last year occurrence from string - regex

Question

I have strings like this:
ACB 01900 X1911D 1910 1955-2011 3424 2135 1934 foobar
I'm trying to get the last occurrence of a single year (from 1900 to 2050), so I need to extract only 1934 from that string.
I'm trying with:
grep -P -o '\s(19|20)[0-9]{2}\s(?!\s(19|20)[0-9]{2}\s)'
or
grep -P -o '((19|20)[0-9]{2})(?!\s\1\s)'
But it matches: 1910 and 1934
Here's the Regex101 example:
https://regex101.com/r/UetMl0/3
https://regex101.com/r/UetMl0/4
Plus: how can I extract the year without the surrounding spaces without doing an extra grep to filter them?

Answer 1

Have you ever heard this saying:
Some people, when confronted with a problem, think
“I know, I'll use regular expressions.” Now they have two problems.
Keep it simple - you're interested in finding a number between 2 numbers so just use a numeric comparison, not a regexp:
$ awk -v min=1900 -v max=2050 '{yr=""; for (i=1;i<=NF;i++) if ( ($i ~ /^[0-9]{4}$/) && ($i >= min) && ($i <= max) ) yr=$i; print yr}' file
1934
You didn't say what to do if no date within your range is present so the above outputs a blank line if that happens but is easily tweaked to do anything else.
To change the above script to find the first instead of the last date is trivial (move the print inside the if), to use different start or end dates in your range is trivial (change the min and/or max values), etc., etc. which is a strong indication that this is the right approach. Try changing any of those requirements with a regexp-based solution.

Answer 2

I don't see a way to do this with grep because it doesn't let you output just one of the capture groups, only the whole match.
Wit perl I'd do something like
perl -lpe 'if (/^.*\b(19\d\d|20(?:0-4\d|50))\b/) { print $1 }'
Idea: Use ^.* (greedy) to consume as much of the string up front as possible, thus finding the last possible match. Use \b (word boundary) around the matched number to prevent matching 01900 or X1911D. Only print the first capture group ($1).
I tried to implement your requirement of 1900-2050; if that's too complicated, ((?:19|20)\d\d) will do (but also match e.g. 2099).

Answer 3

The regex to do your task using grep can be as follows:
\b(?:19\d{2}|20[0-4]\d|2050)\b(?!.*\b(?:19\d{2}|20[0-4]\d|2050)\b)
Details:
\b - Word boundary.
(?: - Start of a non-capturing group, needed as a container for
alternatives.
19\d{2}| - The first alternative (1900 - 1999).
20[0-4]\d| - The second alternative (2000 - 2049).
2050 - The third alternative, just 2050.
) - End of the non-capturing group.
\b - Word boundary.
(?! - Negative lookahead for:
.* - A sequence of any chars, meaning actually "what follows
can occur anywhere further".
\b(?:19\d{2}|20[0-4]\d|2050)\b - The same expression as before.
) - End of the negative lookahead.
The word boundary anchors provide that you will not match numbers - parts
of longer words, e.g. X1911D.
The negative lookahead provides that you will match just the last
occurrence of the required year.
If you can use other tool than grep, supporting call to a previous
numbered group (?n), where n is the number of another capturing
group, the regex can be a bit simpler:
(\b(?:19\d{2}|20[0-4]\d|2050)\b)(?!.*(?1))
Details:
(\b(?:19\d{2}|20[0-4]\d|2050)\b) - The regex like before, but
enclosed within a capturing group (it will be "called" later).
(?!.*(?1)) - Negative lookahead for capturing group No 1,
located anywhere further.
This way you avoid writing the same expression again.
For a working example in regex101 see https://regex101.com/r/fvVnZl/1

Answer 4

You may use a PCRE regex without any groups to only return the last occurrence of a pattern you need if you prepend the pattern with ^.*\K, or, in your case, since you expect a whitespace boundary, ^(?:.*\s)?\K:
grep -Po '^(?:.*\s)?\K(?:19\d{2}|20(?:[0-4]\d|50))(?!\S)' file
See the regex demo.
Details
^ - start of line
(?:.*\s)? - an optional non-capturing group matching 1 or 0 occurrences of
.* - any 0+ chars other than line break chars, as many as possible
\s - a whitespace char
\K - match reset operator discarding the text matched so far
(?:19\d{2}|20(?:[0-4]\d|50)) - 19 and any two digits or 20 followed with either a digit from 0 to 4 and then any digit (00 to 49) or 50.
(?!\S) - a whitespace or end of string.
See an online demo:
s="ACB 01900 X1911D 1910 1955-2011 3424 2135 1934 foobar"
grep -Po '^(?:.*\s)?\K(?:19\d{2}|20(?:[0-4]\d|50))(?!\S)' <<< "$s"
# => 1934