API URL matching pattern is too rigid

[FabulousCupcake]

Unsure if I should file this here or kanasimi/CeJS; the issue seems to be this regexp pattern: https://github.com/kanasimi/CeJS/blob/521b966ccf2810455c9d89ec893f478f06d4575a/application/net/wiki/namespace.js#L399

Steps to reproduce

Simply try getting a page from a domain that does not have subdomain or domain extension, or is only an IP, or have a port number attached to it:

const Wikiapi = require("wikiapi");

(async () => {
  const url = 'https://gbf.wiki/api.php';
  // These will also fail
  // const url = 'http://localhost/api.php';
  // const url = 'http://127.0.0.1/api.php';
  // const url = 'https://www.example.com:8080/api.php';

  const wiki = new Wikiapi(url);
  const data = await wiki.page('Main_Page');
  console.log(data.wikitext);
})();

Expected

It just works

Actual

Get the following error:

api_URL: Unknown project: [https://gbf.wiki/api.php]! Using default API URL.

And it simply queries the default (en wikipedia) instead.

Workaround

I can still get it working by running a local reverse proxy with nginx and modifying /etc/hosts, but this is very non-ideal.

[hugolpz]

If both https:// and /api.php are present, continue.

[FabulousCupcake]

If both https:// and /api.php are present, continue.

I'm not sure what you're trying to convey; as shown in the example it does not work simply when https:// and /api.php are present.

To me, the pattern /^(https?:)?(?:\/\/)?(([a-z][a-z\d\-]{0,14})\.([a-z]+)+(?:\.[a-z]+)+)/i feels unnecessary and overly rigid; if you need to grab the parts of the url it's probably better to let the URL API do the job — and to check if it's a valid mediawiki api endpoint, it's probably best to just assume the url is correct, send some queries, and see if it fails.

[hugolpz]

I suggest the pattern should just check if http*:// and /api.php are present. But Kanasimi with find something smarter ; )

[FabulousCupcake]

Late reply but can confirm that it works now, thank you!