display.precision not honored for column headers

[pblelloch]

Code Sample, a copy-pastable example if possible

import pandas as pd
import numpy as np
pd.options.display.precision = 3
labels = np.random.rand(10)
p = pd.DataFrame(np.random.randn(10,10),index=labels,columns=labels)
p

Problem description

The above code works just fine in a standard console, but in a jupyter notebook the float valued column headers are not displayed with a precision of 3. This appears to be due to the notebook_repr_html option, since setting that to False again honors the precision. Note that the same applies to other formatting options, and if I switch to a MultiIndex it looks fine. I posted this as a question on the pydata discussion site and Joris Van den Bossche confirmed that it was an issue and suggested that I report it here.

Expected Output

Output of pd.show_versions()

[paste the output of pd.show_versions() here below this line]

[jorisvandenbossche]

Thanks for reporting!

Note that the same applies to other formatting options

What do you mean exactly with this? That also other formatting options don't work with the notebook repr?

[pblelloch]

Yes. I couldn't remember them off the top of my head, but I think that there's a float_format option and an eng_float_format, or something like that. All behave the same way. Thanks, Paul

…

Sent from my iPhone On Aug 18, 2017, at 7:01 AM, Joris Van den Bossche <notifications@github.com<mailto:notifications@github.com>> wrote: Thanks for reporting! Note that the same applies to other formatting options What do you mean exactly with this? That also other formatting options don't work with the notebook repr? - You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#17280 (comment)>, or mute the thread<https://github.com/notifications/unsubscribe-auth/AEG2KrjA9m1I944PDF6vhXglCny7ADzRks5sZZkQgaJpZM4O7jej>.

[matanox]

I believe this is not fixed in pandas 0.23.0 at least. Unless it should be seen as a Jupyter notebook issue after all. If we print(p) rather than simply p, the issue does not manifest (but in that case the display is more primitive than without the print).