Show capturing groups in --json mode

[pmkap]

I couldn't find any way to show capturing groups in --json mode.

Motivation:
Sometimes I'm interested in a group. After parsing the json, I need to filter the group in an additional step. Possibly with the same regex as in the first step. In such cases it would be nice if this was directly supported.

Example how this feature could look like:

echo 'Hello! !World! !foo!' | rg --json '!(\w+?)!' | jq

"submatches": [
  {
    "match": {
      "text": "!World!"
    },
    "groups" : [
      {
        "1": "World"
      }
    ]
  },
  {
    "match": {
      "text": "!foo!"
    },
    "groups" : [
      {
        "1": "foo"
      }
    ]
  }
]

[BurntSushi]

Related #1872

Could you please provide an end-to-end use case where you'd want this? My suspicion is that it isn't necessary. The other issue here is that resolving capturing groups can be slow, so it would need to be behind a flag, i.e., --json-captures.

[pmkap]

Thank you for your help!

Like you suspected, I actually did find a solution (without the --json flag).

My use case was the following:
I am using Logseq for taking notes. It has pages that can reference each other with either [[ref]] of #ref. The idea that I'm playing around with is to search a directory for all those references with rg and hook the results into vim's completion.

The solution I now have is rg -o -e '\[\[(?P<g1>.+?)\]\]' -e '#(?P<g2>[^\s#]+)' -r '$g1$g2'

I stumbled upon the --json flag in the manpage and thought it was a good idea to use...

[acheronfail]

I'd like to consider re-opening this issue, since I have a use case for it!
I have built a tool that wraps ripgrep - it's called repgrep: https://github.com/acheronfail/repgrep/.

If we could somehow provide capturing groups in the JSON output (don't mind if it's gated behind a flag or something) then it would enable using repgrep to replace capturing groups, so users could match on something like foo (\w+) and then, when replacing with repgrep, they could use something like bar $1 to use the capturing group in the replacement text.

---

I suppose the only way I could work around this, is if I used ripgrep as a lib (which still seems to be in the experimental phase). That would require a significant refactor of my tool, though - and until libripgrep is stable, I don't think it's a good idea for repgrep.

[BurntSushi]

@acheronfail Is it possible for you to just re-run the regex on the matched lines to get capture groups?

[acheronfail]

@BurntSushi that is a possible workaround, yes.

In fact, it does seem like this is the strategy that VSCode uses:

• runs ripgrep when it's using the node backend
• when replacing, uses JavaScript regexes to perform replacements

So, my program doesn't depend on any regular expression functionality right now, so including a regex crate and using that to match on the lines would definitely be a solution to the issue. I can't imagine it would be that hard, perhaps the only thing is detecting whether a regular expression with groups was passed... I imagine a regex crate would be able to tell me this in some way.

In terms of a performance trade off - I don't think it would be that bad. Since I only need to perform the regular expression matches on visible lines (repgrep is a terminal user interface) I can make it fast enough.

The main bottleneck in performance in repgrep though, is reading the JSON output itself , but that's only really fixed by using ripgrep as a lib.

Long story short: @BurntSushi I think it makes sense for repgrep just to re-run a regular expression on each matched line as needed! Forget about my comment, but I really appreciate you taking the time to humour me! ❤️