1The output format from "git-diff-index", "git-diff-tree" and 2"git-diff-files" are very similar. 3 4These commands all compare two sets of things; what is 5compared differs: 6 7git-diff-index <tree-ish>:: 8 compares the <tree-ish> and the files on the filesystem. 9 10git-diff-index --cached <tree-ish>:: 11 compares the <tree-ish> and the index. 12 13git-diff-tree [-r] <tree-ish-1> <tree-ish-2> [<pattern>...]:: 14 compares the trees named by the two arguments. 15 16git-diff-files [<pattern>...]:: 17 compares the index and the files on the filesystem. 18 19 20An output line is formatted this way: 21 22------------------------------------------------ 23in-place edit :100644 100644 bcd1234... 0123456... M file0 24copy-edit :100644 100644 abcd123... 1234567... C68 file1 file2 25rename-edit :100644 100644 abcd123... 1234567... R86 file1 file3 26create :000000 100644 0000000... 1234567... A file4 27delete :100644 000000 1234567... 0000000... D file5 28unmerged :000000 000000 0000000... 0000000... U file6 29------------------------------------------------ 30 31That is, from the left to the right: 32 33. a colon. 34. mode for "src"; 000000 if creation or unmerged. 35. a space. 36. mode for "dst"; 000000 if deletion or unmerged. 37. a space. 38. sha1 for "src"; 0\{40\} if creation or unmerged. 39. a space. 40. sha1 for "dst"; 0\{40\} if creation, unmerged or "look at work tree". 41. a space. 42. status, followed by optional "score" number. 43. a tab or a NUL when '-z' option is used. 44. path for "src" 45. a tab or a NUL when '-z' option is used; only exists for C or R. 46. path for "dst"; only exists for C or R. 47. an LF or a NUL when '-z' option is used, to terminate the record. 48 49<sha1> is shown as all 0's if a file is new on the filesystem 50and it is out of sync with the index. 51 52Example: 53 54------------------------------------------------ 55:100644 100644 5be4a4...... 000000...... M file.c 56------------------------------------------------ 57 58When `-z` option is not used, TAB, LF, and backslash characters 59in pathnames are represented as `\t`, `\n`, and `\\`, 60respectively. 61 62 63Generating patches with -p 64-------------------------- 65 66When "git-diff-index", "git-diff-tree", or "git-diff-files" are run 67with a '-p' option, they do not produce the output described above; 68instead they produce a patch file. 69 70The patch generation can be customized at two levels. 71 721. When the environment variable 'GIT_EXTERNAL_DIFF' is not set, 73 these commands internally invoke "diff" like this: 74 75 diff -L a/<path> -L b/<path> -pu <old> <new> 76+ 77For added files, `/dev/null` is used for <old>. For removed 78files, `/dev/null` is used for <new> 79+ 80The "diff" formatting options can be customized via the 81environment variable 'GIT_DIFF_OPTS'. For example, if you 82prefer context diff: 83 84 GIT_DIFF_OPTS=-c git-diff-index -p HEAD 85 86 872. When the environment variable 'GIT_EXTERNAL_DIFF' is set, the 88 program named by it is called, instead of the diff invocation 89 described above. 90+ 91For a path that is added, removed, or modified, 92'GIT_EXTERNAL_DIFF' is called with 7 parameters: 93 94 path old-file old-hex old-mode new-file new-hex new-mode 95+ 96where: 97 98 <old|new>-file:: are files GIT_EXTERNAL_DIFF can use to read the 99 contents of <old|new>, 100 <old|new>-hex:: are the 40-hexdigit SHA1 hashes, 101 <old|new>-mode:: are the octal representation of the file modes. 102 103+ 104The file parameters can point at the user's working file 105(e.g. `new-file` in "git-diff-files"), `/dev/null` (e.g. `old-file` 106when a new file is added), or a temporary file (e.g. `old-file` in the 107index). 'GIT_EXTERNAL_DIFF' should not worry about unlinking the 108temporary file --- it is removed when 'GIT_EXTERNAL_DIFF' exits. 109 110For a path that is unmerged, 'GIT_EXTERNAL_DIFF' is called with 1 111parameter, <path>. 112 113 114git specific extension to diff format 115------------------------------------- 116 117What -p option produces is slightly different from the 118traditional diff format. 119 1201. It is preceded with a "git diff" header, that looks like 121 this: 122 123 diff --git a/file1 b/file2 124+ 125The `a/` and `b/` filenames are the same unless rename/copy is 126involved. Especially, even for a creation or a deletion, 127`/dev/null` is _not_ used in place of `a/` or `b/` filenames. 128+ 129When rename/copy is involved, `file1` and `file2` show the 130name of the source file of the rename/copy and the name of 131the file that rename/copy produces, respectively. 