1
.. -*- mode: indented-text; compile-command: "make -C doc" -*-
9
See also various low-level TODOs in the source code. Try looking in
10
the list archive or on gmane.org for previous discussion of these
13
These are classified by approximate size: an hour or less, a day or
14
less, and several days or more.
20
* Merging add of a new file clashing with an existing file doesn't
21
work; add gets an error that it's already versioned and the merge
24
* Merge should ignore the destination's working directory, otherwise
25
we get an error about the statcache when pulling from a remote
28
* Add of a file that was present in the base revision should put back
31
* Handle diff of files which do not have a trailing newline; probably
32
requires patching difflib to get it exactly right, or otherwise
33
calling out to GNU diff.
35
* Should be able to copy files between branches to preserve their
36
file-id (and perhaps eventually parentage.)
38
* -r option should take a revision-id as well as a revno.
40
* allow ``bzr st -r 300`` to show a summary of changes since then.
42
* ``bzr info`` should count only people with distinct email addresses as
43
different committers. (Or perhaps only distinct userids?)
45
* On Windows, command-line arguments should be `glob-expanded`__,
46
because the shell doesn't do this. However, there are probably some
47
commands where this shouldn't be done, such as 'bzr ignore', because
48
we want to accept globs.
50
* ``bzr ignore`` command that just adds a line to the ``.bzrignore`` file
51
and makes it versioned. Fix this to break symlinks.
53
* Any useful sanity checks in 'bzr ignore'? Perhaps give a warning if
54
they try to add a single file which is already versioned, or if they
55
add a pattern which already exists, or if it looks like they gave an
58
__ http://mail.python.org/pipermail/python-list/2001-April/037847.html
60
* Separate read and write version checks?
62
* ``bzr status DIR`` should give status on all files under that
65
* ``bzr log DIR`` should give changes to any files within DIR.
67
* ``bzr inventory -r REV`` and perhaps unify this with ``bzr ls``,
68
giving options to display ids, types, etc.
70
* Split BzrError into various more specific subclasses for different
71
errors people might want to catch.
73
* If the export destination ends in '.tar', '.tar.gz', etc then create
74
a tarball instead of a directory. (Need to actually make a
75
temporary directory and then tar that up.)
77
http://www.gelato.unsw.edu.au/archives/git/0504/2194.html
79
* RemoteBranch could maintain a cache either in memory or on disk. We
80
know more than an external cache might about which files are
81
immutable and which can vary. On the other hand, it's much simpler
82
to just use an external proxy cache.
84
Perhaps ~/.bzr/http-cache. Baz has a fairly simple cache under
85
~/.arch-cache, containing revision information encoded almost as a
86
bunch of archives. Perhaps we could simply store full paths.
88
* Maybe also store directories in the statcache so that we can quickly
89
identify that they still exist.
91
* Diff should show timestamps; for files from the working directory we
92
can use the file itself; for files from a revision we should use the
93
commit time of the revision.
95
* Perhaps split command infrastructure from the actual command
98
* Cleaner support for negative boolean options like --no-recurse.
100
* Statcache should possibly map all file paths to / separators
102
* quotefn doubles all backslashes on Windows; this is probably not the
103
best thing to do. What would be a better way to safely represent
104
filenames? Perhaps we could doublequote things containing spaces,
105
on the principle that filenames containing quotes are unlikely?
106
Nice for humans; less good for machine parsing.
108
* Patches should probably use only forward slashes, even on Windows,
109
otherwise Unix patch can't apply them. (?)
111
* Branch.update_revisions() inefficiently fetches revisions from the
112
remote server twice; once to find out what text and inventory they
113
need and then again to actually get the thing. This is a bit
116
One complicating factor here is that we don't really want to have
117
revisions present in the revision-store until all their constituent
118
parts are also stored.
120
The basic problem is that RemoteBranch.get_revision() and similar
121
methods return object, but what we really want is the raw XML, which
122
can be popped into our own store. That needs to be refactored.
124
* ``bzr status FOO`` where foo is ignored should say so.
126
* ``bzr mkdir A...`` should just create and add A.
128
* Guard against repeatedly merging any particular patch.
135
* Merge revert patch.
137
* ``bzr mv`` that does either rename or move as in Unix.
139
* More efficient diff of only selected files. We should be able to
140
just get the id for the selected files, look up their location and
141
diff just those files. No need to traverse the entire inventories.
143
* ``bzr status DIR`` or ``bzr diff DIR`` should report on all changes
144
under that directory.
146
* Fix up Inventory objects to represent root object as an entry.
148
* Don't convert entire entry from ElementTree to an object when it is
149
read in, but rather wait until the program actually wants to know
152
* Extract changes from one revision to the next to a text form
153
suitable for transmission over email.
157
- Selected-file commit
159
- Impossible selected-file commit: adding things in non-versioned
160
directories, crossing renames, etc.
162
* Write a reproducible benchmark, perhaps importing various kernel versions.
164
* Directly import diffs! It seems a bit redundant to need to rescan
165
the directory to work out what files diff added/deleted/changed when
166
all the information is there in the diff in the first place.
167
Getting the exact behaviour for added/deleted subdirectories etc
170
At the very least we could run diffstat over the diff, or perhaps
171
read the status output from patch. Just knowing which files might
172
be modified would be enough to guide the add and commit.
174
Given this we might be able to import patches at 1/second or better.
176
* Get branch over http.
178
* Pull pure updates over http.
180
* revfile compression.
182
* Split inventory into per-directory files.
184
* Fix ignore file parsing:
186
- fnmatch is not the same as unix patterns
188
- perhaps add extended globs from rsh/rsync
190
- perhaps a pattern that matches only directories or non-directories
192
* Consider using Python logging library as well as/instead of
195
* Commands should give some progress indication by default.
