1Tweaking diff output 2==================== 3June 2005 4 5 6Introduction 7------------ 8 9The diff commands git-diff-index, git-diff-files, and 10git-diff-tree can be told to manipulate differences they find 11in unconventional ways before showing diff(1) output. The 12manipulation is collectively called "diffcore transformation". 13This short note describes what they are and how to use them to 14produce diff outputs that are easier to understand than the 15conventional kind. 16 17 18The chain of operation 19---------------------- 20 21The git-diff-* family works by first comparing two sets of 22files: 23 24 - git-diff-index compares contents of a "tree" object and the 25 working directory (when '\--cached' flag is not used) or a 26 "tree" object and the index file (when '\--cached' flag is 27 used); 28 29 - git-diff-files compares contents of the index file and the 30 working directory; 31 32 - git-diff-tree compares contents of two "tree" objects. 33 34In all of these cases, the commands themselves compare 35corresponding paths in the two sets of files. The result of 36comparison is passed from these commands to what is internally 37called "diffcore", in a format similar to what is output when 38the -p option is not used. E.g. 39 40------------------------------------------------ 41in-place edit :100644 100644 bcd1234... 0123456... M file0 42create :000000 100644 0000000... 1234567... A file4 43delete :100644 000000 1234567... 0000000... D file5 44unmerged :000000 000000 0000000... 0000000... U file6 45------------------------------------------------ 46 47The diffcore mechanism is fed a list of such comparison results 48(each of which is called "filepair", although at this point each 49of them talks about a single file), and transforms such a list 50into another list. There are currently 6 such transformations: 51 52- diffcore-pathspec 53- diffcore-break 54- diffcore-rename 55- diffcore-merge-broken 56- diffcore-pickaxe 57- diffcore-order 58 59These are applied in sequence. The set of filepairs git-diff-\* 60commands find are used as the input to diffcore-pathspec, and 61the output from diffcore-pathspec is used as the input to the 62next transformation. The final result is then passed to the 63output routine and generates either diff-raw format (see Output 64format sections of the manual for git-diff-\* commands) or 65diff-patch format. 66 67 68diffcore-pathspec 69----------------- 70 71The first transformation in the chain is diffcore-pathspec, and 72is controlled by giving the pathname parameters to the 73git-diff-* commands on the command line. The pathspec is used 74to limit the world diff operates in. It removes the filepairs 75outside the specified set of pathnames. 76 77Implementation note. For performance reasons, git-diff-tree 78uses the pathname parameters on the command line to cull set of 79filepairs it feeds the diffcore mechanism itself, and does not 80use diffcore-pathspec, but the end result is the same. 81 82 83diffcore-break 84-------------- 85 86The second transformation in the chain is diffcore-break, and is 87controlled by the -B option to the git-diff-* commands. This is 88used to detect a filepair that represents "complete rewrite" and 89break such filepair into two filepairs that represent delete and 90create. E.g. If the input contained this filepair: 91 92------------------------------------------------ 93:100644 100644 bcd1234... 0123456... M file0 94------------------------------------------------ 95 96and if it detects that the file "file0" is completely rewritten, 97it changes it to: 98 99------------------------------------------------ 100:100644 000000 bcd1234... 0000000... D file0 101:000000 100644 0000000... 0123456... A file0 102------------------------------------------------ 103 104For the purpose of breaking a filepair, diffcore-break examines 105the extent of changes between the contents of the files before 106and after modification (i.e. the contents that have "bcd1234..." 107and "0123456..." as their SHA1 content ID, in the above 108example). The amount of deletion of original contents and 109insertion of new material are added together, and if it exceeds 110the "break score", the filepair is broken into two. The break 111score defaults to 50% of the size of the smaller of the original 112and the result (i.e. if the edit shrinks the file, the size of 113the result is used; if the edit lengthens the file, the size of 114the original is used), and can be customized by giving a number 115after "-B" option (e.g. "-B75" to tell it to use 75%). 116 117 118diffcore-rename 119--------------- 120 121This transformation is used to detect renames and copies, and is 122controlled by the -M option (to detect renames) and the -C option 123(to detect copies as well) to the git-diff-* commands. If the 124input contained these filepairs: 125 126------------------------------------------------ 127:100644 000000 0123456... 0000000... D fileX 128:000000 100644 0000000... 0123456... A file0 129------------------------------------------------ 130 131and the contents of the deleted file fileX is similar enough to 132the contents of the created file file0, then rename detection 133merges these filepairs and creates: 134 135------------------------------------------------ 136:100644 100644 0123456... 