]>
Commit | Line | Data |
---|---|---|
6e411d20 SP |
1 | git-fast-import(1) |
2 | ================== | |
3 | ||
4 | NAME | |
5 | ---- | |
6 | git-fast-import - Backend for fast Git data importers. | |
7 | ||
8 | ||
9 | SYNOPSIS | |
10 | -------- | |
11 | frontend | 'git-fast-import' [options] | |
12 | ||
13 | DESCRIPTION | |
14 | ----------- | |
15 | This program is usually not what the end user wants to run directly. | |
16 | Most end users want to use one of the existing frontend programs, | |
17 | which parses a specific type of foreign source and feeds the contents | |
18 | stored there to git-fast-import (gfi). | |
19 | ||
20 | gfi reads a mixed command/data stream from standard input and | |
21 | writes one or more packfiles directly into the current repository. | |
22 | When EOF is received on standard input, fast import writes out | |
23 | updated branch and tag refs, fully updating the current repository | |
24 | with the newly imported data. | |
25 | ||
26 | The gfi backend itself can import into an empty repository (one that | |
27 | has already been initialized by gitlink:git-init[1]) or incrementally | |
28 | update an existing populated repository. Whether or not incremental | |
29 | imports are supported from a particular foreign source depends on | |
30 | the frontend program in use. | |
31 | ||
32 | ||
33 | OPTIONS | |
34 | ------- | |
63e0c8b3 SP |
35 | --date-format=<fmt>:: |
36 | Specify the type of dates the frontend will supply to | |
37 | gfi within `author`, `committer` and `tagger` commands. | |
38 | See ``Date Formats'' below for details about which formats | |
39 | are supported, and their syntax. | |
40 | ||
7073e69e SP |
41 | --force:: |
42 | Force updating modified existing branches, even if doing | |
43 | so would cause commits to be lost (as the new commit does | |
44 | not contain the old commit). | |
45 | ||
6e411d20 SP |
46 | --max-pack-size=<n>:: |
47 | Maximum size of each output packfile, expressed in MiB. | |
48 | The default is 4096 (4 GiB) as that is the maximum allowed | |
49 | packfile size (due to file format limitations). Some | |
50 | importers may wish to lower this, such as to ensure the | |
51 | resulting packfiles fit on CDs. | |
52 | ||
53 | --depth=<n>:: | |
54 | Maximum delta depth, for blob and tree deltification. | |
55 | Default is 10. | |
56 | ||
57 | --active-branches=<n>:: | |
58 | Maximum number of branches to maintain active at once. | |
59 | See ``Memory Utilization'' below for details. Default is 5. | |
60 | ||
61 | --export-marks=<file>:: | |
62 | Dumps the internal marks table to <file> when complete. | |
63 | Marks are written one per line as `:markid SHA-1`. | |
64 | Frontends can use this file to validate imports after they | |
65 | have been completed. | |
66 | ||
c499d768 SP |
67 | --quiet:: |
68 | Disable all non-fatal output, making gfi silent when it | |
69 | is successful. This option disables the output shown by | |
70 | \--stats. | |
71 | ||
72 | --stats:: | |
73 | Display some basic statistics about the objects gfi has | |
74 | created, the packfiles they were stored into, and the | |
75 | memory used by gfi during this run. Showing this output | |
76 | is currently the default, but can be disabled with \--quiet. | |
77 | ||
78 | ||
6e411d20 SP |
79 | Performance |
80 | ----------- | |
81 | The design of gfi allows it to import large projects in a minimum | |
82 | amount of memory usage and processing time. Assuming the frontend | |
83 | is able to keep up with gfi and feed it a constant stream of data, | |
84 | import times for projects holding 10+ years of history and containing | |
85 | 100,000+ individual commits are generally completed in just 1-2 | |
86 | hours on quite modest (~$2,000 USD) hardware. | |
87 | ||
88 | Most bottlenecks appear to be in foreign source data access (the | |
89 | source just cannot extract revisions fast enough) or disk IO (gfi | |
90 | writes as fast as the disk will take the data). Imports will run | |
91 | faster if the source data is stored on a different drive than the | |
92 | destination Git repository (due to less IO contention). | |
93 | ||
94 | ||
95 | Development Cost | |
96 | ---------------- | |
97 | A typical frontend for gfi tends to weigh in at approximately 200 | |
98 | lines of Perl/Python/Ruby code. Most developers have been able to | |
99 | create working importers in just a couple of hours, even though it | |
100 | is their first exposure to gfi, and sometimes even to Git. This is | |
101 | an ideal situation, given that most conversion tools are throw-away | |
102 | (use once, and never look back). | |
103 | ||
104 | ||
105 | Parallel Operation | |
106 | ------------------ | |
107 | Like `git-push` or `git-fetch`, imports handled by gfi are safe to | |
108 | run alongside parallel `git repack -a -d` or `git gc` invocations, | |
109 | or any other Git operation (including `git prune`, as loose objects | |
110 | are never used by gfi). | |
111 | ||
7073e69e SP |
112 | gfi does not lock the branch or tag refs it is actively importing. |
113 | After the import, during its ref update phase, gfi tests each | |
114 | existing branch ref to verify the update will be a fast-forward | |
115 | update (the commit stored in the ref is contained in the new | |
116 | history of the commit to be written). If the update is not a | |
117 | fast-forward update, gfi will skip updating that ref and instead | |
118 | prints a warning message. gfi will always attempt to update all | |
