Bad rendering of VersionInfo in Pandas DataFrame

[scls19fr]

Hello,

I noticed that VersionInfo are rendered badly in Pandas DataFrame.
(see scls19fr/arduino_libraries_search#8 )

They are rendered as tuple.

In [1]: import pandas as pd
In [2]: import semver
In [3]: versions = ['4.0.0', '5.0.0', '5.0.1', '10.0.0']
In [4]: df = pd.DataFrame(versions, columns=['versions'])
In [5]: df  # version as string
In [6]: df.sort_values(by='versions')  # so they are not ordered correctly
Out[6]:
  versions
3   10.0.0
0    4.0.0
1    5.0.0
2    5.0.1
In [7]: df['versions'] = df['versions'].map(semver.parse_version_info)  # parse versions as `VersionInfo`
In [8]: df.sort_values(by='versions')  # VersionInfo are ordered correctly but are badly rendered
Out[8]:
                 versions
0   (4, 0, 0, None, None)
1   (5, 0, 0, None, None)
2   (5, 0, 1, None, None)
3  (10, 0, 0, None, None)

I don't know what method Pandas use for rendering them.
What I see is that it's neither using __str__ neither using __repr__ method

In [9]: v = df['versions'].iloc[-1]

In [10]: v
Out[10]: VersionInfo(major=10, minor=0, patch=0, prerelease=None, build=None)

In [11]: v.__str__()
Out[11]: '10.0.0'

In [12]: v.__repr__()
Out[12]: 'VersionInfo(major=10, minor=0, patch=0, prerelease=None, build=None)'

VersionInfo inherits from namedtuple so this is probably related.

Any idea?

[scls19fr]

Issue opened upstream at pandas-dev/pandas#21208