Fix regression in naturalsize for float

[hugovk]

Fixes #239.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 99.49%. Comparing base (a0602c7) to head (11e62ee).
Report is 19 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #240   +/-   ##
=======================================
  Coverage   99.49%   99.49%           
=======================================
  Files          11       11           
  Lines         796      796           
=======================================
  Hits          792      792           
  Misses          4        4           

Flag	Coverage Δ
macos-latest	96.83% <100.00%> (ø)
ubuntu-latest	96.83% <100.00%> (ø)
windows-latest	94.36% <100.00%> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[davidboweninrupt]

Hi @hugovk

That's actually wonderful.

I can accept either behaviour, but I wanted it to be stable. Your additional tests really help with that. Ironically the change in behaviour popped up because my tests spotted it.

My local fix was also to convert the bytes to an integer.

I do have cases where I'm looking at the average-bytes-per-second so it might not be an integer. However, naturalsize does say it's trying to be like a file size so integers do make sense.

The proposed fix does an int rather than rounding it and as we know:

int(2.7)
2

which is tolerable, but not what I'd expect from humanize. That's mostly because I've been so impressed by how humanize has given me great outputs for so long. It looks like you are chatting about rounding down in naturaldelta too too.

So, in summary, I'd be fine with it going in this way.

I really appreciate you getting to this so quickly.

[hugovk]

Yeah, and it matches the behaviour of the earlier percent formatting:

>>> "%d" % 2.7
'2'
>>> int(2.7)
2

I'm open to changing the rounding, but let's do it in a more intentional way than a bad formatting upgrade :)

[davidboweninrupt]

Nice, thanks @hugovk

[hugovk]

You're welcome! 🚀