RFE: ignore build system input .in extension when determining the type

[scop]

Would be good to strip known template/backup filename extensions before attempting a match against a filename/extension glob.

Here's an implementation of that I did some time ago for bat: sharkdp/bat#1687

[epage]

Template files are a weird case because they contain multiple languages. Which language should we check?

For backup files, I assume we would instead want to ignore all of them.

[scop]

Right, ignoring backups would make sense for typos.

Regarding templates, the case I'm mostly interested in is the .in generic build system input one; for that I think the right thing to do is to strip the suffix and decide based on what's left in the name after that. I don't remember coming across one that would actually have multiple languages in it, but just some placeholders to be replaced.

For this specific case, using appropriate type specific extend-globs like

[type.sh]
extend-glob = ["*.sh.in"]

...does seem to serve as a workaround for the time being.

[epage]

I don't remember coming across one that would actually have multiple languages in it, but just some placeholders to be replaced.

I was referring to when you have a template language mixed in with the generated output.

You could say php/html is an example of this. See also liquid, tera, ninja, handlebars, etc.

[epage]

Another aspect of backups is that hopefully they are gitignored (in the per-user ignore file) which typos will then ignore by defalt.

[scop]

Right, agreed on both, and the PHP example is an interesting one.

Anyway as far as I'm concerned, this RFE reduces to just stripping the literal trailling .in suffix (recursively, as at least *.in.in are a pattern in use in some cases too) before attempting filename based type finding.

[epage]

I'd be fine with just stripping .in