132 1332. It is followed by one or more extended header lines: 134 135 old mode <mode> 136 new mode <mode> 137 deleted file mode <mode> 138 new file mode <mode> 139 copy from <path> 140 copy to <path> 141 rename from <path> 142 rename to <path> 143 similarity index <number> 144 dissimilarity index <number> 145 index <hash>..<hash> <mode> 146 1473. TAB, LF, double quote and backslash characters in pathnames 148 are represented as `\t`, `\n`, `\"` and `\\`, respectively. 149 If there is need for such substitution then the whole 150 pathname is put in double quotes. 151 152 153combined diff format 154-------------------- 155 156git-diff-tree and git-diff-files can take '-c' or '--cc' option 157to produce 'combined diff', which looks like this: 158 159------------ 160diff --combined describe.c 161index fabadb8,cc95eb0..4866510 162--- a/describe.c 163+++ b/describe.c 164@@@ -98,20 -98,12 +98,20 @@@ 165 return (a_date > b_date) ? -1 : (a_date == b_date) ? 0 : 1; 166 } 167 168- static void describe(char *arg) 169 -static void describe(struct commit *cmit, int last_one) 170++static void describe(char *arg, int last_one) 171 { 172 + unsigned char sha1[20]; 173 + struct commit *cmit; 174 struct commit_list *list; 175 static int initialized = 0; 176 struct commit_name *n; 177 178 + if (get_sha1(arg, sha1) < 0) 179 + usage(describe_usage); 180 + cmit = lookup_commit_reference(sha1); 181 + if (!cmit) 182 + usage(describe_usage); 183 + 184 if (!initialized) { 185 initialized = 1; 186 for_each_ref(get_name); 187------------ 188 1891. It is preceded with a "git diff" header, that looks like 190 this (when '-c' option is used): 191 192 diff --combined file 193+ 194or like this (when '--cc' option is used): 195 196 diff --c file 197 1982. It is followed by one or more extended header lines 199 (this example shows a merge with two parents): 200 201 index <hash>,<hash>..<hash> 202 mode <mode>,<mode>..<mode> 203 new file mode <mode> 204 deleted file mode <mode>,<mode> 205+ 206The `mode <mode>,<mode>..<mode>` line appears only if at least one of 207the <mode> is diferent from the rest. Extended headers with 208information about detected contents movement (renames and 209copying detection) are designed to work with diff of two 210<tree-ish> and are not used by combined diff format. 211 2123. It is followed by two-line from-file/to-file header 213 214 --- a/file 215 +++ b/file 216+ 217Similar to two-line header for traditional 'unified' diff 218format, `/dev/null` is used to signal created or deleted 219files. 220 2214. Chunk header format is modified to prevent people from 222 accidentally feeding it to `patch -p1`. Combined diff format 223 was created for review of merge commit changes, and was not 224 meant for apply. The change is similar to the change in the 225 extended 'index' header: 226 227 @@@ <from-file-range> <from-file-range> <to-file-range> @@@ 228+ 229There are (number of parents + 1) `@` characters in the chunk 230header for combined diff format. 231 232Unlike the traditional 'unified' diff format, which shows two 233files A and B with a single column that has `-` (minus -- 234appears in A but removed in B), `+` (plus -- missing in A but 235added to B), or `" "` (space -- unchanged) prefix, this format 236compares two or more files file1, file2,... with one file X, and 237shows how X differs from each of fileN. One column for each of 238fileN is prepended to the output line to note how X's line is 239different from it. 240 241A `-` character in the column N means that the line appears in 242fileN but it does not appear in the result. A `+` character 243in the column N means that the line appears in the last file, 244and fileN does not have that line (in other words, the line was 245added, from the point of view of that parent). 246 247In the above example output, the function signature was changed 248from both files (hence two `-` removals from both file1 and 249file2, plus `++` to mean one line that was added does not appear 250in either file1 nor file2). Also two other lines are the same 251from file1 but do not appear in file2 (hence prefixed with ` +`). 252 253When shown by `git diff-tree -c`, it compares the parents of a 254merge commit with the merge result (i.e. file1..fileN are the 255parents). When shown by `git diff-files -c`, it compares the 256two unresolved merge parents with the working tree file 257(i.e. file1 is stage 2 aka "our version", file2 is stage 3 aka 258"their version"). 259