Merge branch 'nd/sparse'
[git/git.git] / Documentation / git-read-tree.txt
1 git-read-tree(1)
2 ================
3
4 NAME
5 ----
6 git-read-tree - Reads tree information into the index
7
8
9 SYNOPSIS
10 --------
11 'git read-tree' [[-m [--trivial] [--aggressive] | --reset | --prefix=<prefix>]
12 [-u [--exclude-per-directory=<gitignore>] | -i]]
13 [--index-output=<file>] [--no-sparse-checkout]
14 <tree-ish1> [<tree-ish2> [<tree-ish3>]]
15
16
17 DESCRIPTION
18 -----------
19 Reads the tree information given by <tree-ish> into the index,
20 but does not actually *update* any of the files it "caches". (see:
21 linkgit:git-checkout-index[1])
22
23 Optionally, it can merge a tree into the index, perform a
24 fast-forward (i.e. 2-way) merge, or a 3-way merge, with the `-m`
25 flag. When used with `-m`, the `-u` flag causes it to also update
26 the files in the work tree with the result of the merge.
27
28 Trivial merges are done by 'git-read-tree' itself. Only conflicting paths
29 will be in unmerged state when 'git-read-tree' returns.
30
31 OPTIONS
32 -------
33 -m::
34 Perform a merge, not just a read. The command will
35 refuse to run if your index file has unmerged entries,
36 indicating that you have not finished previous merge you
37 started.
38
39 --reset::
40 Same as -m, except that unmerged entries are discarded
41 instead of failing.
42
43 -u::
44 After a successful merge, update the files in the work
45 tree with the result of the merge.
46
47 -i::
48 Usually a merge requires the index file as well as the
49 files in the working tree are up to date with the
50 current head commit, in order not to lose local
51 changes. This flag disables the check with the working
52 tree and is meant to be used when creating a merge of
53 trees that are not directly related to the current
54 working tree status into a temporary index file.
55
56 -v::
57 Show the progress of checking files out.
58
59 --trivial::
60 Restrict three-way merge by 'git-read-tree' to happen
61 only if there is no file-level merging required, instead
62 of resolving merge for trivial cases and leaving
63 conflicting files unresolved in the index.
64
65 --aggressive::
66 Usually a three-way merge by 'git-read-tree' resolves
67 the merge for really trivial cases and leaves other
68 cases unresolved in the index, so that Porcelains can
69 implement different merge policies. This flag makes the
70 command to resolve a few more cases internally:
71 +
72 * when one side removes a path and the other side leaves the path
73 unmodified. The resolution is to remove that path.
74 * when both sides remove a path. The resolution is to remove that path.
75 * when both sides adds a path identically. The resolution
76 is to add that path.
77
78 --prefix=<prefix>/::
79 Keep the current index contents, and read the contents
80 of named tree-ish under directory at `<prefix>`. The
81 original index file cannot have anything at the path
82 `<prefix>` itself, and have nothing in `<prefix>/`
83 directory. Note that the `<prefix>/` value must end
84 with a slash.
85
86 --exclude-per-directory=<gitignore>::
87 When running the command with `-u` and `-m` options, the
88 merge result may need to overwrite paths that are not
89 tracked in the current branch. The command usually
90 refuses to proceed with the merge to avoid losing such a
91 path. However this safety valve sometimes gets in the
92 way. For example, it often happens that the other
93 branch added a file that used to be a generated file in
94 your branch, and the safety valve triggers when you try
95 to switch to that branch after you ran `make` but before
96 running `make clean` to remove the generated file. This
97 option tells the command to read per-directory exclude
98 file (usually '.gitignore') and allows such an untracked
99 but explicitly ignored file to be overwritten.
100
101 --index-output=<file>::
102 Instead of writing the results out to `$GIT_INDEX_FILE`,
103 write the resulting index in the named file. While the
104 command is operating, the original index file is locked
105 with the same mechanism as usual. The file must allow
106 to be rename(2)ed into from a temporary file that is
107 created next to the usual index file; typically this
108 means it needs to be on the same filesystem as the index
109 file itself, and you need write permission to the
110 directories the index file and index output file are
111 located in.
