Support partial patterns on web pages

[rhadamanthe]

Consider the following URL pattern: https?://(www\.)?toto\.com/media/[^"]+ It sometimes happen pages reference absolute ou relative links on the page. As an example, toto.com may have page that only contain /media/test.jpg or ../../media/test.jpg.

In such cases, HG ++ does not find anything. The search pattern becomes invalid.

---

There are 3 solutions to this.

1. Introduce a new property in the dictionary to specify the search pattern is restricted to a given host. Here, the restriction would be about the toto.com host, and the search pattern would become .(*/)?media/^"]+.

2. Implement a smart guess to deduce relative patterns from a global one. Depending on the search pattern, it might be complicated.

3. Rework the dictionary. Add a domain property. And rename the URL pattern to path pattern.

---

With the third option, it becomes much more simple to handle all these cases.

• In a given page, we build an URL pattern from both the domain and path pattern.
• If the current URL matches the domain, we can also directly search for the path domains.

Code modifications are not the most complicated. We will have to rewrite the dictionary.