How does one extract a unified-diff style patch subset? - diff

Question

Every time I want to take a subset of a patch, I'm forced to write a script to only extract the indices that I want.
e.g. I have a patch that applies to sub directories
'yay' and 'foo'.
Is there a way to create a new patch or apply only a subset of a patch? i.e. create a new patch from the existing patch that only takes all indices that are under sub directory 'yay'. Or all indices that are not under sub directory 'foo'
If I have a patch like ( excuse the below pseudo-patch):
Index : foo/bar
yada
yada
- asdf
+ jkl
yada
yada
Index : foo/bah
blah
blah
- 28
+ 29
blah
blah
blah
Index : yay/team
go
huskies
- happy happy
+ joy joy
cougars
suck
How can I extract or apply only the 'yay' subdirectory like:
Index : yay/team
go
huskies
- happy happy
+ joy joy
cougars
suck
I know if I script up a solution I'll be re-inventing the wheel...

Answer 1

Take a look at the filterdiff utility, which is part of patchutils.
For example, if you have the following patch:
$ cat example.patch
diff -Naur orig/a/bar new/a/bar
--- orig/a/bar 2009-12-02 12:41:38.353745751 -0800
+++ new/a/bar 2009-12-02 12:42:17.845745951 -0800
## -1,3 +1,3 ##
4
-5
+e
6
diff -Naur orig/a/foo new/a/foo
--- orig/a/foo 2009-12-02 12:41:32.845745768 -0800
+++ new/a/foo 2009-12-02 12:42:25.697995617 -0800
## -1,3 +1,3 ##
1
2
-3
+c
diff -Naur orig/b/baz new/b/baz
--- orig/b/baz 2009-12-02 12:41:42.993745756 -0800
+++ new/b/baz 2009-12-02 12:42:37.585745735 -0800
## -1,3 +1,3 ##
-7
+z
8
9
Then you can run the following command to extract the patch for only things in the a directory like this:
$ cat example.patch | filterdiff -i 'new/a/*'
--- orig/a/bar 2009-12-02 12:41:38.353745751 -0800
+++ new/a/bar 2009-12-02 12:42:17.845745951 -0800
## -1,3 +1,3 ##
4
-5
+e
6
--- orig/a/foo 2009-12-02 12:41:32.845745768 -0800
+++ new/a/foo 2009-12-02 12:42:25.697995617 -0800
## -1,3 +1,3 ##
1
2
-3
+c

Answer 2

Here's my quick and dirty Perl solution.
perl -ne '#a = split /^Index :/m, join "", <>; END { for(#a) {print "Index :", $_ if (m, yay/team,)}}' < foo.patch

Answer 3

In response to sigjuice's request in the comments, I'm posting my script solution. It isn't 100% bullet proof, and I'll probably use filterdiff instead.
base_usage_str=r'''
python %prog index_regex patch_file
description:
Extracts all indices from a patch-file matching 'index_regex'
e.g.
python %prog '^evc_lib' p.patch > evc_lib_p.patch
Will extract all indices which begin with evc_lib.
-or-
python %prog '^(?!evc_lib)' p.patch > not_evc_lib_p.patch
Will extract all indices which do *not* begin with evc_lib.
authors:
Ross Rogers, 2009.04.02
'''
import re,os,sys
from optparse import OptionParser
def main():
parser = OptionParser(usage=base_usage_str)
(options, args) = parser.parse_args(args=sys.argv[1:])
if len(args) != 2:
parser.print_help()
if len(args) == 0:
sys.exit(0)
else:
sys.exit(1)
(index_regex,patch_file) = args
sys.stderr.write('Extracting patches for indices found by regex:%s\n'%index_regex)
#print 'user_regex',index_regex
user_index_match_regex = re.compile(index_regex)
# Index: verification/ring_td_cte/tests/mmio_wr_td_target.e
# --- sw/cfg/foo.xml 2009-04-30 17:59:11 -07:00
# +++ sw/cfg/foo.xml 2009-05-11 09:26:58 -07:00
index_cre = re.compile(r'''(?:(?<=^)|(?<=\n))(--- (?:.*\n){2,}?(?![ #\+\-]))''')
patch_file = open(patch_file,'r')
all_patch_sets = index_cre.findall(patch_file.read())
patch_file.close()
for file_edit in all_patch_sets:
# extract index subset
index_path = re.compile('\+\+\+ (?P<index>[\w_\-/\.]+)').search(file_edit).group('index').strip()
if user_index_match_regex.search(index_path):
sys.stderr.write("Index regex matched index: "+index_path+"\n")
print file_edit,
if __name__ == '__main__':
main()