112
113 --no-sparse-checkout::
114 Disable sparse checkout support even if `core.sparseCheckout`
115 is true.
116
117 <tree-ish#>::
118 The id of the tree object(s) to be read/merged.
119
120
121 Merging
122 -------
123 If `-m` is specified, 'git-read-tree' can perform 3 kinds of
124 merge, a single tree merge if only 1 tree is given, a
125 fast-forward merge with 2 trees, or a 3-way merge if 3 trees are
126 provided.
127
128
129 Single Tree Merge
130 ~~~~~~~~~~~~~~~~~
131 If only 1 tree is specified, 'git-read-tree' operates as if the user did not
132 specify `-m`, except that if the original index has an entry for a
133 given pathname, and the contents of the path matches with the tree
134 being read, the stat info from the index is used. (In other words, the
135 index's stat()s take precedence over the merged tree's).
136
137 That means that if you do a `git read-tree -m <newtree>` followed by a
138 `git checkout-index -f -u -a`, the 'git-checkout-index' only checks out
139 the stuff that really changed.
140
141 This is used to avoid unnecessary false hits when 'git-diff-files' is
142 run after 'git-read-tree'.
143
144
145 Two Tree Merge
146 ~~~~~~~~~~~~~~
147
148 Typically, this is invoked as `git read-tree -m $H $M`, where $H
149 is the head commit of the current repository, and $M is the head
150 of a foreign tree, which is simply ahead of $H (i.e. we are in a
151 fast-forward situation).
152
153 When two trees are specified, the user is telling 'git-read-tree'
154 the following:
155
156 1. The current index and work tree is derived from $H, but
157 the user may have local changes in them since $H;
158
159 2. The user wants to fast-forward to $M.
160
161 In this case, the `git read-tree -m $H $M` command makes sure
162 that no local change is lost as the result of this "merge".
163 Here are the "carry forward" rules:
164
165 I (index) H M Result
166 -------------------------------------------------------
167 0 nothing nothing nothing (does not happen)
168 1 nothing nothing exists use M
169 2 nothing exists nothing remove path from index
170 3 nothing exists exists, use M if "initial checkout"
171 H == M keep index otherwise
172 exists fail
173 H != M
174
175 clean I==H I==M
176 ------------------
177 4 yes N/A N/A nothing nothing keep index
178 5 no N/A N/A nothing nothing keep index
179
180 6 yes N/A yes nothing exists keep index
181 7 no N/A yes nothing exists keep index
182 8 yes N/A no nothing exists fail
183 9 no N/A no nothing exists fail
184
185 10 yes yes N/A exists nothing remove path from index
186 11 no yes N/A exists nothing fail
187 12 yes no N/A exists nothing fail
188 13 no no N/A exists nothing fail
189
190 clean (H=M)
191 ------
192 14 yes exists exists keep index
193 15 no exists exists keep index
194
195 clean I==H I==M (H!=M)
196 ------------------
197 16 yes no no exists exists fail
198 17 no no no exists exists fail
199 18 yes no yes exists exists keep index
200 19 no no yes exists exists keep index
201 20 yes yes no exists exists use M
202 21 no yes no exists exists fail
203
204 In all "keep index" cases, the index entry stays as in the
205 original index file. If the entry were not up to date,
206 'git-read-tree' keeps the copy in the work tree intact when
207 operating under the -u flag.
208
209 When this form of 'git-read-tree' returns successfully, you can
210 see what "local changes" you made are carried forward by running
211 `git diff-index --cached $M`. Note that this does not
212 necessarily match `git diff-index --cached $H` would have
213 produced before such a two tree merge. This is because of cases
214 18 and 19 --- if you already had the changes in $M (e.g. maybe
215 you picked it up via e-mail in a patch form), `git diff-index
216 --cached $H` would have told you about the change before this
217 merge, but it would not show in `git diff-index --cached $M`
218 output after two-tree merge.