119 | branch refs, and does not stop on the first failure. | |
120 | ||
c499d768 SP |
121 | Branch updates can be forced with \--force, but its recommended that |
122 | this only be used on an otherwise quiet repository. Using \--force | |
7073e69e | 123 | is not necessary for an initial import into an empty repository. |
6e411d20 SP |
124 | |
125 | ||
126 | Technical Discussion | |
127 | -------------------- | |
128 | gfi tracks a set of branches in memory. Any branch can be created | |
129 | or modified at any point during the import process by sending a | |
130 | `commit` command on the input stream. This design allows a frontend | |
131 | program to process an unlimited number of branches simultaneously, | |
132 | generating commits in the order they are available from the source | |
133 | data. It also simplifies the frontend programs considerably. | |
134 | ||
135 | gfi does not use or alter the current working directory, or any | |
136 | file within it. (It does however update the current Git repository, | |
137 | as referenced by `GIT_DIR`.) Therefore an import frontend may use | |
138 | the working directory for its own purposes, such as extracting file | |
139 | revisions from the foreign source. This ignorance of the working | |
140 | directory also allows gfi to run very quickly, as it does not | |
141 | need to perform any costly file update operations when switching | |
142 | between branches. | |
143 | ||
144 | Input Format | |
145 | ------------ | |
146 | With the exception of raw file data (which Git does not interpret) | |
147 | the gfi input format is text (ASCII) based. This text based | |
148 | format simplifies development and debugging of frontend programs, | |
149 | especially when a higher level language such as Perl, Python or | |
150 | Ruby is being used. | |
151 | ||
152 | gfi is very strict about its input. Where we say SP below we mean | |
153 | *exactly* one space. Likewise LF means one (and only one) linefeed. | |
154 | Supplying additional whitespace characters will cause unexpected | |
155 | results, such as branch names or file names with leading or trailing | |
156 | spaces in their name, or early termination of gfi when it encounters | |
157 | unexpected input. | |
158 | ||
63e0c8b3 SP |
159 | Date Formats |
160 | ~~~~~~~~~~~~ | |
161 | The following date formats are supported. A frontend should select | |
162 | the format it will use for this import by passing the format name | |
c499d768 | 163 | in the \--date-format=<fmt> command line option. |
63e0c8b3 SP |
164 | |
165 | `raw`:: | |
9b92c82f | 166 | This is the Git native format and is `<time> SP <offutc>`. |
c499d768 | 167 | It is also gfi's default format, if \--date-format was |
63e0c8b3 SP |
168 | not specified. |
169 | + | |
170 | The time of the event is specified by `<time>` as the number of | |
171 | seconds since the UNIX epoch (midnight, Jan 1, 1970, UTC) and is | |
172 | written as an ASCII decimal integer. | |
173 | + | |
9b92c82f SP |
174 | The local offset is specified by `<offutc>` as a positive or negative |
175 | offset from UTC. For example EST (which is 5 hours behind UTC) | |
176 | would be expressed in `<tz>` by ``-0500'' while UTC is ``+0000''. | |
177 | The local offset does not affect `<time>`; it is used only as an | |
178 | advisement to help formatting routines display the timestamp. | |
63e0c8b3 | 179 | + |
9b92c82f SP |
180 | If the local offset is not available in the source material, use |
181 | ``+0000'', or the most common local offset. For example many | |
63e0c8b3 SP |
182 | organizations have a CVS repository which has only ever been accessed |
183 | by users who are located in the same location and timezone. In this | |
f842fdb0 | 184 | case a reasonable offset from UTC could be assumed. |
63e0c8b3 SP |
185 | + |
186 | Unlike the `rfc2822` format, this format is very strict. Any | |
187 | variation in formatting will cause gfi to reject the value. | |
188 | ||
189 | `rfc2822`:: | |
190 | This is the standard email format as described by RFC 2822. | |
191 | + | |
192 | An example value is ``Tue Feb 6 11:22:18 2007 -0500''. The Git | |
f842fdb0 | 193 | parser is accurate, but a little on the lenient side. It is the |
63e0c8b3 SP |
194 | same parser used by gitlink:git-am[1] when applying patches |
195 | received from email. | |
196 | + | |
197 | Some malformed strings may be accepted as valid dates. In some of | |
198 | these cases Git will still be able to obtain the correct date from | |
199 | the malformed string. There are also some types of malformed | |
200 | strings which Git will parse wrong, and yet consider valid. | |
201 | Seriously malformed strings will be rejected. | |
202 | + | |
9b92c82f SP |
203 | Unlike the `raw` format above, the timezone/UTC offset information |
204 | contained in an RFC 2822 date string is used to adjust the date | |
205 | value to UTC prior to storage. Therefore it is important that | |
206 | this information be as accurate as possible. | |
207 | + | |
f842fdb0 | 208 | If the source material uses RFC 2822 style dates, |
63e0c8b3 SP |
209 | the frontend should let gfi handle the parsing and conversion |
210 | (rather than attempting to do it itself) as the Git parser has | |
211 | been well tested in the wild. | |
212 | + | |
213 | Frontends should prefer the `raw` format if the source material | |
f842fdb0 SP |
214 | already uses UNIX-epoch format, can be coaxed to give dates in that |
