search

tahoe-lafs / trac2gitea

Utility for migrating Trac projects into Gitea.
GNU General Public License v3.0
0 stars 0 forks source link

Add support for keywords migration as labels #28

Open btlogy opened 1 month ago

btlogy commented 1 month ago

Trac tickets have a keywords field which are "useful for searching and report generation" about tickets related to a the same topic without having to create a component or a milestone (which may requires some privileges).

As today, there are already a few fields from Track tickets which are converted in Gitea labels:

From here:

https://github.com/tahoe-lafs/trac2gitea/blob/f013209fe7461359865579a28a161bb74c3abf37/importer/label.go#L42-L70

It would make sense to use a new type of label to migrate the keywords. Though, because of their "free" natures, their number might be quite large and generate a long list of label in Gitea, we might want to enable this feature only when requested and also propose a (default?) prefix such as keyword/.

btlogy commented 2 weeks ago

The first idea is obviously to create a label per keyword... But the keywords being a free string in Trac, this will result in creating 1846 new labels in the case I'm working on! We may consider an alternative implementation which would just add the keywords string at the end of the description! But this approach could make it harder to query the issues (not sure how the Gitea/Forgejo search feature will work it out).

btlogy commented 2 weeks ago

As suggested my @meejah in the last N&B, Twisted has likely faced some similar concern when they have decided to migrated from Trac to GitHub. In their case, they have chosen to add those keywords as part of some searchable metadata in the description (e.g. https://github.com/twisted/twisted/issues/11338). But I'm not sure how those data are searched...

btlogy commented 2 weeks ago

I've tried to query those Twisted searchable metadata, but yet failed to find any issues where they would have used those keywords...

In addition, as far as I can tell, search queries are not as advanced on Forgejo as on GitHub and seem to be limited to the title of the issues (not covering the description or the comment)!

meejah commented 2 weeks ago

This is the repo you want, I believe: https://github.com/chevah/trac-to-github

meejah commented 2 weeks ago

In addition, as far as I can tell, search queries are not as advanced on Forgejo as on GitHub and seem to be limited to the title of the issues (not covering the description or the comment)!

Please add this as another "con" for Forgejo / codeberg.

btlogy commented 2 weeks ago

In addition, as far as I can tell, search queries are not as advanced on Forgejo as on GitHub and seem to be limited to the title of the issues (not covering the description or the comment)!

Please add this as another "con" for Forgejo / codeberg.

If we were to compare with GitHub, advanced issue search queries (e.g.: with regex) and code search would indeed be a con for Gitea/Forgejo and any SaaS serving them indeed. But the main idea was to compare Gitea/Forgejo on self-hosted vs on SaaS.

Never the less, this is an interesting user story to cover assuming Trac was allowing to search in comments. And I was surprised by the fact that Forgejo/Gitea search did not seem to cover it.

As it turned out, this feature is available in the latest Forgejo running now on Codeberg. And not yet in our self-hosted instance in NixOS v24.05. But should be available this month in v24.11.

btlogy commented 2 weeks ago

Assuming we can search in comments (see above), we then could decide to dump the keywords as metadata in the description.

btlogy commented 2 weeks ago

This is the repo you want, I believe: https://github.com/chevah/trac-to-github

Thank you @meejah: this tool is actually parsing the keywords as labels:

https://github.com/chevah/trac-to-github/blob/f7f9c35c9c778a54c8dcf4357e74e6e210a24280/ticket_migrate.py#L778-L812

btlogy commented 1 week ago

In addition, as far as I can tell, search queries are not as advanced on Forgejo as on GitHub and seem to be limited to the title of the issues (not covering the description or the comment)!

Please add this as another "con" for Forgejo / codeberg.

I've verified this apparent lack of "search in comments" and found out that is was caused by a broken indexer (bleeve by default) on the Forgejo instance we are using (it was working as expected on Gitea). I've fixed it, so we could also search for metadata if needed, w/o waiting for a more recent version of Forgejo.

This being said, the migration from keywords to labels is working well enough IMHO. So we may not need to dump metadata to search for those keywords (maybe still needed for something else, but nothing obvious to me yet).