Regex for getting multiple words after a delimiter

Question

I have been trying to get the separate groups from the below string using regex in PCRE:

drop = blah blah blah something keep = bar foo nlah aaaa rename = (a=b d=e) obs=4 where = (foo > 45 and bar == 35)

Groups I am trying to make is like:

1. drop = blah blah blah something
2. keep = bar foo nlah aaaa
3. rename = (a=b d=e)
4. obs=4
5. where = (foo > 45 and bar == 35)

I have written a regex using recursion but for some reason recursion is partially working for selecting multiple words after drop like it's selecting just first 3 words (blah blah blah) and not the 4th one. I have looked through various stackoverflow questions and have tried using positive lookahead also but this is the closest I could get to and now I am stuck because I am unable to understand what I am doing wrong.

Same can be seen here: RegEx Demo.

Any help on this or understanding what I am doing wrong is appreciated.

The subroutine looks redundant here, try `(?i)(drop|keep|where|rename|obs)\s*=\s*(?|(\w+(?:\s+\w+)*)(?=\s+\w+\s+=|$)|$(.*?)$)` ([demo](https://regex101.com/r/qAe88o/1)). — Wiktor Stribiżew, Sep 15 '20 at 07:51
Just an idea but maybe you can split on a space where the right conditions are met? Like [so](https://regex101.com/r/Z94EY2/1)? — JvdV, Sep 15 '20 at 07:55
@WiktorStribiżew This works..!! I wasn't aware of things like "Branch Reset Groups". Thanks for this. I am gonna look more about this and see if I can use these in future. :) — Frosty, Sep 15 '20 at 08:35

Jan · Answer 1 · 2020-09-15T12:50:59.903

You could use the newer regex module with DEFINE:

(?(DEFINE)
    (?<key>\w+)
    (?<sep>\s*=\s*)
    (?<value>(?:(?!(?&key)(?&sep))[^()=])+)
    (?<par>\((?:[^()]+|(?&par))+\))
)
(?P<k>(?&key))(?&sep)(?P<v>(?:(?&value)|(?&par)))

See a demo on regex101.com.

In Python this could be:

import regex as re

data = """
drop = blah blah blah something keep = bar foo nlah aaaa rename = (a=b d=e) obs=4 where = (foo > 45 and bar == 35)
"""

rx = re.compile(r'''
(?(DEFINE)
    (?<key>\w+)
    (?<sep>\s*=\s*)
    (?<value>(?:(?!(?&key)(?&sep))[^()=])+)
    (?<par>\((?:[^()]+|(?&par))+\))
)

(?P<k>(?&key))(?&sep)(?P<v>(?:(?&value)|(?&par)))''', re.X)

result = {m.group('k').strip(): m.group('v').strip()
          for m in rx.finditer(data)}

print(result)

And yields

{'drop': 'blah blah blah something', 'keep': 'bar foo nlah aaaa', 'rename': '(a=b d=e)', 'obs': '4', 'where': '(foo > 45 and bar == 35)'}

Thanks..!! this works too. I have been avoiding to build a parser for this because it's not really a bigger functionality that needs to be used at too many places so decided to go with regex instead. for multiple parentheses. I will check but it's a good starting point for me. Can you also please explain the named regex once. That would help it understand more. Thanks again..!! :) — Frosty, Sep 15 '20 at 08:42

score 1 · Accepted Answer · answered Sep 16 '20 at 13:39

You can use a branch reset group solution:

(?i)\b(drop|keep|where|rename|obs)\s*=\s*(?|(\w+(?:\s+\w+)*)(?=\s+\w+\s+=|$)|\((.*?)\))

See the PCRE regex demo

Details

(?i) - case insensitive mode on
\b - a word boundary
(drop|keep|where|rename|obs) - Group 1: any of the words in the group
\s*=\s* - a = char enclosed with 0+ whitespace chars
(?| - start of a branch reset group:
- (\w+(?:\s+\w+)*) - Group 2: one or more word chars followed with zero or more repetitions of one or more whitespaces and one or more word chars
- (?=\s+\w+\s+=|$) - up to one or more whitespaces, one or more word chars, one or more whitespaces, and =, or end of string
- | - or
  - $(.*?)$ - (, then Group 2 capturing any zero or more chars other than line break chars, as few as possible and then )
) - end of the branch reset group.

See Python demo:

import regex
pattern = r"(?i)\b(drop|keep|where|rename|obs)\s*=\s*(?|(\w+(?:\s+\w+)*)(?=\s+\w+\s+=|$)|\((.*?)\))"
text = "drop = blah blah blah something keep = bar foo nlah aaaa rename = (a=b d=e) obs=4 where = (foo > 45 and bar == 35)"
print( [x.group() for x in regex.finditer(pattern, text)] )
# => ['drop = blah blah blah something', 'keep = bar foo nlah aaaa', 'rename = (a=b d=e)', 'obs=4', 'where = (foo > 45 and bar == 35)']
print( regex.findall(pattern, text) )
# => [('drop', 'blah blah blah something'), ('keep', 'bar foo nlah aaaa'), ('rename', 'a=b d=e'), ('obs', '4'), ('where', 'foo > 45 and bar == 35')]

Regex for getting multiple words after a delimiter

2 Answers2