Spec pandas[excel] gets old version of Pandas

[notatallshaw]

Since 2.0 Pandas have added several optional extras: https://pandas.pydata.org/docs/getting_started/install.html#optional-dependencies

One common one is pandas[excel], but when I try and run rip against this spec I get Pandas 1.5.3 which is before this optional extra was added, no explanation is given:

$ cargo r -- pandas[excel]
    Finished dev [unoptimized + debuginfo] target(s) in 0.07s
     Running `target/debug/rip 'pandas[excel]'`
Resolved environment:
- pandas[excel]

Name             Version
numpy            1.26.0
pandas[excel]    1.5.3
python-dateutil  2.8.2
pytz             2023.3.post1
six              1.16.0

Running pandas without the optional extra gets the expected Pandas version:

$ cargo r -- pandas
    Finished dev [unoptimized + debuginfo] target(s) in 0.07s
     Running `target/debug/rip pandas`
Resolved environment:
- pandas

Name             Version
numpy            1.26.0
pandas           2.1.1
python-dateutil  2.8.2
pytz             2023.3.post1
six              1.16.0
tzdata           2023.3

[baszalmstra]

Thats not good! Ill take a look!

[notatallshaw]

Thanks, also I think the output message shoud highlight when it failed to get an extra:

The report gives:

pandas[excel]    1.5.3

But this extra did not exist for 1.5.3, I beleive the spec on this is extras are considered optional, so it's technically correct to be able to install pandas without the extra excel even when you specify it (which personally I wish all installers provided an option to override thie behavior and fail to install when the extra isn't present), but at the very least the user should be alerted.

[baszalmstra]

After a little investigation, it turns out that this is the same result pip returns:

> pip install pandas[excel] --only-binary :all: --dry-run
...
Would install numpy-1.26.0 pandas-1.5.3 python-dateutil-2.8.2 pytz-2023.3.post1 six-1.16.0

We currently don't deal with sdists at all yet (that's why I use --only-binary :all:).

When we do consider sdists with pip we get the result you'd expect:

> pip install pandas[excel] --dry-run
...
Would install XlsxWriter-3.1.5 defusedxml-0.7.1 et-xmlfile-1.1.0 numpy-1.26.0 odfpy-1.4.1 openpyxl-3.1.2 pandas-2.1.1 python-dateutil-2.8.2 pytz-2023.3.post1 pyxlsb-1.0.10 six-1.16.0 tzdata-2023.3 xlrd-2.0.1

I see your point with regard to the report being incorrect. Pip indeed prints a warning:

WARNING: pandas 1.5.3 does not provide the extra 'excel'

I will add a warning to the resolution.

[notatallshaw]

Thanks for the update, I will compare to Pip with --only-binary :all: from now on.

[wolfv]

If you have hints on how we can make sdists work we would also gladly take them (#32)

[baszalmstra]

Im going to close this issue in favor of #32