Nan value in dataframe editor crash Spyder

[markusr]

Description of your problem

1. Run the following script:

import numpy as np
import pandas as pd

np_array = np.zeros(100)
np_array[10:] = np.nan
dataframe = pd.DataFrame(np_array)

2. Go to the Variable explorer
3. Double click on the dataframe variable
4. Close the table editor window

What is the expected output? What do you see instead?
Spyder completely crashes. This is not always reproducible but opening and closing table editor several times triggers the crash.

Versions and main components

• Spyder Version: 3.1.2
  Python 3.5.2 64bits
  Qt 5.7.0
  PyQt5 5.7
• Operating system: Kubuntu 16.04

Dependencies

jedi >=0.8.1 : 0.10.0 (OK)
matplotlib >=1.0 : 1.5.3 (OK)
nbconvert >=4.0 : 5.1.1 (OK)
numpy >=1.7 : 1.12.0 (OK)
pandas >=0.13.1 : 0.18.1 (OK)
pep8 >=0.6 : 1.7.0 (OK)
psutil >=0.3 : 5.1.3 (OK)
pyflakes >=0.6.0 : 1.5.0 (OK)
pygments >=2.0 : 2.2.0 (OK)
pylint >=0.25 : 1.6.5 (OK)
qtconsole >=4.2.0: 4.2.1 (OK)
rope >=0.9.4 : 0.9.4-1 (OK)
sphinx >=0.6.6 : 1.5.2 (OK)
sympy >=0.7.3 : None (NOK)

[ccordoba12]

I can't reproduce this. How did you install Qt/PyQt?

[markusr]

I don't know how I installed it. All with standard packages and/or pip.

I did some further investigations:

• The format (in variable explorer) has to be %d. Which I used before for a special dataframe.
• There are exceptions in the internal console.

Traceback (most recent call last):
  File "/usr/local/lib/python3.5/dist-packages/spyder/widgets/variableexplorer/dataframeeditor.py", line 277, in data
    return to_qvariant(self._format % value)
ValueError: cannot convert float NaN to integer

[ccordoba12]

@jitseniesen, any ideas?

[jitseniesen]

Yes, I understand it after markusr's explanation. The error message complains that the format %d is for integers and cannot be applied to NaN.

An easy solution for us would be to use the default format (which is %.3g) in this case. Would that be acceptable from a user experience point of view?

[markusr]

An easy solution for us would be to use the default format (which is %.3g) in this case. Would that be acceptable from a user experience point of view?

@jitseniesen Are you suggesting that non-NaN values are shown as %d and NaN values as %.3g (if %d is set as the input format)?

[jitseniesen]

Yes

[markusr]

This sounds like a good solution. I also gave it some thoughts but I couldn't come up with better solution.

Anyhow better than the application crash...

[ccordoba12]

@jitseniesen, could you create a PR against the 3.1.x branch for this? It seems an easy fix but I fail to understand what we should exactly do to fix it :-)