sduduzog/diff-buddy

Diff Buddy!

This is a small utility to help with diffs on really large pull requests. To use:

1. Clone the repo you're looking at (add and fetch all remotes, if necessary)
2. Run diff-buddy - if you run from within the repo, it will default to diff against master
3. Filter down results iteratively at the command-line with subsequent runs until you end up seeing a useful changeset

Options

diff-buddy has help which can be invoked from the command-line with --help:

diff-buddy {args}

-a, --at [number]               Show the diff at the provided index only
-f, --from-branch [text]        Starting branch for the diff (master)
-h, --help                      shows this help
-I, --ignore-files [text]       Regular expressions for files to ignore completely
-i, --ignore-lines [text]       Regular expressions for lines to ignore when deciding on interesting files
    --ignore-whitespace         Ignore whitespace changes, even those spanning multiple lines, when deciding on interesting files
-l, --limit [number]            Limit the number of files to list (2147483647)
-o, --offset [number]           Start printing from this offset (0)
-r, --repo [text]               (.)
-S, --show-indexes              Show the numeric index for each file
    --show-operations           Show the operation which this file has undergone: (A)dded, (D)eleted, (M)odified, (R)enamed
-s, --show-patches              Whether or not to show patches for files
-t, --to-branch [text]          Later branch to diff with

Negate any flag argument with --no-{option}

(note that flags can be negated with --no-, eg --no-show-indexes)

Building filters

The whole point of diff-buddy is to reduce the number of files you need to look at to understand a large pull request. Sometimes, a pull-request has many files which have changed but in an insignificant way, eg because of a namespace rename or whitespace formatting.

The process, then, is as follows, for slimming down what would be interesting when comparing some satellite branch in a local repo to master:

1. Check out the satellite branch of interest
2. Make sure you've got the latest: git pull --rebase
3. Start with a plain old diff-buddy.exe run in the repository to get a feel for the size of the diff
4. Scan for files which aren't of interest - perhaps you don't care about test files:
   • diff-buddy --ignore-files Tests.*
   • note that --ignore-files and --ignore-lines take regex strings: when working on unix shells like bash and zsh, you may need to surround these with single-quotes to prevent the shell from expanding them
5. Enable patches and start looking at smaller clumps of files:
   • diff-buddy --ignore-files Tests.* --show-patches --limit 10 --offset 0
   • increase the offset by 10 at a time and look for uninteresting patterns to exclude
6. Exclude files with uninteresting changes:
   • diff-buddy --ignore-files Tests.* --ignore-lines ^using --show-patches --limit 10 --offset 10
     • this ignores files where the only changes are to using statements, still only outputting 10 files, after an offset of 10
7. Add to the ignores for a few iterations until you can do
   • diff-buddy --ignore-files {...} --ignore-lines {...}
   • and end up with a small enough set of files that you are willing to review full patches for
8. Now run with your prior commandline and --review to see each file, one-by-one