9

Looking to quantify how much change happened in each changeset. Any quick way to list maybe kb diff between two revisions?

loneboat
  • 2,845
  • 5
  • 28
  • 40

4 Answers4

3

I had the same thought as @shambulator yesterday! So I've added ability to print delta size in bytes as a part of --diffstat output from my somewhat long and clean patch.py utility.

wget https://raw.githubusercontent.com/techtonik/python-patch/master/patch.py
hg diff -c tip | python patch.py --diffstat --
 codereview/views.py | 28 ++++++++++++++++++++++++++++
 index.yaml          | 10 ++++++++++
 2 files changed, 38 insertions(+), 0 deletions(-), +1267 bytes

UPD: Thanks to @Gili and @mforbes there is now a ticket for Mercurial
https://bz.mercurial-scm.org/show_bug.cgi?id=4245

Community
  • 1
  • 1
anatoly techtonik
  • 19,847
  • 9
  • 124
  • 140
2

Perhaps one can use hg bundle to check the size? (I have not checked how consistent this is in terms of the total repository size.)

function revsize() {
  hg bundle -r $1 --base "p1($1)+p2($1)" /dev/stdout | wc -c
}

How it works

This computes the size (in bytes) using wc -c after generating a bundle for the changes between revision REV = $1 (the first argument to the bash function) and its parents "p1(REV)+p2(REV)" (there may be two if it is a merge.) By using /dev/stdout as a file, the result is sent to standard out where it can be piped to wc -c without creating a file on disk.

mforbes
  • 7,062
  • 3
  • 16
  • 21
  • Note that by default, bundles are compressed using bzip2. For no compression, use "-t none" (see hg bundle docs linked above). – Kim Apr 20 '17 at 15:53
  • @Kim Do you know if the differences are stored compressed on disk or not? If not, ``-t none`` is probably best, but this all needs some testing (or an authoritative source.) – mforbes Apr 20 '17 at 23:29
  • According to this ancient [answer](http://stackoverflow.com/a/1270485/3387608), initial files are compressed, but diffs are uncompressed. I haven't verified this independently or found a more recent confirmation though. – Kim Apr 21 '17 at 14:47
1

hg log --stat is the command you're after. See this example:

$ hg log --stat

changeset:   12431:56e146c7beef
user:        flast
date:        Wed Jun 08 16:12:54 2011 +1000
summary:     Fix the frobulate to frob the knob correctly on tuesdays.

 path/to/src/frob/interface.py       |  29 ++++++++++++++++++++---------
 path/to/tests/systest_frob.py       |  14 ++++++++++++++
 2 files changed, 34 insertions(+), 9 deletions(-)
Jerub
  • 41,746
  • 15
  • 73
  • 90
1

I had the same thought only yesterday! I wrote a quick and dirty Python script for determining total file size change given a unified diff on stdin or as files on the command line. To do it for a changeset, you could just:

hg diff -c <cset id> | patchsize.py

Quick and dirty in the sense that it probably doesn't account for platform-specific line endings, and it doesn't parse diffs particularly cleverly. But it's close enough for my purposes.

anton.burger
  • 5,637
  • 32
  • 48
  • 2
    Yep, I use it all the time. But it only goes as far as line insertions and deletions; I was curious about the total change in file sizes introduced by a patch. – anton.burger Jun 10 '11 at 12:50