Read_Ncol crashes when being passed a pandas dataframe

[vtraag]

Describe the bug
When a pandas dataframe is passed to Graph.Read_Ncol it crashes. Of course, a dataframe should not be passed to Graph.Read_Ncol, but this should not cause the Python session to crash, but should be correctly handled.

To reproduce

import igraph as ig
import pandas as pd
df = pd.DataFrame({'from': [1, 2],
                   'to': [2, 3]})
g = ig.Graph.Read_Ncol(df)

Version information
This is using version 0.9.6 installed from conda.

[ntamas]

Can you please try it with the latest dev version? I cannot reproduce it with the dev version on my machine; nevertheless I have added a regression test for it so we can see it in the CI if this fails on some platform.

On my machine, I get an exception like this:

TypeError: argument must be an int, or have a fileno() method

---

Edit: don't bother, there are CI failures so it looks like it's a genuine problem, at least with older versions of Python. I'll investigate a bit more.

[ntamas]

Strangely enough, I have apparently found a completely unrelated bug in igraphmodule_PyObject_to_edge_type_sw_t(), which seems to have fixed the problem on Python 3.7 on my machine. Let's see how the CI tests go now.

Related commits: be01f01, 78adc9e

[ntamas]

@vtraag Can you test now with the latest master to see if the problem is still there on your machine? (CI seems to pass now).

[vtraag]

I can confirm that using the laster master this no longer crashes, I now also receive the error

TypeError: argument must be an int, or have a fileno() method.

Good to see it fixed, even thought it's not clear how it the original problem relates to the fix.