JustAnotherArchivist
543c6b0595
Use temporary directory for Git clone directory
The previous approach was flawed and broke on URLs ending with a slash.
1 year ago
JustAnotherArchivist
47fe0a4e70
Handle URLs with queries and fragments
1 year ago
JustAnotherArchivist
e08919d89f
Fix crash on incremental bundling with warnings
For example, if the HEAD is excluded:
warning: ref 'refs/heads/master' is excluded by the rev-list options
warning: ref 'HEAD' is excluded by the rev-list options
fatal: Refusing to create empty bundle.
The fatal message always appears last (though that's of course undocumented).
1 year ago
JustAnotherArchivist
0da610744c
Keep a record of what HEAD points at
1 year ago
JustAnotherArchivist
9de50bebdb
Fix metadata parsing on field values containing a colon
1 year ago
JustAnotherArchivist
e39548c50b
Fix file extensions
1 year ago
JustAnotherArchivist
7861036624
Add submodule check
#13
1 year ago
JustAnotherArchivist
9a61800758
Refactor Git bundling to allow for verification of the bundle contents
This verifies that all objects from the current clone are in either the dependency bundles or the current bundle. This guarantees that the repo as it has been clone at the time of retrieval can be reconstructed exactly from the bundles.
As a side-effect, if a non-standard Git server were to include objects in a clone pack that are not discoverable from refs, this will fail any attempt to archive such a clone. This could in the future be resolved by adding custom refs for those extra objects.
This also fixes a bug where prior bundles could be included as a dependency even though they contain no relevant data due to their refs (as refs are always listed in the bundle metadata). Instead, dependency detection now operates directly on commit and tag objects, which can only be present in one bundle.
1 year ago
JustAnotherArchivist
f5fe0496f5
Add support for supplying a file-like object as stdin
1 year ago
JustAnotherArchivist
adafd6bd01
Fix race condition in subprocess runner
stdin, stdout, and stderr being closed does not necessarily imply that the process has exited, although it usually does. Still need to explicitly wait for it to terminate after the I/O loop. This matches what the stdlib `subprocess.Popen._communicate` does as well.
1 year ago
JustAnotherArchivist
9a0c8398de
Document minimum Git version
1 year ago
JustAnotherArchivist
3ca99d8839
Require `Storage.search_metadata` to return files in lexicographical order to minimise dependencies between bundles
1 year ago
JustAnotherArchivist
cc7bdbb3f4
Fix tag objects not getting deduplicated
1 year ago
JustAnotherArchivist
f1edf4b752
Fix TypeError due to lack of `glob.glob`'s `root_dir` option on Python 3.9
1 year ago
JustAnotherArchivist
4d6a423fb5
Replace hacky module importing (taken from snscrape commit aa7d7d3d)
1 year ago
JustAnotherArchivist
7eb175fb63
Document how inheritance on Metadata classes works
1 year ago
JustAnotherArchivist
a361fe54e5
Add a metadata version field
1 year ago
JustAnotherArchivist
fb8af13c15
Return all metadata validation errors at the same time
1 year ago
JustAnotherArchivist
811e119835
Add retrieval start/end time metadata fields
1 year ago
JustAnotherArchivist
b0505f94fe
Fix typo in package name
1 year ago
JustAnotherArchivist
eab6db9f27
Better storage metadata search now that the module name is recorded there anyway
1 year ago
JustAnotherArchivist
fa4b60225c
Index → Metadata
'Index' was a misnomer from the start since it contains critical information for the operation that can't be reconstructed (e.g. existing refs).
1 year ago
JustAnotherArchivist
4259d34ec8
Set default ID
1 year ago
JustAnotherArchivist
d5891c795c
More metadata
1 year ago
JustAnotherArchivist
25792d9006
Fix missing inheritance from abc.ABCMeta
1 year ago
JustAnotherArchivist
a910d4851c
Add support for inheritance of index fields; change type of field list to a tuple to lessen the risk of modification
1 year ago
JustAnotherArchivist
2779148a1b
Add .gitignore
1 year ago
JustAnotherArchivist
d5a7d39f74
setup.py → pyproject.toml
1 year ago
JustAnotherArchivist
80995bccde
Add comment about FETCH_HEAD
1 year ago
JustAnotherArchivist
2a9ff2ee15
Support empty incremental bundles
1 year ago
JustAnotherArchivist
0e7b17d3fd
Capture and return stderr
1 year ago
JustAnotherArchivist
a6e256c58f
Fix invalid usage of codearchiver.subprocess
Introduced by 240dcceb
1 year ago
JustAnotherArchivist
8e83c9b7b4
Support incremental Git bundles
Also fix a small discrepancy between the commit list and bundle due to --reflog vs --all
1 year ago
JustAnotherArchivist
021b26973b
Fix handling empty input
1 year ago
JustAnotherArchivist
ed69ba16c9
logger → _logger
1 year ago
JustAnotherArchivist
6f7a95d289
Add --progress option to cloning for more details
1 year ago
JustAnotherArchivist
42e420ad0d
Disable prompts on password-protected repos
1 year ago
JustAnotherArchivist
a9e838adde
Raise exception if file already exists in DirectoryStorage target
1 year ago
JustAnotherArchivist
6af07cb51c
Raise exceptions on fatal errors
1 year ago
JustAnotherArchivist
2257305872
Disallow underscores in module names
Using the preferred file naming scheme of {moduleName}_{someInputURLDerivative}_{date}*, this allows mapping files to modules without ambiguity.
1 year ago
JustAnotherArchivist
4dcac08585
Fix import order
1 year ago
JustAnotherArchivist
0f1f5abc64
Add indices for files
1 year ago
JustAnotherArchivist
e3da8c7736
Use generic alias types
This requires at least Python 3.9.
1 year ago
JustAnotherArchivist
f2d2df9428
Simplify storage design; there is no need for the queue
1 year ago
JustAnotherArchivist
550afa8644
Add storage abstraction
1 year ago
JustAnotherArchivist
06daea162f
Remove GitHub module as it is not ready for use yet
1 year ago
JustAnotherArchivist
240dcceb10
Add subprocess wrapper for logging stderr
1 year ago
JustAnotherArchivist
6fb0ac4e5e
Initial GitHub module only retrieving the actual repository
3 years ago
JustAnotherArchivist
2a2c9373d0
Documentation of the core
3 years ago
JustAnotherArchivist
715420e298
Fix imports in CLI: core and modules aren't needed in the argument parser
3 years ago