hawkey is not published on PyPI and the Fedora package has no .egg-info

[AdamWill]

The former problem makes it difficult to test anything that uses hawkey with tox. The latter means you cannot list hawkey as a requirement in Python projects, because setuptools, pip etc. etc. can't tell that it's installed.

[ignatenkobrain]

1. most probably you should not use hawkey directly (but use DNF)
2. PyPI is for python packages while libdnf is moving a lot and depends on packages which are not on pypi (so it's just useless to have it there)

[AdamWill]

All I wanted was a NEVRA parser, and hawkey is about the only thing that had one.

[kparal]

We also need a NEVRA parser in taskotron (as a python library) and thus we'd also appreciate having hawkey in pypi. Currently we're using hawkey.split_nevra(). Igor, what is the best available replacement if we want to get rid of hawkey dependency? Thanks.

[AdamWill]

Yes, that's the same thing I'm using, that I couldn't find an alternative for. I was expecting there to be some standard Python bindings for typical 'working with RPMs' functions but couldn't find any anywhere.

[ignatenkobrain]

@kparal ideally, you should not do this at all ;) second best option is to expose this function from RPM API... third best option is to implement it yourself

[kparal]

@ignatenkobrain Can you clarify? What do you mean by "you should not do this at all"? Of course I need NEVRA parsing, when I get a file name and need to figure out the details. RPM API doesn't seem to have this, of course it would be nice if it had. That's why we started using hawkey in the first place. I won't even comment on the idea of every program implementing its our NEVRA parser with its own set of bugs...

[ignatenkobrain]

@kparal for example, filename doesn't contain epoch.. Second thing is that filename doesn't tell anything about content.. so you should use RPM to read its headers which means you don't need to split NEVRA because RPM already did it for you..

I don't know your full use-case, so I'm just guessing.

RPM has that function internally, but it is not exposed, so you could work on properly exposing it for instance.

[AdamWill]

We are often working with just file names, not actual files. (Or even just NVRA strings provided by, e.g., Bodhi.)

[kparal]

Correct, most of the time, we receive a string and process it, and based on that, we do stuff. Very often we don't have the actual RPM downloaded (and doing that would be a waste of resources). I'm aware of the epoch, but that's not important for us too often (and when it is, we query koji to learn it).