Non-normalised package name

[jayvdb]

I would like to use the data to correlate with openSUSE package names, which use the 'real' name supplied in setup.py, i.e. not-normalised.

I've been doing a bit of research at hugovk/top-pypi-packages#4 and psincraian/pepy#128, and the raw data from bigquery can include this, with a very small perf hit.

The query only needs to change from selecting file.project to substr(max(file.filename),1,LENGTH(file.project)) , or more likely including both.

Note this does depend on using standard SQL ( #28 ).

Do we know the cost implications of those changes?

[ofek]

Hello there!

I'm unsure of the cost implications of this. Though, I'll approve whatever @hugovk thinks is best 🙂

[ofek]

@hugovk

[hugovk]

I don't know the cost implications, I guess the best way is to test it out.

If it costs more, I'd suggest adding a switch to include the change. I'm nearly always out of quota.

[jaraco]

I'd also be interested in this feature. In coherent-oss/coherent.build#3, I'm building a database that keeps track of the normalized and non-normalized names (along with the Python root names it exposes). It currently resolves the package name from the wheel that it downloads, but there are situations where it would be advantageous to have the original name available prior to resolving the wheel (or when a wheel is unavailable).