#238 closed enhancement (fixed)
Support for downloading tarballs from the repository
Reported by: | Owned by: | Christian Boos | |
---|---|---|---|
Priority: | normal | Milestone: | 0.10 |
Component: | version control/browser | Version: | 0.6 |
Severity: | minor | Keywords: | objectstore tracdiff |
Cc: | Alec Thomas | Branch: | |
Release Notes: | |||
API Changes: | |||
Internal Changes: |
Description
It would be cool to be able to browse the repository and download a tarball of a branch or tag.
Attachments (0)
Change History (26)
comment:1 by , 21 years ago
comment:2 by , 21 years ago
Milestone: | → 0.8 |
---|---|
Priority: | normal → lowest |
comment:3 by , 20 years ago
Milestone: | 0.8 → 0.9 |
---|---|
Summary: | Support for extracting tarballs from the repository → Support for downloading tarballs from the repository |
comment:5 by , 20 years ago
Keywords: | helpwanted added |
---|
This shouldn't be so hard… have a look at the zip file support in Changeset.py for an example of how to package up files in a ZIP file.
comment:6 by , 20 years ago
#1504 has been marked as duplicate of this ticket (although it limits the request to files as opposed to whole directories). It also contains a patch for enabling this.
comment:8 by , 20 years ago
Owner: | changed from | to
---|---|
Status: | new → assigned |
This is implemented in the TracDiff branch.
comment:9 by , 19 years ago
Milestone: | 0.9 → 1.0 |
---|---|
Priority: | lowest → normal |
Severity: | normal → minor |
comment:11 by , 19 years ago
mgood suggested that this feature could put some load on the server. While that's certainly not an issue with small repositories, that can be a problem for big repositories, so it should be possible to disable this feature:
[browser] download_archive = false
However, a better solution might be to allow the download of some selected paths and cache the corresponding .zip files:
[browser] restrict_download = tags/trac-0.8.1, tags/trac-0.8.4, tags/trac-0.9b1
Of course, the revision should be taken into account, which may complicate matters (e.g. cache a maximum of 5 different revisions for each specified path).
comment:12 by , 19 years ago
Keywords: | tracdiff added |
---|
comment:13 by , 19 years ago
I don't know if it would be a good idea to restrict the downloadable directories in the configuration file, since than you would have information in there about the structure of your repository. IMHO it would be better to put the information which directories are downloadable/not downloadable in the repo. itself. Properties would be perfect for this.
comment:14 by , 19 years ago
Milestone: | 1.0 → 0.10 |
---|
answering ttest: not every VersioningSystemBackend has support for properties (e.g. Mercurial doesn't)
comment:15 by , 19 years ago
Priority: | normal → high |
---|
comment:16 by , 19 years ago
Priority: | high → normal |
---|
comment:17 by , 19 years ago
Note that the feature is now available in trunk. I leave this open until there's some consensus about the form that should take the "control" of this feature. What we have so far:
- usage of properties is ruled out because that would be specific to TracSubversion
- some possible TracIni settingss:
enable_download
for enabling or not that featurerestrict_download
using a glob pattern to define the paths where a download is possible. e.g.- "*":: everywhere
- "tags/*":: only below the tags folder
- "":: nowhere
comment:18 by , 19 years ago
Implementation of the previous idea (list of glob patterns)
Index: trac/versioncontrol/web_ui/browser.py =================================================================== --- trac/versioncontrol/web_ui/browser.py (revision 3006) +++ trac/versioncontrol/web_ui/browser.py (working copy) @@ -17,6 +17,8 @@ import re import urllib +import os.path +from fnmatch import fnmatchcase from trac import util from trac.util import sorted @@ -154,7 +156,10 @@ req.hdf['browser.items'] = info req.hdf['browser.changes'] = changes - if node.path != '': + patterns = self.config.get('browser', 'downloadable_paths') + if node.path and patterns and \ + filter(None, [fnmatchcase(node.path, p) for p in + patterns.split(os.path.pathsep)]): zip_href = self.env.href.changeset(rev, node.path, old=rev, old_path='/', # special case #238 format='zip')
Of course, the most difficult here is to find a sensible configuration setting name and a sensible default:
- "*": all folders are downloadable
- "": none are downloadable
comment:20 by , 19 years ago
This feature will probably be used most in root level directories (trunk, branches/0.x-stable, tags/…), not all over the place. In the case where it might be used in a broader fashion (e.g. TrackHacks), it means there are probably a lot of smaller projects and less overhead to create new archives as needed.
I think the globbing is overkill.
How about:
[browser] enable_archive_download = true
comment:21 by , 19 years ago
The problem is that there are no fixed rules for identifying what is a root level directory…
E.g. you would like to make each release (/tags/release-x.y.z
)
downloadable, but certainly not the whole /tags
folder,
which might contain dozens of releases…
Other example, in Trac hacks, the glob pattern would be /*/*
,
i.e. anything at the second level, like /authformplugin/0.9
but not /authformplugin
itself.
IMHO, a global yes/no flag is useless… Sure you may want to disable the feature completely, but when you want to have it, you want to have some reasonable control over it, as examplified above.
With the glob patterns, complete disabling is very easy, simply leave the option empty, and if control is needed, I can't think of a better compromise between simplicity and flexibility than a list of glob patterns…
comment:22 by , 19 years ago
I agree, globs seem like an excellent choice. I certainly wouldn't want people downloading the entire TracHacks repository.
comment:24 by , 19 years ago
Keywords: | objectstore added; helpwanted removed |
---|
Note that in the future, it might be worth to use a cache for the zip files, as they can be costly to create.
comment:25 by , 19 years ago
I've updated to trunk on TracHacks and this is working beautifully, thanks Christian :)
A cache would also be quite useful.
comment:26 by , 16 years ago
Is it possible in a similar way to create "real" tar balls (not .zip files) instead?
While presumable a cool feature, it'll possibly take up a lot of resources and bandwidth.
If done though, adding a "download directory as tar/zip-ball" feature to the Browser would suffice.