215 | format, or its format is easiliy convertible to it, as there is no | |
216 | ambiguity in parsing. | |
63e0c8b3 SP |
217 | |
218 | `now`:: | |
219 | Always use the current time and timezone. The literal | |
220 | `now` must always be supplied for `<when>`. | |
221 | + | |
222 | This is a toy format. The current time and timezone of this system | |
223 | is always copied into the identity string at the time it is being | |
224 | created by gfi. There is no way to specify a different time or | |
225 | timezone. | |
226 | + | |
227 | This particular format is supplied as its short to implement and | |
228 | may be useful to a process that wants to create a new commit | |
229 | right now, without needing to use a working directory or | |
230 | gitlink:git-update-index[1]. | |
231 | + | |
232 | If separate `author` and `committer` commands are used in a `commit` | |
233 | the timestamps may not match, as the system clock will be polled | |
234 | twice (once for each command). The only way to ensure that both | |
235 | author and committer identity information has the same timestamp | |
236 | is to omit `author` (thus copying from `committer`) or to use a | |
237 | date format other than `now`. | |
238 | ||
6e411d20 SP |
239 | Commands |
240 | ~~~~~~~~ | |
241 | gfi accepts several commands to update the current repository | |
242 | and control the current import process. More detailed discussion | |
243 | (with examples) of each command follows later. | |
244 | ||
245 | `commit`:: | |
246 | Creates a new branch or updates an existing branch by | |
247 | creating a new commit and updating the branch to point at | |
248 | the newly created commit. | |
249 | ||
250 | `tag`:: | |
251 | Creates an annotated tag object from an existing commit or | |
252 | branch. Lightweight tags are not supported by this command, | |
253 | as they are not recommended for recording meaningful points | |
254 | in time. | |
255 | ||
256 | `reset`:: | |
257 | Reset an existing branch (or a new branch) to a specific | |
258 | revision. This command must be used to change a branch to | |
259 | a specific revision without making a commit on it. | |
260 | ||
261 | `blob`:: | |
262 | Convert raw file data into a blob, for future use in a | |
263 | `commit` command. This command is optional and is not | |
264 | needed to perform an import. | |
265 | ||
266 | `checkpoint`:: | |
267 | Forces gfi to close the current packfile, generate its | |
268 | unique SHA-1 checksum and index, and start a new packfile. | |
269 | This command is optional and is not needed to perform | |
270 | an import. | |
271 | ||
272 | `commit` | |
273 | ~~~~~~~~ | |
274 | Create or update a branch with a new commit, recording one logical | |
275 | change to the project. | |
276 | ||
277 | .... | |
278 | 'commit' SP <ref> LF | |
279 | mark? | |
63e0c8b3 SP |
280 | ('author' SP <name> SP LT <email> GT SP <when> LF)? |
281 | 'committer' SP <name> SP LT <email> GT SP <when> LF | |
6e411d20 SP |
282 | data |
283 | ('from' SP <committish> LF)? | |
284 | ('merge' SP <committish> LF)? | |
825769a8 | 285 | (filemodify | filedelete | filedeleteall)* |
6e411d20 SP |
286 | LF |
287 | .... | |
288 | ||
289 | where `<ref>` is the name of the branch to make the commit on. | |
290 | Typically branch names are prefixed with `refs/heads/` in | |
291 | Git, so importing the CVS branch symbol `RELENG-1_0` would use | |
292 | `refs/heads/RELENG-1_0` for the value of `<ref>`. The value of | |
293 | `<ref>` must be a valid refname in Git. As `LF` is not valid in | |
294 | a Git refname, no quoting or escaping syntax is supported here. | |
295 | ||
296 | A `mark` command may optionally appear, requesting gfi to save a | |
297 | reference to the newly created commit for future use by the frontend | |
298 | (see below for format). It is very common for frontends to mark | |
299 | every commit they create, thereby allowing future branch creation | |
300 | from any imported commit. | |
301 | ||
302 | The `data` command following `committer` must supply the commit | |
303 | message (see below for `data` command syntax). To import an empty | |
304 | commit message use a 0 length data. Commit messages are free-form | |
305 | and are not interpreted by Git. Currently they must be encoded in | |
306 | UTF-8, as gfi does not permit other encodings to be specified. | |
307 | ||
825769a8 SP |
308 | Zero or more `filemodify`, `filedelete` and `filedeleteall` commands |
309 | may be included to update the contents of the branch prior to | |
310 | creating the commit. These commands may be supplied in any order. | |
311 | However it is recommended that a `filedeleteall` command preceed | |
312 | all `filemodify` commands in the same commit, as `filedeleteall` | |
313 | wipes the branch clean (see below). | |
6e411d20 SP |
314 | |
315 | `author` | |
316 | ^^^^^^^^ | |
317 | An `author` command may optionally appear, if the author information | |
318 | might differ from the committer information. If `author` is omitted | |
319 | then gfi will automatically use the committer's information for | |
320 | the author portion of the commit. See below for a description of | |
321 | the fields in `author`, as they are identical to `committer`. | |
322 | ||
323 | `committer` | |
324 | ^^^^^^^^^^^ | |
325 | The `committer` command indicates who made this commit, and when | |
326 | they made it. | |
327 | ||
328 | Here `<name>` is the person's display name (for example | |
329 | ``Com M Itter'') and `<email>` is the person's email address | |
