Story #5344
Updated by dalley almost 5 years ago
Problem Statement:
API design for copy and upload has thus far been somewhat haphazard and it would be useful to make it more consistent. Additionally there are some deficiencies in some of the current APIs that need to be addressed.
For example, the copy API currently looks something like this
<pre>
POST /pulp/api/v3/rpm/copy/ types=['package', 'modulemd']
</pre>
The problem with this API is that it will be difficult to make it play nicely with filtering, which has yet to be implemented. The filtering use case is essentially "copy units for which metadata matches these filters, e.g. name=walrus and version>1.1.0". However, because different unit types have different metadata fields, allowing the user to specify filters while also allowing more than one type of content to be copied at once would be problematic because there's no good way to determine what filters should apply to what content types. Possibly it could be done, but it would probably not be a good idea. Therefore it is best that the "types" parameter be removed entirely.
Sidenote: In various discussions we've mentioned the utility (and possibly necessity) of being able to have a group of tasks that are atomic and transactional together. Something like this could be applied here in order to replace this functionality (however, it is out of scope for this Story).
There are 2 different alternative APIs I would like to propose:
1. The same API, but with a single type copied at a time. If "all" is specified as the type, it would copy everything. This would IMO cover the most important use cases.
<pre>
POST /pulp/api/v3/rpm/copy/ type='package'
</pre>
Pros:
- More flexible, more similar to the Pulp 2 API
Cons:
- Dissimilar to the other API's in Pulp 3
- "type" is going to be the name of a field on the RPM content itself (inherited from core, formerly _type)
2. Make the type to be copied part of the name of the endpoint itself, like so:
<pre>
POST /pulp/api/v3/rpm/package/copy/
</pre>
Pros:
- Consistent with other Pulp 3 APIs
Cons:
- Not as flexible, no way to copy all units at once via this API
3. Use a complex filter endpoint
<pre>
POST /pulp/api/v3/rpm/copy/ filter={'package': "name='foo', version>='4.5', arch='x86_64'", 'modulemd': "name=ripgrep, stream=master"}
</pre>
Pros:
- Might be able to support all the use cases we want in one API endpoint
Cons:
- Complicated to implement
h2. Use cases
* As a user, I can copy all content of a specific type from one repository to another repository
* As a user, I can copy content matching certain "search criteria" from one repository to another repository
* As a user, when copying content that directly references other content, the referenced content is ?always? copied
* e.g. Modules referencing Modules, Modules referencing RPMs, Erratum referencing RPMs, and {{other types}}
* As a user, I can optionally choose to copy all indirect dependencies of content that is being copied (recursive copy)
* Should this be the default?
* Some content types create a new content unit in the destination repo instead of just copying, e.g. yum_repo_metadata_file
* Special copy needed for modules? (multi-repo)
* Special copy needed for distribution trees?
.