1Tweaking diff output 2==================== 3June 2005 4 5 6Introduction 7------------ 8 9The diff commands git-diff-index, git-diff-files, git-diff-tree, and 10git-diff-stages can be told to manipulate differences they find in 11unconventional ways before showing diff(1) output. The manipulation 12is collectively called "diffcore transformation". This short note 13describes what they are and how to use them to produce diff outputs 14that are easier to understand than the conventional kind. 15 16 17The chain of operation 18---------------------- 19 20The git-diff-* family works by first comparing two sets of 21files: 22 23 - git-diff-index compares contents of a "tree" object and the 24 working directory (when '\--cached' flag is not used) or a 25 "tree" object and the index file (when '\--cached' flag is 26 used); 27 28 - git-diff-files compares contents of the index file and the 29 working directory; 30 31 - git-diff-tree compares contents of two "tree" objects; 32 33 - git-diff-stages compares contents of blobs at two stages in an 34 unmerged index file. 35 36In all of these cases, the commands themselves compare 37corresponding paths in the two sets of files. The result of 38comparison is passed from these commands to what is internally 39called "diffcore", in a format similar to what is output when 40the -p option is not used. E.g. 41 42------------------------------------------------ 43in-place edit :100644 100644 bcd1234... 0123456... M file0 44create :000000 100644 0000000... 1234567... A file4 45delete :100644 000000 1234567... 0000000... D file5 46unmerged :000000 000000 0000000... 0000000... U file6 47------------------------------------------------ 48 49The diffcore mechanism is fed a list of such comparison results 50(each of which is called "filepair", although at this point each 51of them talks about a single file), and transforms such a list 52into another list. There are currently 6 such transformations: 53 54- diffcore-pathspec 55- diffcore-break 56- diffcore-rename 57- diffcore-merge-broken 58- diffcore-pickaxe 59- diffcore-order 60 61These are applied in sequence. The set of filepairs git-diff-\* 62commands find are used as the input to diffcore-pathspec, and 63the output from diffcore-pathspec is used as the input to the 64next transformation. The final result is then passed to the 65output routine and generates either diff-raw format (see Output 66format sections of the manual for git-diff-\* commands) or 67diff-patch format. 68 69 70diffcore-pathspec: For Ignoring Files Outside Our Consideration 71--------------------------------------------------------------- 72 73The first transformation in the chain is diffcore-pathspec, and 74is controlled by giving the pathname parameters to the 75git-diff-* commands on the command line. The pathspec is used 76to limit the world diff operates in. It removes the filepairs 77outside the specified set of pathnames. E.g. If the input set 78of filepairs included: 79 80------------------------------------------------ 81:100644 100644 bcd1234... 0123456... M junkfile 82------------------------------------------------ 83 84but the command invocation was "git-diff-files myfile", then the 85junkfile entry would be removed from the list because only "myfile" 86is under consideration. 87 88Implementation note. For performance reasons, git-diff-tree 89uses the pathname parameters on the command line to cull set of 90filepairs it feeds the diffcore mechanism itself, and does not 91use diffcore-pathspec, but the end result is the same. 92 93 94diffcore-break: For Splitting Up "Complete Rewrites" 95---------------------------------------------------- 96 97The second transformation in the chain is diffcore-break, and is 98controlled by the -B option to the git-diff-* commands. This is 99used to detect a filepair that represents "complete rewrite" and 100break such filepair into two filepairs that represent delete and 101create. E.g. If the input contained this filepair: 102 103------------------------------------------------ 104:100644 100644 bcd1234... 0123456... M file0 105------------------------------------------------ 106 107and if it detects that the file "file0" is completely rewritten, 108it changes it to: 109 110------------------------------------------------ 111:100644 000000 bcd1234... 0000000... D file0 112:000000 100644 0000000... 0123456... A file0 113------------------------------------------------ 114 115For the purpose of breaking a filepair, diffcore-break examines 116the extent of changes between the contents of the files before 117and after modification (i.e. the contents that have "bcd1234..." 118and "0123456..." as their SHA1 content ID, in the above 119example). The amount of deletion of original contents and 120insertion of new material are added together, and if it exceeds 121the "break score", the filepair is broken into two. The break 122score defaults to 50% of the size of the smaller of the original 123and the result (i.e. if the edit shrinks the file, the size of 124the result is used; if the edit lengthens the file, the size of 125the original is used), and can be customized by giving a number 126after "-B" option (e.g. "-B75" to tell it to use 75%). 127 128 129diffcore-rename: For Detection Renames and Copies 130------------------------------------------------- 131 132This transformation is used to detect renames and copies, and is 133controlled by the -M option (to detect renames) and the -C option 134(to detect copies as well) to the git-diff-* commands. If the 135input contained these filepairs: 136 137------------------------------------------------ 138:100644 000000 0123456... 0000000... D fileX 139:000000 100644 0000000... 0123456... A file0 140------------------------------------------------ 141 142and the contents of the deleted file fileX is similar enough to 143the contents of the created file file0, then rename detection 144merges these filepairs and creates: 145 146------------------------------------------------ 147:100644 100644 0123456... 