wojciech.kapcia@tigase.net opened 4 years ago
|
|||||||
Chat transcript:
|
|||||||
I'm wondering if we shouldn't reorganise the schema files at the same time as well, possibly group them by database/component (i.e. mysql/muc/xxx.sql, mysql/offline/xxx.sql). One of the pain points previously was a mess we had in schema files so with multiple "atomic" schema changes this could became a problem again. What do you think @andrzej.wojcik ? |
|||||||
We could do that, but copying from database to database/component/ should be done by distribution packaging. |
|||||||
Agreed but we would still should divide per database in component repositories (it would be simpler to configure distribution packaging). |
|||||||
I know it would be simpler but somehow I do not like to be forced to create "component" directory in the database directory in the repository. At this point (as we are switching from component version based versioning to schema based versioning) maybe we should have database/repo instead of database/component and that would be more acceptable for me and it would keep each repo schema separated. |
|||||||
I'm starting to wonder if we should not keep schema for each database separated |
|||||||
OK, I think we miscomunicated here. What I was proposing to:
For example
a) in tigase-server we would have |
|||||||
well, ok. I'm just not sure how would it look while merged for distribution but the idea looks ok. |
|||||||
In distribution it would be:
|
|||||||
wojciech.kapcia@tigase.net batch edited 7 months ago
|
Type |
New Feature
|
Priority |
Normal
|
Assignee | |
Version |
Candidate for next major release
|
Currently we version schema based on associated component version, which doesn't seem ideal - it requires adding empty files without any changes for example.
A better approach would be to version schema based on required changes in the database (more atomic changes) - it would simplify a lot of things:
upgrade-schema
to collect the files anyway)