As input I have several strings containing dates in different formats like
I use dateutil.parser.parse to recognize the dates in the strings.
In the next step I want to remove the dates from the strings. Result should be
Is there a simple way to achieve this?
You can use the fuzzy_with_tokens option to dateutil.parser.parse:
from dateutil.parser import parse
dtstrs = [
"Peter drinks tea at 16:45",
"My birthday is on 08-07-1990",
"On Sat 11 July I'll be back home",
]
out = [
parse(dtstr, fuzzy_with_tokens=True)
for dtstr in dtstrs
]
Result:
[(datetime.datetime(2018, 7, 17, 16, 45), ('Peter drinks tea at ',)),
(datetime.datetime(1990, 8, 7, 0, 0), ('My birthday is on ',)),
(datetime.datetime(2018, 7, 11, 0, 0), ('On ', ' ', " I'll be back home"))]
When fuzzy_with_tokens is true, the parser returns a tuple of a datetime and a tuple of ignored tokens (with the used tokens removed). You can join them back into a string like this:
>>> ['<missing>'.join(x[1]) for x in out]
['Peter drinks tea at ',
'My birthday is on ',
"On <missing> <missing> I'll be back home"]
I'll note that the fuzzy parsing logic is not amazingly reliable, because it's very difficult to pick out only valid components from a string and use them. If you change the person drinking tea to someone named April, for example:
>>> dt, tokens = parse("April drinks tea at 16:45", fuzzy_with_tokens=True)
>>> print(dt)
2018-04-17 16:45:00
>>> print('<missing>'.join(tokens))
drinks tea at
So I would urge some caution with this approach (though I can't really recommend a better approach, this is just a hard problem).