330 | (``cm@example.com''). `LT` and `GT` are the literal less-than (\x3c) | |
331 | and greater-than (\x3e) symbols. These are required to delimit | |
332 | the email address from the other fields in the line. Note that | |
333 | `<name>` is free-form and may contain any sequence of bytes, except | |
334 | `LT` and `LF`. It is typically UTF-8 encoded. | |
335 | ||
63e0c8b3 | 336 | The time of the change is specified by `<when>` using the date format |
c499d768 | 337 | that was selected by the \--date-format=<fmt> command line option. |
63e0c8b3 SP |
338 | See ``Date Formats'' above for the set of supported formats, and |
339 | their syntax. | |
6e411d20 SP |
340 | |
341 | `from` | |
342 | ^^^^^^ | |
343 | Only valid for the first commit made on this branch by this | |
344 | gfi process. The `from` command is used to specify the commit | |
345 | to initialize this branch from. This revision will be the first | |
346 | ancestor of the new commit. | |
347 | ||
348 | Omitting the `from` command in the first commit of a new branch will | |
349 | cause gfi to create that commit with no ancestor. This tends to be | |
350 | desired only for the initial commit of a project. Omitting the | |
351 | `from` command on existing branches is required, as the current | |
352 | commit on that branch is automatically assumed to be the first | |
353 | ancestor of the new commit. | |
354 | ||
355 | As `LF` is not valid in a Git refname or SHA-1 expression, no | |
356 | quoting or escaping syntax is supported within `<committish>`. | |
357 | ||
358 | Here `<committish>` is any of the following: | |
359 | ||
360 | * The name of an existing branch already in gfi's internal branch | |
361 | table. If gfi doesn't know the name, its treated as a SHA-1 | |
362 | expression. | |
363 | ||
364 | * A mark reference, `:<idnum>`, where `<idnum>` is the mark number. | |
365 | + | |
366 | The reason gfi uses `:` to denote a mark reference is this character | |
367 | is not legal in a Git branch name. The leading `:` makes it easy | |
368 | to distingush between the mark 42 (`:42`) and the branch 42 (`42` | |
369 | or `refs/heads/42`), or an abbreviated SHA-1 which happened to | |
370 | consist only of base-10 digits. | |
371 | + | |
372 | Marks must be declared (via `mark`) before they can be used. | |
373 | ||
374 | * A complete 40 byte or abbreviated commit SHA-1 in hex. | |
375 | ||
376 | * Any valid Git SHA-1 expression that resolves to a commit. See | |
377 | ``SPECIFYING REVISIONS'' in gitlink:git-rev-parse[1] for details. | |
378 | ||
379 | The special case of restarting an incremental import from the | |
380 | current branch value should be written as: | |
381 | ---- | |
382 | from refs/heads/branch^0 | |
383 | ---- | |
209f1298 | 384 | The `{caret}0` suffix is necessary as gfi does not permit a branch to |
6e411d20 | 385 | start from itself, and the branch is created in memory before the |
209f1298 | 386 | `from` command is even read from the input. Adding `{caret}0` will force |
6e411d20 SP |
387 | gfi to resolve the commit through Git's revision parsing library, |
388 | rather than its internal branch table, thereby loading in the | |
389 | existing value of the branch. | |
390 | ||
391 | `merge` | |
392 | ^^^^^^^ | |
393 | Includes one additional ancestor commit, and makes the current | |
394 | commit a merge commit. An unlimited number of `merge` commands per | |
395 | commit are permitted by gfi, thereby establishing an n-way merge. | |
396 | However Git's other tools never create commits with more than 15 | |
397 | additional ancestors (forming a 16-way merge). For this reason | |
398 | it is suggested that frontends do not use more than 15 `merge` | |
399 | commands per commit. | |
400 | ||
401 | Here `<committish>` is any of the commit specification expressions | |
402 | also accepted by `from` (see above). | |
403 | ||
404 | `filemodify` | |
ef94edb5 | 405 | ^^^^^^^^^^^^ |
6e411d20 SP |
406 | Included in a `commit` command to add a new file or change the |
407 | content of an existing file. This command has two different means | |
408 | of specifying the content of the file. | |
409 | ||
410 | External data format:: | |
411 | The data content for the file was already supplied by a prior | |
412 | `blob` command. The frontend just needs to connect it. | |
413 | + | |
414 | .... | |
415 | 'M' SP <mode> SP <dataref> SP <path> LF | |
416 | .... | |
417 | + | |
418 | Here `<dataref>` can be either a mark reference (`:<idnum>`) | |
419 | set by a prior `blob` command, or a full 40-byte SHA-1 of an | |
420 | existing Git blob object. | |
421 | ||
422 | Inline data format:: | |
423 | The data content for the file has not been supplied yet. | |
424 | The frontend wants to supply it as part of this modify | |
425 | command. | |
426 | + | |
427 | .... | |
428 | 'M' SP <mode> SP 'inline' SP <path> LF | |
429 | data | |
430 | .... | |
431 | + | |
432 | See below for a detailed description of the `data` command. | |
433 | ||
434 | In both formats `<mode>` is the type of file entry, specified | |
435 | in octal. Git only supports the following modes: | |
436 | ||
437 | * `100644` or `644`: A normal (not-executable) file. The majority | |
438 | of files in most projects use this mode. If in doubt, this is | |
439 | what you want. | |
440 | * `100755` or `755`: A normal, but executable, file. | |
9981b6d9 | 441 | * `120000`: A symlink, the content of the file will be the link target. |
6e411d20 SP |
442 | |