0123456... R100 fileX file0 137------------------------------------------------ 138 139When the "-C" option is used, the original contents of modified 140files and contents of unchanged files are considered as 141candidates of the source files in rename/copy operation, in 142addition to the deleted files. If the input were like these 143filepairs, that talk about a modified file fileY and a newly 144created file file0: 145 146------------------------------------------------ 147:100644 100644 0123456... 1234567... M fileY 148:000000 100644 0000000... 0123456... A file0 149------------------------------------------------ 150 151the original contents of fileY and the resulting contents of 152file0 are compared, and if they are similar enough, they are 153changed to: 154 155------------------------------------------------ 156:100644 100644 0123456... 1234567... M fileY 157:100644 100644 0123456... 0123456... C100 fileY file0 158------------------------------------------------ 159 160In both rename and copy detection, the same "extent of changes" 161algorithm used in diffcore-break is used to determine if two 162files are "similar enough", and can be customized to use 163similarity score different from the default 50% by giving a 164number after "-M" or "-C" option (e.g. "-M8" to tell it to use 1658/10 = 80%). 166 167Note. When the "-C" option is used with `\--find-copies-harder` 168option, git-diff-\* commands feed unmodified filepairs to 169diffcore mechanism as well as modified ones. This lets the copy 170detector consider unmodified files as copy source candidates at 171the expense of making it slower. Without `\--find-copies-harder`, 172git-diff-\* commands can detect copies only if the file that was 173copied happened to have been modified in the same changeset. 174 175 176diffcore-merge-broken 177--------------------- 178 179This transformation is used to merge filepairs broken by 180diffcore-break, and were not transformed into rename/copy by 181diffcore-rename, back into a single modification. This always 182runs when diffcore-break is used. 183 184For the purpose of merging broken filepairs back, it uses a 185different "extent of changes" computation from the ones used by 186diffcore-break and diffcore-rename. It counts only the deletion 187from the original, and does not count insertion. If you removed 188only 10 lines from a 100-line document, even if you added 910 189new lines to make a new 1000-line document, you did not do a 190complete rewrite. diffcore-break breaks such a case in order to 191help diffcore-rename to consider such filepairs as candidate of 192rename/copy detection, but if filepairs broken that way were not 193matched with other filepairs to create rename/copy, then this 194transformation merges them back into the original 195"modification". 196 197The "extent of changes" parameter can be tweaked from the 198default 80% (that is, unless more than 80% of the original 199material is deleted, the broken pairs are merged back into a 200single modification) by giving a second number to -B option, 201like these: 202 203* -B50/60 (give 50% "break score" to diffcore-break, use 60% 204 for diffcore-merge-broken). 205 206* -B/60 (the same as above, since diffcore-break defaults to 50%). 207 208Note that earlier implementation left a broken pair as a separate 209creation and deletion patches. This was unnecessary hack and 210the latest implementation always merges all the broken pairs 211back into modifications, but the resulting patch output is 212formatted differently to still let the reviewing easier for such 213a complete rewrite by showing the entire contents of old version 214prefixed with '-', followed by the entire contents of new 215version prefixed with '+'. 216 217 218diffcore-pickaxe 219---------------- 220 221This transformation is used to find filepairs that represent 222changes that touch a specified string, and is controlled by the 223-S option and the `\--pickaxe-all` option to the git-diff-* 224commands. 225 226When diffcore-pickaxe is in use, it checks if there are 227filepairs whose "original" side has the specified string and 228whose "result" side does not. Such a filepair represents "the 229string appeared in this changeset". It also checks for the 230opposite case that loses the specified string. 231 232When `\--pickaxe-all` is not in effect, diffcore-pickaxe leaves 233only such filepairs that touches the specified string in its 234output. When `\--pickaxe-all` is used, diffcore-pickaxe leaves all 235filepairs intact if there is such a filepair, or makes the 236output empty otherwise. The latter behaviour is designed to 237make reviewing of the changes in the context of the whole 238changeset easier. 239 240 241diffcore-order 242-------------- 243 244This is used to reorder the filepairs according to the user's 245(or project's) taste, and is controlled by the -O option to the 246git-diff-* commands. 247 248This takes a text file each of whose line is a shell glob 249pattern. Filepairs that match a glob pattern on an earlier line 250in the file are output before ones that match a later line, and 251filepairs that do not match any glob pattern are output last. 252 253As an example, typical orderfile for the core GIT probably 254would look like this: 255 256------------------------------------------------ 257 README 258 Makefile 259 Documentation 260 *.h 261 *.c 262 t 263------------------------------------------------ 264