Master format free space

[jkonecny12]

We are using magic numbers acquired by testing to reduce the size depending on a used format.

Also add support for Size operators (add,mul...) by float.

Related: rhbz#1224048
Main patch will go to the anaconda later.

[bcl]

Shouldn't be named 'Class' if it is a constant.

[vpodzime]

Yeah and I think MetadataSizeFactor or something like that would be a better name.

[jkonecny12]

OK I will use MetadataSizeFactor.

[jkonecny12]

Yeah sorry my bad.

[jkonecny12]

I have updated the PR. I hope I did it right :) .

[vpodzime]

Looks good to me now.

[jkonecny12]

@vpodzime thank you I'll change it locally.

[jkonecny12]

Pushed

[dwlehman]

@jkonecny12 Can you please do a followup that removes the float stuff from blivet.size.Size and instead uses decimal.Decimal for the factors in blivet.formats.fs?

[jkonecny12]

Hi @dwlehman, yeah I can do that but why we need that? Isn't it better to do the conversion on one place for every use instead on many places in blivet and anaconda?

[jkonecny12]

This PR is pushed already but I'm reopening the PR to complete discussion on possible follow-up.

[vpodzime]

@jkonecny12 Can you please do a followup that removes the float stuff from blivet.size.Size and instead uses decimal.Decimal for the factors in blivet.formats.fs?

What would be the benefit of this? I know Decimal intentionally doesn't implement those mathematical operations with floats because they result in a lost of precision, but is Size really the same case? We convert everything to whole bytes anyway so what's the point in enforcing those "infinite precision" rules? Could be I'm missing something here, so let me know. It would be a great help for me for thinking about the GLib implementation of Size.

[dwlehman]

It is about loss of precision from the use of float. We stopped using float for anything related to size some time ago. I forgot that we do round to whole bytes, which means this may be more about convention/principle than practical need.

[vpodzime]

I forgot that we do round to whole bytes, which means this may be more about convention/principle than practical need.

Good to know, I wasn't sure I wasn't missing something here. Let's discuss the approach to this at the meeting.

[jkonecny12]

Output from the meeting is this follow-up #229 .
Closing this issue again.