443 | In both formats `<path>` is the complete path of the file to be added | |
444 | (if not already existing) or modified (if already existing). | |
445 | ||
446 | A `<path>` string must use UNIX-style directory seperators (forward | |
447 | slash `/`), may contain any byte other than `LF`, and must not | |
448 | start with double quote (`"`). | |
449 | ||
450 | If an `LF` or double quote must be encoded into `<path>` shell-style | |
451 | quoting should be used, e.g. `"path/with\n and \" in it"`. | |
452 | ||
453 | The value of `<path>` must be in canoncial form. That is it must not: | |
454 | ||
455 | * contain an empty directory component (e.g. `foo//bar` is invalid), | |
456 | * end with a directory seperator (e.g. `foo/` is invalid), | |
457 | * start with a directory seperator (e.g. `/foo` is invalid), | |
458 | * contain the special component `.` or `..` (e.g. `foo/./bar` and | |
459 | `foo/../bar` are invalid). | |
460 | ||
461 | It is recommended that `<path>` always be encoded using UTF-8. | |
462 | ||
6e411d20 | 463 | `filedelete` |
ef94edb5 | 464 | ^^^^^^^^^^^^ |
6e411d20 SP |
465 | Included in a `commit` command to remove a file from the branch. |
466 | If the file removal makes its directory empty, the directory will | |
467 | be automatically removed too. This cascades up the tree until the | |
468 | first non-empty directory or the root is reached. | |
469 | ||
470 | .... | |
471 | 'D' SP <path> LF | |
472 | .... | |
473 | ||
474 | here `<path>` is the complete path of the file to be removed. | |
475 | See `filemodify` above for a detailed description of `<path>`. | |
476 | ||
825769a8 SP |
477 | `filedeleteall` |
478 | ^^^^^^^^^^^^^^^ | |
479 | Included in a `commit` command to remove all files (and also all | |
480 | directories) from the branch. This command resets the internal | |
481 | branch structure to have no files in it, allowing the frontend | |
482 | to subsequently add all interesting files from scratch. | |
483 | ||
484 | .... | |
485 | 'deleteall' LF | |
486 | .... | |
487 | ||
488 | This command is extremely useful if the frontend does not know | |
489 | (or does not care to know) what files are currently on the branch, | |
490 | and therefore cannot generate the proper `filedelete` commands to | |
491 | update the content. | |
492 | ||
493 | Issuing a `filedeleteall` followed by the needed `filemodify` | |
494 | commands to set the correct content will produce the same results | |
495 | as sending only the needed `filemodify` and `filedelete` commands. | |
496 | The `filedeleteall` approach may however require gfi to use slightly | |
497 | more memory per active branch (less than 1 MiB for even most large | |
498 | projects); so frontends that can easily obtain only the affected | |
499 | paths for a commit are encouraged to do so. | |
500 | ||
6e411d20 SP |
501 | `mark` |
502 | ~~~~~~ | |
503 | Arranges for gfi to save a reference to the current object, allowing | |
504 | the frontend to recall this object at a future point in time, without | |
505 | knowing its SHA-1. Here the current object is the object creation | |
506 | command the `mark` command appears within. This can be `commit`, | |
507 | `tag`, and `blob`, but `commit` is the most common usage. | |
508 | ||
509 | .... | |
510 | 'mark' SP ':' <idnum> LF | |
511 | .... | |
512 | ||
513 | where `<idnum>` is the number assigned by the frontend to this mark. | |
ef94edb5 SP |
514 | The value of `<idnum>` is expressed as an ASCII decimal integer. |
515 | The value 0 is reserved and cannot be used as | |
6e411d20 SP |
516 | a mark. Only values greater than or equal to 1 may be used as marks. |
517 | ||
518 | New marks are created automatically. Existing marks can be moved | |
519 | to another object simply by reusing the same `<idnum>` in another | |
520 | `mark` command. | |
521 | ||
522 | `tag` | |
523 | ~~~~~ | |
524 | Creates an annotated tag referring to a specific commit. To create | |
525 | lightweight (non-annotated) tags see the `reset` command below. | |
526 | ||
527 | .... | |
528 | 'tag' SP <name> LF | |
529 | 'from' SP <committish> LF | |
63e0c8b3 | 530 | 'tagger' SP <name> SP LT <email> GT SP <when> LF |
6e411d20 SP |
531 | data |
532 | LF | |
533 | .... | |
534 | ||
535 | where `<name>` is the name of the tag to create. | |
536 | ||
537 | Tag names are automatically prefixed with `refs/tags/` when stored | |
538 | in Git, so importing the CVS branch symbol `RELENG-1_0-FINAL` would | |
539 | use just `RELENG-1_0-FINAL` for `<name>`, and gfi will write the | |
540 | corresponding ref as `refs/tags/RELENG-1_0-FINAL`. | |
541 | ||
542 | The value of `<name>` must be a valid refname in Git and therefore | |
543 | may contain forward slashes. As `LF` is not valid in a Git refname, | |
544 | no quoting or escaping syntax is supported here. | |
545 | ||
546 | The `from` command is the same as in the `commit` command; see | |
547 | above for details. | |
548 | ||
549 | The `tagger` command uses the same format as `committer` within | |
550 | `commit`; again see above for details. | |
551 | ||
552 | The `data` command following `tagger` must supply the annotated tag | |
553 | message (see below for `data` command syntax). To import an empty | |
554 | tag message use a 0 length data. Tag messages are free-form and are | |
555 | not interpreted by Git. Currently they must be encoded in UTF-8, | |
556 | as gfi does not permit other encodings to be specified. | |
557 | ||
