Get list of tasks numbers from git log based on message

Question

Task number = JIRA issue number = **** (E.g.: 7600)

Let's suppose that I have a list of commits having the following messages:

PRJ-7600 - first message
PRJ-8283 - second message
PRJ-8283 - third message
PRJ-1001 - fourth message
PRJ-8283 - fifth message
PRJ-7600 - sixth message

where the first one is for the oldest commit.

Wanted output:

1001
7600
8283

I listed my commits using the following command:

git log --author="First Last" --oneline --grep=PRJ --pretty=format:"%s" | sort

where

• committer = author (in this case)
• --grep=PRJ is specified to ignore the comments that were automatically generated ("Merge branch ...") (alternative to --no-merges)
• --pretty=format:"%s" shows only the message (removes the hash)

Actual output:

PRJ-1001 - fourth message
PRJ-7600 - first message
PRJ-7600 - sixth message
PRJ-8283 - fifth message
PRJ-8283 - second message
PRJ-8283 - third message

Is it possible to extract those numbers (probably using regex or something like substring) showing them only once?

Details:

• Windows 7
• git 1.9.5 (msysgit) -> used from cmd, not from Git Bash console

Answer 1

This will do it in either bash or git bash:

git log --author="First Last" --oneline --grep=PRJ --pretty=format:"%s" | sort | cut --delimiter='-' --fields=2 | uniq

So building on the first section posted in the question, the extra is:

| cut --delimiter='-' --fields=2 | uniq

this pipes the sorted output to cut which extracts the 2nd field delimited by a hyphen "-" and the result is then pipes to uniq to display the distinct values.

This solution has a weakness in the form of the delimiter used for cut - if the format of the log message changes, then it may break. A better solution would be to use a regex search (instead of cut) for the issue key ("/PRJ-.+\s/" I think...) and output the number part.

EDIT

So after a bit of digging, it is possible to do this a little more reliably using grep to find the item key (PRJ in this case):

git log ... | grep -oP --regexp="PRJ-\K\d+" | uniq

-o tells grep to output only the matched part of the line
-P is use the PCRE (perl/PHP) flavour of regex, thus enabling us to use the
\K option which causes the matches prior (to that point) to be excluded