Pulp

• What should the name of the field be? href_in_list or id_in_list?

The name of the field should be "href and it should support the "exact" and "in" filters. This would allow you to filter on a list of hrefs with the query "/?href__in=<href1>,<href2>

My filter overhaul recently did something similar with all of the existing filters, e.g. "/?name__in=foo,foo2". It also works for hrefs, though, and I implemented "content__in" to search for repo versions containing multiple content hrefs.

So I thought about this a bit more and I think we should try to nail down some actual use cases for this functionality.

The main purpose of filtering is a human-friendly way of figuring out which objects meet certain human-facing criteria. In order to obtain your list of HREFs you're probably doing some other filter operation first, using dates or names or numbers or online=True or something. But in that case, you could just combine your filters together instead of trying to extract the HREFs and then do a separate filter operation on those HREFs.

e.g. instead of doing

/repository/<uuid>/versions/?number__gte=5
{
   blah blah blah
}

HREFS go into new filter

repository/<uuid>/versions/?content=<content_href>&href__in=<blah1_href>,<blah2_href>,<blah3_href>
{
    blah blah blah
}

Why not just encourage

/repository/<uuid>/versions/?number__gte=5&content=<content_href>

An HREF is a glorified primary key, it seems silly to have the user be chucking those around unless they have nothing else to use, such as with ContentUnits (since the user-friendly names are attached to Artifacts, of which there can be many). That's the only model I can actually think of where it makes sense.

I think the main use case is Katello. They store ids for objects and often times will want to look up lists of objects by their ids/hrefs. Since katello is a multi-tennant app, the names Katello is using for resources in Pulp are auto-generated so Katello uses ID to identify objects and often they are dealing with 100s or even 1000s of objects they want to get back. Filtering by ids/hrefs lets them submit a single request.

I'm not too concerned about getting this done anytime soon which is why it's post-MVP but we definitely will need this kind of filter I think. We can confirm with Katello about which specific objects they need to filter by ids/hrefs on.