feature request: have a wildcard domain matcher

[GlenDC]

Currently, the domain matchers (e.g., used in route_upstream) support only child_match and exact_match. However, there’s no support for wildcard matching. For example, something like:

foo*.site.bar.com

would ideally match:

foo.site.bar.com  
foo1.site.bar.com  
foofoo.site.bar.com

The exact syntax or feature set isn’t particularly important to me—what matters more is having the capability to perform wildcard-based filtering, especially in cases where exact_match or child_match aren’t sufficient.

I understand that wildcard matching may be slower by nature, but I still believe it could be quite useful for certain edge cases.

[zh-jq-b]

Yes it would be useful. It can use a radix_trie::Trie<u8, ArcEscaper> to do the match check. Can you submit a PR?

[GlenDC]

At the moment, I’m unfortunately swamped. If you haven’t had a chance to get to it by the time I need it, I’ll definitely take care of it myself.

That said, could you elaborate a bit more on how exactly you envision this? Your proposal still leaves quite a bit of room for interpretation, and I’d like to make sure we’re aligned before moving forward.

[zh-jq-b]

The wildcard_match code would be much like the child_match code, the difference is:

child_match insert example.net as net.example. to a Trie<String, ArcEscaper>,
wildcard_match insert a*.example.net as net.example.a to a Trie<String, ArcEscaper>.

*.example.net seems to be a special case as it should match both example.net and a.example.net.

[zh-jq]

I have added 'regex match' support in this PR #667. I think this can be used to do a wildcard match.

[GlenDC]

Seems to work. However

g3/g3proxy/src/config/escaper/route_upstream/regex_match.rs

Line 207 in 622c4b1

let prefix = domain.split_at(domain.len() - prefix_len).0;

currently can panic in case prefix length is larger than domain length.

Example regex used in escaper that can trigger it:

regex_match:
  - next: forward-protoparser
    rules:
      - parent: foobar.com
        regex: abcdfoobaserver-dk\.clients\d+

output:

thread 'main#13' panicked at g3proxy/src/config/escaper/route_upstream/regex_match.rs:207:46:
attempt to subtract with overflow

That said it does work and the regex correctly uses the "next" escaper.

[zh-jq]

the panic will be fixed by #671

[GlenDC]

Panic seems to be fixed indeed. I'm okay with closing this issue as done, as it seems to be implemented correctly as far as I can see?