Feature request: extend Match class to access group's start/end index

[yan-zaitsev]

dart --version: Dart SDK version: 2.10.5 (stable) (Tue Jan 19 13:05:37 2021 +0100) on "macos_x64"

My feature request is to extend
Match class: https://github.com/dart-lang/sdk/blob/master/sdk/lib/core/pattern.dart
so it will be possible to get start/end position of founded groups.

My case:

I want to find some substring in text and bold it in UI for user so he will see entered search text in founded texts.
I am doing it using simple Regex:

 final stringRanges = RegExp("${RegExp.escape(searchInput)}", caseSensitive: false, unicode: true)
        .allMatches(text)
        .toList();

It will return me List<Match> which I could render in UI.

Problem

When I want to use more complex Regex with groups:

final stringRanges = RegExp("begin(${RegExp.escape(searchInput)})end", caseSensitive: false, unicode: true)
        .allMatches(text)
        .toList();

At this moment, stringRanges contains more complex Match with groups. I want to parse every group in match with its start/end positions.

Solution

Match class has method String? group(int group); to get string value of founded group.
I will happy to have api like Match? groupMatch(int group); to get more information about single group.
For me it will be enough to have simpler api:

class StringRange {
  int start;
  int end;
 StringRange(this.start, this.end);
}

abstract class Match {
...
StringRange? groupRange(int group);
...
}

[lrhn]

The reason that functionality is not available is that it's not possible in JavaScript.
The JavaScript match object only provides access to captures as strings, not their position in the original string.

[yan-zaitsev]

@lrhn I don't know JS very well, but, maybe, something from https://stackoverflow.com/questions/1985594/how-to-find-indices-of-groups-in-javascript-regular-expressions-match could be useful?

If it is not possible, please, close the task

[iarkh]

@iarkh I don't know JS very well, but, maybe, something from https://stackoverflow.com/questions/1985594/how-to-find-indices-of-groups-in-javascript-regular-expressions-match could be useful?

If it is not possible, please, close the task

@yan-zaitsev, seems like this is a question for @lrhn, isn't it?

[lrhn]

I have thought about whether we could reasonably make capture group indices available in JS before, and have not found a good way.
The approach in the stackoverflow link here works for linear regexps, but captures can be inside repetitions as well, and while you could potentially unroll some RegExp repetitions to make it possible to capture "everything up to the capture group" as a separate capture group, I can guarantee that I can create RegExps where that's not possible. And it would be a rewrite that we have to do at run-time in JS compiled code. We'd be better off creating our own JS-compiled RegExp implementation if that's what we wanted (we don't want to do that, the one in browsers is pretty efficient).

So, if you really, really want to know the indices of RegExp capture groups, you'll have to massage your RegExp yourself to capture "everything before the capture" as well. It's not something we can automate in an efficient and consistent way.

So, closing as not planned (until JS changes to make the information available).
(It is annoying because the underlying engine in the VM, and in Chrome and Firefox, does have the information. It remembers the position of the captures and can even create the capture strings lazily from those indices when you first ask for them).

[mraleph]

@lrhn RegExp Match Indices are at Stage 4 in TC-39 (see https://github.com/tc39/proposal-regexp-match-indices) and are shipping in M91 and FF Nightly and Safari preview. So maybe you could reconsider closing this issue.

[lrhn]

That's great news. We'll still have to decide when we think the browsers we support are sufficiently updated to make the switch
(and then it'll probably be a breaking change to the RegExpMatch class. There are other null-safety related changes I'd like to make if we are breaking it anyway).

[lrhn]

Proof of concept: https://dart-review.googlesource.com/c/sdk/+/212582

(Definitely needs more design work, and then we need to decide whether we think it's well-enough supported on the web).