Extracting a JSON substring with POSIX.2 regex

Question

I need to extract a field value from a JSON string using POSIX.2 extended regular expressions (man 7 regex on Linux): eg. from {"field":"value"}

I'm used to working with Perl-style regex when a simple /"field":"(.*)"/ would do the trick, but the POSIX regex don't seem to have a notion to express capture groups.

Is there a trick to craft a POSIX pattern that will match exactly the JSON value ?

Perhaps, `"field":"([^"]*)"` will work if there are no `"` inside the value. — Wiktor Stribiżew, Oct 25 '16 at 20:05
Parsing JSON with a regexp is very error-prone. Why not just use a JSON parser library? — Barmar, Oct 25 '16 at 20:05
@Wiktor: That would still match the "field" part. I strictly need the value. — Gene Vincent, Oct 25 '16 at 20:06
You will have in Group 1 or not? What is your real environment? — Wiktor Stribiżew, Oct 25 '16 at 20:06
There are no anchors in the regexp you showed. Are you talking about the parentheses used for the capture group? — Barmar, Oct 25 '16 at 20:07
@Barmar: Thats right, I mean capture groups. This is inside a legacy system where I don't get to pick additional libraries. Thus the old POSIX.2 syntax. — Gene Vincent, Oct 25 '16 at 20:07
The capture group is the problem. My regex is valid, but matches too much. — Gene Vincent, Oct 25 '16 at 20:12
Matches too much? You used `"field":"([^"]*)"` and it works as appropriate? Or did you have to escape the parentheses? — Wiktor Stribiżew, Oct 25 '16 at 21:12

Barmar · Accepted Answer · 2016-10-25T20:23:39.333

0

In POSIX "basic" regular expressions, capture groups are expressed with \(...\), so what you want is:

"field":"\([^"]*\)"

Note that you have to use [^"]* to prevent the pattern from matching across multiple strings (POSIX regular expressions don't have non-greedy quantifiers).

In "extended" regular expressions, you don't need the backslashes. But you still need to use [^"] instead of . to keep from matching too much.

edited Oct 25 '16 at 20:23

answered Oct 25 '16 at 20:11

Barmar

741,623
53
500
612

The `\(...\)` is BRE, not ERE syntax. No idea if it really helps. – Wiktor Stribiżew Oct 25 '16 at 20:21

Extracting a JSON substring with POSIX.2 regex

1 Answers1