA-Z0-9@_]"
> 
> This assumes that @"x1 and @'y2 cannot occur in a syntactically valid
> program.

No, these aren’t valid.

> Remember: the patterns can be loose; they do not have to
> validate the input, but can assume that it is syntactically valid.

Ok

> Does the language not have any two-character operators, such as '<=‚?

It has, I should add them as well (it even has 3 letter operators).

--

Łukasz Niemier lukasz@niemier.pl