219
220 Case #3 is slightly tricky and needs explanation. The result from this
221 rule logically should be to remove the path if the user staged the removal
222 of the path and then switching to a new branch. That however will prevent
223 the initial checkout from happening, so the rule is modified to use M (new
224 tree) only when the contents of the index is empty. Otherwise the removal
225 of the path is kept as long as $H and $M are the same.
226
227 3-Way Merge
228 ~~~~~~~~~~~
229 Each "index" entry has two bits worth of "stage" state. stage 0 is the
230 normal one, and is the only one you'd see in any kind of normal use.
231
232 However, when you do 'git-read-tree' with three trees, the "stage"
233 starts out at 1.
234
235 This means that you can do
236
237 ----------------
238 $ git read-tree -m <tree1> <tree2> <tree3>
239 ----------------
240
241 and you will end up with an index with all of the <tree1> entries in
242 "stage1", all of the <tree2> entries in "stage2" and all of the
243 <tree3> entries in "stage3". When performing a merge of another
244 branch into the current branch, we use the common ancestor tree
245 as <tree1>, the current branch head as <tree2>, and the other
246 branch head as <tree3>.
247
248 Furthermore, 'git-read-tree' has special-case logic that says: if you see
249 a file that matches in all respects in the following states, it
250 "collapses" back to "stage0":
251
252 - stage 2 and 3 are the same; take one or the other (it makes no
253 difference - the same work has been done on our branch in
254 stage 2 and their branch in stage 3)
255
256 - stage 1 and stage 2 are the same and stage 3 is different; take
257 stage 3 (our branch in stage 2 did not do anything since the
258 ancestor in stage 1 while their branch in stage 3 worked on
259 it)
260
261 - stage 1 and stage 3 are the same and stage 2 is different take
262 stage 2 (we did something while they did nothing)
263
264 The 'git-write-tree' command refuses to write a nonsensical tree, and it
265 will complain about unmerged entries if it sees a single entry that is not
266 stage 0.
267
268 OK, this all sounds like a collection of totally nonsensical rules,
269 but it's actually exactly what you want in order to do a fast
270 merge. The different stages represent the "result tree" (stage 0, aka
271 "merged"), the original tree (stage 1, aka "orig"), and the two trees
272 you are trying to merge (stage 2 and 3 respectively).
273
274 The order of stages 1, 2 and 3 (hence the order of three
275 <tree-ish> command line arguments) are significant when you
276 start a 3-way merge with an index file that is already
277 populated. Here is an outline of how the algorithm works:
278
279 - if a file exists in identical format in all three trees, it will
280 automatically collapse to "merged" state by 'git-read-tree'.
281
282 - a file that has _any_ difference what-so-ever in the three trees
283 will stay as separate entries in the index. It's up to "porcelain
284 policy" to determine how to remove the non-0 stages, and insert a
285 merged version.
286
287 - the index file saves and restores with all this information, so you
288 can merge things incrementally, but as long as it has entries in
289 stages 1/2/3 (i.e., "unmerged entries") you can't write the result. So
290 now the merge algorithm ends up being really simple:
291
292 * you walk the index in order, and ignore all entries of stage 0,
293 since they've already been done.
294
295 * if you find a "stage1", but no matching "stage2" or "stage3", you
296 know it's been removed from both trees (it only existed in the
297 original tree), and you remove that entry.
298
299 * if you find a matching "stage2" and "stage3" tree, you remove one
300 of them, and turn the other into a "stage0" entry. Remove any
301 matching "stage1" entry if it exists too. .. all the normal
302 trivial rules ..
303
304 You would normally use 'git-merge-index' with supplied
305 'git-merge-one-file' to do this last step. The script updates
306 the files in the working tree as it merges each path and at the
307 end of a successful merge.
308
309 When you start a 3-way merge with an index file that is already
310 populated, it is assumed that it represents the state of the
311 files in your work tree, and you can even have files with
312 changes unrecorded in the index file. It is further assumed
313 that this state is "derived" from the stage 2 tree. The 3-way
314 merge refuses to run if it finds an entry in the original index
315 file that does not match stage 2.
