fix: ignore .ts and .js extensions form imports

[nohehf]

Simple quick fix for #435, this can be evolved further but I believe will work in most currently broken cases.

[hendrikvanantwerpen]

I think this is the right approach, but the regex seems off. See my comments below.

[hendrikvanantwerpen]

I think the idea is correct, bit this doesn't look quite right. What about this?

Suggested change

	"([^/|(\.j|ts)]+)/?" {
	"([^/]+)(/|\.ts|\.tsx|\.js|\.jsx\)?" {

[nohehf]

Actually my regex is a bit of a hack but as .js (and .ts, ...) are included in this broad pattern: ([^/]+) it will be included in the first group if this is optional: (/|\.ts|\.tsx|\.js|\.jsx\) (eg. it will still match the extention). Working on something cleaner.

[nohehf]

adding lazy matching with ? in the first group seems to fix the issue:
^([^/]+?)(\.js|\.ts|\.jsx|\.tsx)?$seems ok, updating the pr

[nohehf]

It seems the regex parser does not support non greedy matching ? (...?)

[hendrikvanantwerpen]

Ah, I see the problem! An alternative might be to keep the original regex, and strip any of the extensions off inside the body using (replace ...). That might be easier.

[nohehf]

I also tried a way using regexes and replace, but it seems the regex in tsg does not strictly behave as in rust:
Given the regex \.[^\.\\/]+$

scan "./bar" {
    "\.[^\.\\/]+$" {
      print $0
    }
 }

Outputs /bar
But using the same regex in rust it does not match (as expected):
https://regex101.com/r/Rwm8lq/1 (check the unit tests tab)
(also tested on https://rustexp.lpil.uk/)

[nohehf]

I believe the implementation of path-dir is kind of wrong:

functions.add(
    "path-dir".into(),
    path_fn(|p| p.parent().map(|s| s.as_os_str().to_os_string())),
);

For instance I expect the dir of . to be . and the dir of .. to be .., but their parents are both None.

[nohehf]

I got it to work by removing the explicit extensions (.js, .ts, .jsx, .tsx, not all extensions) with: (path-normalize (replace mod_path "\.(js|ts|jsx|tsx)$" "")), but the questions above still remain and should be considered I believe

[nohehf]

(tho this can be merged I believe)

[hendrikvanantwerpen]

My kingdom for a small coherent theory of paths 😭. Thanks for figuring out these cases. I'll put them in an issue so we document this in a more visible place.