regex extract different parts of a string in consistent order

Question

I have a list of strings

my_strings = [
    "2002-03-04 with Matt",
    "Important: 2016-01-23 with Mary",
    "with Tom on 2015-06-30",
]

I want to extract:

• date (always in yyyy-mm-dd format)
• person (always in with person) but I don't want to keep "with"

I could do:

import re
pattern = r'.*(\d{4}-\d{2}-\d{2}).*with \b([^\b]+)\b.*'
matched = [re.match(pattern, x).groups() for x in my_strings]

but it fails because pattern doesn't match "with Tom on 2015-06-30" .

Question s

How do I specify the regex pattern to be indifferent to the order in which date or person appear in the string?

and

How do I ensure that the groups() method returns them in the same order every time?

I expect the output to look like this?

[('2002-03-04', 'Matt'), ('2016-01-23', 'Mary'), ('2015-06-30', 'Tom')]

Answer 1

What about doing it with 2 separate regex?

my_strings = [
    "2002-03-04 with Matt",
    "Important: 2016-01-23 with Mary",
    "with Tom on 2015-06-30",
]
import re

pattern = r'.*(\d{4}-\d{2}-\d{2})'
dates = [re.match(pattern, x).groups()[0] for x in my_strings]

pattern = r'.*with (\w+).*'
persons = [re.match(pattern, x).groups()[0] for x in my_strings]

output = zip(dates, persons)
print output
## [('2002-03-04', 'Matt'), ('2016-01-23', 'Mary'), ('2015-06-30', 'Tom')]

Answer 2

This should work:

my_strings = [
    "2002-03-04 with Matt",
    "Important: 2016-01-23 with Mary",
    "with Tom on 2015-06-30",
]

import re

alternates = r"(?:\b(\d{4}-\d\d-\d\d)\b|with (\w+)|.)*"

for tc in my_strings:
    print(tc)
    m = re.match(alternates, tc)
    if m:
        print("\t", m.group(1))
        print("\t", m.group(2))

Output is:

$ python test.py
2002-03-04 with Matt
     2002-03-04
     Matt
Important: 2016-01-23 with Mary
     2016-01-23
     Mary
with Tom on 2015-06-30
     2015-06-30
     Tom

However, something like this is not totally intuitive. I encourage you to try using named groups if at all possible.

Answer 3

Just for education reasons, a non-regex approach could involve using dateutil parser in a "fuzzy" mode to extract the dates and the nltk toolkit with the named entity recognition to extract names. Complete code:

import nltk
from nltk import pos_tag, ne_chunk
from nltk.tokenize import SpaceTokenizer
from dateutil.parser import parse


def extract_names(text):
    tokenizer = SpaceTokenizer()
    toks = tokenizer.tokenize(text)
    pos = pos_tag(toks)
    chunked_nes = ne_chunk(pos)

    return [' '.join(map(lambda x: x[0], ne.leaves())) for ne in chunked_nes if isinstance(ne, nltk.tree.Tree)]

my_strings = [
    "2002-03-04 with Matt",
    "Important: 2016-01-23 with Mary",
    "with Tom on 2015-06-30"
]

for s in my_strings:
    print(parse(s, fuzzy=True))
    print(extract_names(s))

Prints:

2002-03-04 00:00:00
['Matt']
2016-01-23 00:00:00
['Mary']
2015-06-30 00:00:00
['Tom']

That's probably an over-complication though.

Answer 4

If you use Python's new regex module, you can use conditionals to get
a guaranteed match on 2 items.

I'd think this is more like a standard to do out-of-order matching.

(?:.*?(?:(?(1)(?!))\\b(\\d{4}-\\d\\d-\\d\\d)\\b|(?(2)(?!))with[ ](\\w+))){2}

Expanded

 (?:
      .*? 
      (?:
           (?(1)(?!))
           \b 
           ( \d{4} - \d\d - \d\d )       # (1)
           \b 
        |  (?(2)(?!))
           with [ ] 
           ( \w+ )                       # (2)
      )
 ){2}