558 | Signing annotated tags during import from within gfi is not | |
559 | supported. Trying to include your own PGP/GPG signature is not | |
560 | recommended, as the frontend does not (easily) have access to the | |
561 | complete set of bytes which normally goes into such a signature. | |
562 | If signing is required, create lightweight tags from within gfi with | |
563 | `reset`, then create the annotated versions of those tags offline | |
564 | with the standard gitlink:git-tag[1] process. | |
565 | ||
566 | `reset` | |
567 | ~~~~~~~ | |
568 | Creates (or recreates) the named branch, optionally starting from | |
569 | a specific revision. The reset command allows a frontend to issue | |
570 | a new `from` command for an existing branch, or to create a new | |
571 | branch from an existing commit without creating a new commit. | |
572 | ||
573 | .... | |
574 | 'reset' SP <ref> LF | |
575 | ('from' SP <committish> LF)? | |
576 | LF | |
577 | .... | |
578 | ||
579 | For a detailed description of `<ref>` and `<committish>` see above | |
580 | under `commit` and `from`. | |
581 | ||
582 | The `reset` command can also be used to create lightweight | |
583 | (non-annotated) tags. For example: | |
584 | ||
585 | ==== | |
586 | reset refs/tags/938 | |
587 | from :938 | |
588 | ==== | |
589 | ||
590 | would create the lightweight tag `refs/tags/938` referring to | |
591 | whatever commit mark `:938` references. | |
592 | ||
593 | `blob` | |
594 | ~~~~~~ | |
595 | Requests writing one file revision to the packfile. The revision | |
596 | is not connected to any commit; this connection must be formed in | |
597 | a subsequent `commit` command by referencing the blob through an | |
598 | assigned mark. | |
599 | ||
600 | .... | |
601 | 'blob' LF | |
602 | mark? | |
603 | data | |
604 | .... | |
605 | ||
606 | The mark command is optional here as some frontends have chosen | |
607 | to generate the Git SHA-1 for the blob on their own, and feed that | |
608 | directly to `commit`. This is typically more work than its worth | |
609 | however, as marks are inexpensive to store and easy to use. | |
610 | ||
611 | `data` | |
612 | ~~~~~~ | |
613 | Supplies raw data (for use as blob/file content, commit messages, or | |
614 | annotated tag messages) to gfi. Data can be supplied using an exact | |
615 | byte count or delimited with a terminating line. Real frontends | |
616 | intended for production-quality conversions should always use the | |
617 | exact byte count format, as it is more robust and performs better. | |
618 | The delimited format is intended primarily for testing gfi. | |
619 | ||
ef94edb5 SP |
620 | Exact byte count format:: |
621 | The frontend must specify the number of bytes of data. | |
622 | + | |
6e411d20 SP |
623 | .... |
624 | 'data' SP <count> LF | |
625 | <raw> LF | |
626 | .... | |
ef94edb5 | 627 | + |
6e411d20 | 628 | where `<count>` is the exact number of bytes appearing within |
ef94edb5 SP |
629 | `<raw>`. The value of `<count>` is expressed as an ASCII decimal |
630 | integer. The `LF` on either side of `<raw>` is not | |
6e411d20 SP |
631 | included in `<count>` and will not be included in the imported data. |
632 | ||
ef94edb5 SP |
633 | Delimited format:: |
634 | A delimiter string is used to mark the end of the data. | |
635 | gfi will compute the length by searching for the delimiter. | |
636 | This format is primarly useful for testing and is not | |
637 | recommended for real data. | |
638 | + | |
6e411d20 SP |
639 | .... |
640 | 'data' SP '<<' <delim> LF | |
641 | <raw> LF | |
642 | <delim> LF | |
643 | .... | |
ef94edb5 | 644 | + |
6e411d20 SP |
645 | where `<delim>` is the chosen delimiter string. The string `<delim>` |
646 | must not appear on a line by itself within `<raw>`, as otherwise | |
647 | gfi will think the data ends earlier than it really does. The `LF` | |
648 | immediately trailing `<raw>` is part of `<raw>`. This is one of | |
649 | the limitations of the delimited format, it is impossible to supply | |
650 | a data chunk which does not have an LF as its last byte. | |
651 | ||
652 | `checkpoint` | |
653 | ~~~~~~~~~~~~ | |
820b9310 SP |
654 | Forces gfi to close the current packfile, start a new one, and to |
655 | save out all current branch refs, tags and marks. | |
6e411d20 SP |
656 | |
657 | .... | |
658 | 'checkpoint' LF | |
659 | LF | |
660 | .... | |
661 | ||
820b9310 SP |
662 | Note that gfi automatically switches packfiles when the current |
663 | packfile reaches \--max-pack-size, or 4 GiB, whichever limit is | |
664 | smaller. During an automatic packfile switch gfi does not update | |
665 | the branch refs, tags or marks. | |
666 | ||
667 | As a `checkpoint` can require a significant amount of CPU time and | |
668 | disk IO (to compute the overall pack SHA-1 checksum, generate the | |
669 | corresponding index file, and update the refs) it can easily take | |
670 | several minutes for a single `checkpoint` command to complete. | |
671 | ||
672 | Frontends may choose to issue checkpoints during extremely large | |
673 | and long running imports, or when they need to allow another Git | |
674 | process access to a branch. However given that a 30 GiB Subversion | |
675 | repository can be loaded into Git through gfi in about 3 hours, | |
676 | explicit checkpointing may not be necessary. | |
677 | ||
678 | ||
bdd9f424 SP |
679 | Tips and Tricks |
680 | --------------- | |
681 | The following tips and tricks have been collected from various | |