197
- But quieten this with ``--silent``.
199
* Change to using gettext message localization.
201
* Make a clearer separation between internal and external bzrlib
202
interfaces. Make internal interfaces use protected names. Write at
203
least some documentation for those APIs, probably as docstrings.
205
Consider using ZopeInterface definitions for the external interface;
206
I think these are already used in PyBaz. They allow automatic
207
checking of the interface but may be unfamiliar to general Python
208
developers, so I'm not really keen.
210
* Commands to dump out all command help into a manpage or HTML file or
213
* Handle symlinks in the working directory; at the very least it
214
should be possible for them to be present and ignored/unknown
215
without causing assertion failures.
217
Eventually symlinks should be versioned.
219
* Allow init in a subdirectory to create a nested repository, but only
220
if the subdirectory is not already versioned. Perhaps also require
221
a ``--nested`` to protect against confusion.
225
* More test framework:
227
- Class that describes the state of a working tree so we can just
230
* There are too many methods on Branch() that really manipulate the
231
WorkingTree. They should be moved across.
233
Also there are some methods which are duplicated on Tree and
234
Inventory objects, and it should be made more clear which ones are
235
proxies and which ones behave differently, and how.
237
* Try using XSLT to add some formatting to REST-generated HTML. Or
238
maybe write a small Python program that specifies a header and foot
239
for the pages and calls into the docutils libraries.
241
* --format=xml for log, status and other commands.
243
* Attempting to explicitly add a file that's already added should give
244
a warning; however there should be no warning for directories (since
245
we scan for new children) or files encountered in a directory that's
248
* Better handling of possible collisions on case-losing filesystems;
249
make sure a single file does not get added twice under different
252
* Clean up XML inventory:
254
- Use nesting rather than parent_id pointers.
256
- Hold the ElementTree in memory in the Inventory object and work
257
directly on that, rather than converting into Python objects every
258
time it is read in. Probably still exposoe it through some kind of
259
object interface though, but perhaps that should just be a proxy
262
- Less special cases for the root directory.
264
* Perhaps inventories should remember the revision in which each file
265
was last changed, as well as its current state? This is a bit
266
redundant but might often be interested to know.
268
* stat cache should perhaps only stat files as necessary, rather than
269
doing them all up-front. On the other hand, that disallows the
270
opimization of stating them in inode order.
272
* It'd be nice to pipeline multiple HTTP requests. Often we can
273
predict what will be wanted in future: all revisions, or all texts
274
in a particular revision, etc.
276
urlgrabber's docs say they are working on batched downloads; we
277
could perhaps ride on that or just create a background thread (ew).
279
* Paranoid mode where we never trust SHA-1 matches.
281
* Don't commit if there are no changes unless forced.
283
* --dry-run mode for commit? (Or maybe just run with
284
check-command=false?)
286
* Generally, be a bit more verbose unless --silent is specified.
288
* Function that finds all changes to files under a given directory;
289
perhaps log should use this if a directory is given.
291
* XML attributes might have trouble with filenames containing \n and
292
\r. Do we really want to support this? I think perhaps not.
294
* Remember execute bits, so that exports will work OK.
296
* Unify smart_add and plain Branch.add(); perhaps smart_add should
297
just build a list of files to add and pass that to the regular add
300
* Function to list a directory, saying in which revision each file was
301
last modified. Useful for web and gui interfaces, and slow to
302
compute one file at a time.
304
* unittest is standard, but the results are kind of ugly; would be
305
nice to make it cleaner.
307
* Check locking is correct during merge-related operations.
309
* Perhaps attempts to get locks should timeout after some period of
310
time, or at least display a progress message.
312
* Split out upgrade functionality from check command into a separate
315
* Don't pass around command classes but rather pass objects. This'd
316
make it cleaner to construct objects wrapping external commands.
318
* Track all merged-in revisions in a versioned add-only metafile.
324
* Generate annotations from current file relative to previous
327
- Is it necessary to store any kind of annotation where data was
330
* Update revfile_ format and make it active:
332
- Texts should be identified by something keyed on the revision, not
333
an individual text-id. This is much more useful for annotate I
334
think; we want to map back to the revision that last changed it.
336
- Access revfile revisions through the Tree/Store classes.
338
- Check them from check commands.
342
.. _revfile: revfile.html
344
* Hooks for pre-commit, post-commit, etc.
346
Consider the security implications; probably should not enable hooks
347
for remotely-fetched branches by default.
349
* Pre-commit check. If this hook is defined, it needs to be handled
350
specially: create a temporary directory containing the tree as it
351
will be after the commit. This means excluding any ignored/unknown
352
files, and respecting selective commits. Run the pre-commit check
353
(e.g. compile and run test suite) in there.
355
Possibly this should be done by splitting the commit function into
356
several parts (under a single interface). It is already rather
357
large. Decomposition:
359
- find tree modifications and prepare in-memory inventory
361
- export that inventory to a temporary directory
363
- run the test in that temporary directory
365
- if that succeeded, continue to actually finish the commit
367
What should be done with the text of modified files while this is
368
underway? I don't think we want to count on holding them in memory
369
and we can't trust the working files to stay in one place so I
370
suppose we need to move them into the text store, or otherwise into
371
a temporary directory.
373
If the commit does not actually complete, we would rather the
374
content was not left behind in the stores.
378
* GUI (maybe in Python GTK+?)
380
* C library interface
382
* Expansion of $Id$ keywords within working files. Perhaps do this in
383
exports first as a simpler case because then we don't need to deal
384
with removing the tags on the way back in.
added
removed
TODO