316
317 This is done to prevent you from losing your work-in-progress
318 changes, and mixing your random changes in an unrelated merge
319 commit. To illustrate, suppose you start from what has been
320 committed last to your repository:
321
322 ----------------
323 $ JC=`git rev-parse --verify "HEAD^0"`
324 $ git checkout-index -f -u -a $JC
325 ----------------
326
327 You do random edits, without running 'git-update-index'. And then
328 you notice that the tip of your "upstream" tree has advanced
329 since you pulled from him:
330
331 ----------------
332 $ git fetch git://.... linus
333 $ LT=`cat .git/FETCH_HEAD`
334 ----------------
335
336 Your work tree is still based on your HEAD ($JC), but you have
337 some edits since. Three-way merge makes sure that you have not
338 added or modified index entries since $JC, and if you haven't,
339 then does the right thing. So with the following sequence:
340
341 ----------------
342 $ git read-tree -m -u `git merge-base $JC $LT` $JC $LT
343 $ git merge-index git-merge-one-file -a
344 $ echo "Merge with Linus" | \
345 git commit-tree `git write-tree` -p $JC -p $LT
346 ----------------
347
348 what you would commit is a pure merge between $JC and $LT without
349 your work-in-progress changes, and your work tree would be
350 updated to the result of the merge.
351
352 However, if you have local changes in the working tree that
353 would be overwritten by this merge, 'git-read-tree' will refuse
354 to run to prevent your changes from being lost.
355
356 In other words, there is no need to worry about what exists only
357 in the working tree. When you have local changes in a part of
358 the project that is not involved in the merge, your changes do
359 not interfere with the merge, and are kept intact. When they
360 *do* interfere, the merge does not even start ('git-read-tree'
361 complains loudly and fails without modifying anything). In such
362 a case, you can simply continue doing what you were in the
363 middle of doing, and when your working tree is ready (i.e. you
364 have finished your work-in-progress), attempt the merge again.
365
366
367 Sparse checkout
368 ---------------
369
370 "Sparse checkout" allows to sparsely populate working directory.
371 It uses skip-worktree bit (see linkgit:git-update-index[1]) to tell
372 Git whether a file on working directory is worth looking at.
373
374 "git read-tree" and other merge-based commands ("git merge", "git
375 checkout"...) can help maintaining skip-worktree bitmap and working
376 directory update. `$GIT_DIR/info/sparse-checkout` is used to
377 define the skip-worktree reference bitmap. When "git read-tree" needs
378 to update working directory, it will reset skip-worktree bit in index
379 based on this file, which uses the same syntax as .gitignore files.
380 If an entry matches a pattern in this file, skip-worktree will be
381 set on that entry. Otherwise, skip-worktree will be unset.
382
383 Then it compares the new skip-worktree value with the previous one. If
384 skip-worktree turns from unset to set, it will add the corresponding
385 file back. If it turns from set to unset, that file will be removed.
386
387 While `$GIT_DIR/info/sparse-checkout` is usually used to specify what
388 files are in. You can also specify what files are _not_ in, using
389 negate patterns. For example, to remove file "unwanted":
390
391 ----------------
392 *
393 !unwanted
394 ----------------
395
396 Another tricky thing is fully repopulating working directory when you
397 no longer want sparse checkout. You cannot just disable "sparse
398 checkout" because skip-worktree are still in the index and you working
399 directory is still sparsely populated. You should re-populate working
400 directory with the `$GIT_DIR/info/sparse-checkout` file content as
401 follows:
402
403 ----------------
404 *
405 ----------------
406
407 Then you can disable sparse checkout. Sparse checkout support in "git
408 read-tree" and similar commands is disabled by default. You need to
409 turn `core.sparseCheckout` on in order to have sparse checkout
410 support.
411
412
413 SEE ALSO
414 --------
415 linkgit:git-write-tree[1]; linkgit:git-ls-files[1];
416 linkgit:gitignore[5]
417
418
419 Author
420 ------
421 Written by Linus Torvalds <torvalds@osdl.org>
422
423 Documentation
424 --------------
425 Documentation by David Greaves, Junio C Hamano and the git-list <git@vger.kernel.org>.
426
427 GIT
428 ---
429 Part of the linkgit:git[1] suite