682 | users of gfi, and are offered here as suggestions. | |
683 | ||
684 | Use One Mark Per Commit | |
685 | ~~~~~~~~~~~~~~~~~~~~~~~ | |
686 | When doing a repository conversion, use a unique mark per commit | |
687 | (`mark :<n>`) and supply the \--export-marks option on the command | |
688 | line. gfi will dump a file which lists every mark and the Git | |
689 | object SHA-1 that corresponds to it. If the frontend can tie | |
690 | the marks back to the source repository, it is easy to verify the | |
691 | accuracy and completeness of the import by comparing each Git | |
692 | commit to the corresponding source revision. | |
693 | ||
694 | Coming from a system such as Perforce or Subversion this should be | |
695 | quite simple, as the gfi mark can also be the Perforce changeset | |
696 | number or the Subversion revision number. | |
697 | ||
698 | Freely Skip Around Branches | |
699 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~ | |
700 | Don't bother trying to optimize the frontend to stick to one branch | |
701 | at a time during an import. Although doing so might be slightly | |
702 | faster for gfi, it tends to increase the complexity of the frontend | |
703 | code considerably. | |
704 | ||
705 | The branch LRU builtin to gfi tends to behave very well, and the | |
706 | cost of activating an inactive branch is so low that bouncing around | |
707 | between branches has virtually no impact on import performance. | |
708 | ||
709 | Use Tag Fixup Branches | |
710 | ~~~~~~~~~~~~~~~~~~~~~~ | |
711 | Some other SCM systems let the user create a tag from multiple | |
712 | files which are not from the same commit/changeset. Or to create | |
713 | tags which are a subset of the files available in the repository. | |
714 | ||
715 | Importing these tags as-is in Git is impossible without making at | |
716 | least one commit which ``fixes up'' the files to match the content | |
717 | of the tag. Use gfi's `reset` command to reset a dummy branch | |
718 | outside of your normal branch space to the base commit for the tag, | |
719 | then commit one or more file fixup commits, and finally tag the | |
720 | dummy branch. | |
721 | ||
722 | For example since all normal branches are stored under `refs/heads/` | |
723 | name the tag fixup branch `TAG_FIXUP`. This way it is impossible for | |
724 | the fixup branch used by the importer to have namespace conflicts | |
725 | with real branches imported from the source (the name `TAG_FIXUP` | |
726 | is not `refs/heads/TAG_FIXUP`). | |
727 | ||
728 | When committing fixups, consider using `merge` to connect the | |
729 | commit(s) which are supplying file revisions to the fixup branch. | |
730 | Doing so will allow tools such as gitlink:git-blame[1] to track | |
731 | through the real commit history and properly annotate the source | |
732 | files. | |
733 | ||
734 | After gfi terminates the frontend will need to do `rm .git/TAG_FIXUP` | |
735 | to remove the dummy branch. | |
736 | ||
737 | Import Now, Repack Later | |
738 | ~~~~~~~~~~~~~~~~~~~~~~~~ | |
739 | As soon as gfi completes the Git repository is completely valid | |
740 | and ready for use. Typicallly this takes only a very short time, | |
741 | even for considerably large projects (100,000+ commits). | |
742 | ||
743 | However repacking the repository is necessary to improve data | |
744 | locality and access performance. It can also take hours on extremely | |
745 | large projects (especially if -f and a large \--window parameter is | |
746 | used). Since repacking is safe to run alongside readers and writers, | |
747 | run the repack in the background and let it finish when it finishes. | |
748 | There is no reason to wait to explore your new Git project! | |
749 | ||
750 | If you choose to wait for the repack, don't try to run benchmarks | |
751 | or performance tests until repacking is completed. gfi outputs | |
752 | suboptimal packfiles that are simply never seen in real use | |
753 | situations. | |
754 | ||
755 | Repacking Historical Data | |
756 | ~~~~~~~~~~~~~~~~~~~~~~~~~ | |
757 | If you are repacking very old imported data (e.g. older than the | |
758 | last year), consider expending some extra CPU time and supplying | |
759 | \--window=50 (or higher) when you run gitlink:git-repack[1]. | |
760 | This will take longer, but will also produce a smaller packfile. | |
761 | You only need to expend the effort once, and everyone using your | |
762 | project will benefit from the smaller repository. | |
763 | ||
764 | ||
6e411d20 SP |
765 | Packfile Optimization |
766 | --------------------- | |
767 | When packing a blob gfi always attempts to deltify against the last | |
768 | blob written. Unless specifically arranged for by the frontend, | |
769 | this will probably not be a prior version of the same file, so the | |
770 | generated delta will not be the smallest possible. The resulting | |
771 | packfile will be compressed, but will not be optimal. | |
772 | ||
773 | Frontends which have efficient access to all revisions of a | |
774 | single file (for example reading an RCS/CVS ,v file) can choose | |
775 | to supply all revisions of that file as a sequence of consecutive | |
776 | `blob` commands. This allows gfi to deltify the different file | |
777 | revisions against each other, saving space in the final packfile. | |
778 | Marks can be used to later identify individual file revisions during | |
779 | a sequence of `commit` commands. | |
780 | ||