0123456... R100 fileX file0 148------------------------------------------------ 149 150When the "-C" option is used, the original contents of modified files, 151and deleted files (and also unmodified files, if the 152"\--find-copies-harder" option is used) are considered as candidates 153of the source files in rename/copy operation. If the input were like 154these filepairs, that talk about a modified file fileY and a newly 155created file file0: 156 157------------------------------------------------ 158:100644 100644 0123456... 1234567... M fileY 159:000000 100644 0000000... bcd3456... A file0 160------------------------------------------------ 161 162the original contents of fileY and the resulting contents of 163file0 are compared, and if they are similar enough, they are 164changed to: 165 166------------------------------------------------ 167:100644 100644 0123456... 1234567... M fileY 168:100644 100644 0123456... bcd3456... C100 fileY file0 169------------------------------------------------ 170 171In both rename and copy detection, the same "extent of changes" 172algorithm used in diffcore-break is used to determine if two 173files are "similar enough", and can be customized to use 174a similarity score different from the default of 50% by giving a 175number after the "-M" or "-C" option (e.g. "-M8" to tell it to use 1768/10 = 80%). 177 178Note. When the "-C" option is used with `\--find-copies-harder` 179option, git-diff-\* commands feed unmodified filepairs to 180diffcore mechanism as well as modified ones. This lets the copy 181detector consider unmodified files as copy source candidates at 182the expense of making it slower. Without `\--find-copies-harder`, 183git-diff-\* commands can detect copies only if the file that was 184copied happened to have been modified in the same changeset. 185 186 187diffcore-merge-broken: For Putting "Complete Rewrites" Back Together 188-------------------------------------------------------------------- 189 190This transformation is used to merge filepairs broken by 191diffcore-break, and not transformed into rename/copy by 192diffcore-rename, back into a single modification. This always 193runs when diffcore-break is used. 194 195For the purpose of merging broken filepairs back, it uses a 196different "extent of changes" computation from the ones used by 197diffcore-break and diffcore-rename. It counts only the deletion 198from the original, and does not count insertion. If you removed 199only 10 lines from a 100-line document, even if you added 910 200new lines to make a new 1000-line document, you did not do a 201complete rewrite. diffcore-break breaks such a case in order to 202help diffcore-rename to consider such filepairs as candidate of 203rename/copy detection, but if filepairs broken that way were not 204matched with other filepairs to create rename/copy, then this 205transformation merges them back into the original 206"modification". 207 208The "extent of changes" parameter can be tweaked from the 209default 80% (that is, unless more than 80% of the original 210material is deleted, the broken pairs are merged back into a 211single modification) by giving a second number to -B option, 212like these: 213 214* -B50/60 (give 50% "break score" to diffcore-break, use 60% 215 for diffcore-merge-broken). 216 217* -B/60 (the same as above, since diffcore-break defaults to 50%). 218 219Note that earlier implementation left a broken pair as a separate 220creation and deletion patches. This was an unnecessary hack and 221the latest implementation always merges all the broken pairs 222back into modifications, but the resulting patch output is 223formatted differently for easier review in case of such 224a complete rewrite by showing the entire contents of old version 225prefixed with '-', followed by the entire contents of new 226version prefixed with '+'. 227 228 229diffcore-pickaxe: For Detecting Addition/Deletion of Specified String 230--------------------------------------------------------------------- 231 232This transformation is used to find filepairs that represent 233changes that touch a specified string, and is controlled by the 234-S option and the `\--pickaxe-all` option to the git-diff-* 235commands. 236 237When diffcore-pickaxe is in use, it checks if there are 238filepairs whose "original" side has the specified string and 239whose "result" side does not. Such a filepair represents "the 240string appeared in this changeset". It also checks for the 241opposite case that loses the specified string. 242 243When `\--pickaxe-all` is not in effect, diffcore-pickaxe leaves 244only such filepairs that touch the specified string in its 245output. When `\--pickaxe-all` is used, diffcore-pickaxe leaves all 246filepairs intact if there is such a filepair, or makes the 247output empty otherwise. The latter behaviour is designed to 248make reviewing of the changes in the context of the whole 249changeset easier. 250 251 252diffcore-order: For Sorting the Output Based on Filenames 253--------------------------------------------------------- 254 255This is used to reorder the filepairs according to the user's 256(or project's) taste, and is controlled by the -O option to the 257git-diff-* commands. 258 259This takes a text file each of whose lines is a shell glob 260pattern. Filepairs that match a glob pattern on an earlier line 261in the file are output before ones that match a later line, and 262filepairs that do not match any glob pattern are output last. 263 264As an example, a typical orderfile for the core git probably 265would look like this: 266 267------------------------------------------------ 268README 269Makefile 270Documentation 271*.h 272*.c 273t 274------------------------------------------------ 275