781 | The packfile(s) created by gfi do not encourage good disk access | |
782 | patterns. This is caused by gfi writing the data in the order | |
783 | it is received on standard input, while Git typically organizes | |
784 | data within packfiles to make the most recent (current tip) data | |
785 | appear before historical data. Git also clusters commits together, | |
786 | speeding up revision traversal through better cache locality. | |
787 | ||
788 | For this reason it is strongly recommended that users repack the | |
789 | repository with `git repack -a -d` after gfi completes, allowing | |
790 | Git to reorganize the packfiles for faster data access. If blob | |
791 | deltas are suboptimal (see above) then also adding the `-f` option | |
792 | to force recomputation of all deltas can significantly reduce the | |
793 | final packfile size (30-50% smaller can be quite typical). | |
794 | ||
bdd9f424 | 795 | |
6e411d20 SP |
796 | Memory Utilization |
797 | ------------------ | |
798 | There are a number of factors which affect how much memory gfi | |
799 | requires to perform an import. Like critical sections of core | |
800 | Git, gfi uses its own memory allocators to ammortize any overheads | |
801 | associated with malloc. In practice gfi tends to ammoritize any | |
802 | malloc overheads to 0, due to its use of large block allocations. | |
803 | ||
804 | per object | |
805 | ~~~~~~~~~~ | |
806 | gfi maintains an in-memory structure for every object written in | |
807 | this execution. On a 32 bit system the structure is 32 bytes, | |
808 | on a 64 bit system the structure is 40 bytes (due to the larger | |
809 | pointer sizes). Objects in the table are not deallocated until | |
810 | gfi terminates. Importing 2 million objects on a 32 bit system | |
811 | will require approximately 64 MiB of memory. | |
812 | ||
813 | The object table is actually a hashtable keyed on the object name | |
814 | (the unique SHA-1). This storage configuration allows gfi to reuse | |
815 | an existing or already written object and avoid writing duplicates | |
816 | to the output packfile. Duplicate blobs are surprisingly common | |
817 | in an import, typically due to branch merges in the source. | |
818 | ||
819 | per mark | |
820 | ~~~~~~~~ | |
821 | Marks are stored in a sparse array, using 1 pointer (4 bytes or 8 | |
822 | bytes, depending on pointer size) per mark. Although the array | |
823 | is sparse, frontends are still strongly encouraged to use marks | |
824 | between 1 and n, where n is the total number of marks required for | |
825 | this import. | |
826 | ||
827 | per branch | |
828 | ~~~~~~~~~~ | |
829 | Branches are classified as active and inactive. The memory usage | |
830 | of the two classes is significantly different. | |
831 | ||
832 | Inactive branches are stored in a structure which uses 96 or 120 | |
833 | bytes (32 bit or 64 bit systems, respectively), plus the length of | |
834 | the branch name (typically under 200 bytes), per branch. gfi will | |
835 | easily handle as many as 10,000 inactive branches in under 2 MiB | |
836 | of memory. | |
837 | ||
838 | Active branches have the same overhead as inactive branches, but | |
839 | also contain copies of every tree that has been recently modified on | |
840 | that branch. If subtree `include` has not been modified since the | |
841 | branch became active, its contents will not be loaded into memory, | |
842 | but if subtree `src` has been modified by a commit since the branch | |
843 | became active, then its contents will be loaded in memory. | |
844 | ||
845 | As active branches store metadata about the files contained on that | |
846 | branch, their in-memory storage size can grow to a considerable size | |
847 | (see below). | |
848 | ||
849 | gfi automatically moves active branches to inactive status based on | |
850 | a simple least-recently-used algorithm. The LRU chain is updated on | |
851 | each `commit` command. The maximum number of active branches can be | |
c499d768 | 852 | increased or decreased on the command line with \--active-branches=. |
6e411d20 SP |
853 | |
854 | per active tree | |
855 | ~~~~~~~~~~~~~~~ | |
856 | Trees (aka directories) use just 12 bytes of memory on top of the | |
857 | memory required for their entries (see ``per active file'' below). | |
858 | The cost of a tree is virtually 0, as its overhead ammortizes out | |
859 | over the individual file entries. | |
860 | ||
861 | per active file entry | |
862 | ~~~~~~~~~~~~~~~~~~~~~ | |
863 | Files (and pointers to subtrees) within active trees require 52 or 64 | |
864 | bytes (32/64 bit platforms) per entry. To conserve space, file and | |
865 | tree names are pooled in a common string table, allowing the filename | |
866 | ``Makefile'' to use just 16 bytes (after including the string header | |
867 | overhead) no matter how many times it occurs within the project. | |
868 | ||
869 | The active branch LRU, when coupled with the filename string pool | |
870 | and lazy loading of subtrees, allows gfi to efficiently import | |
871 | projects with 2,000+ branches and 45,114+ files in a very limited | |
872 | memory footprint (less than 2.7 MiB per active branch). | |
873 | ||
874 | ||
875 | Author | |
876 | ------ | |
877 | Written by Shawn O. Pearce <spearce@spearce.org>. | |
878 | ||
879 | Documentation | |
880 | -------------- | |
881 | Documentation by Shawn O. Pearce <spearce@spearce.org>. | |
882 | ||
883 | GIT | |
884 | --- | |
885 | Part of the gitlink:git[7] suite | |
886 |