Improper handling of Series in Series constructor

[wesm]

This looks like a regression that did not have a unit test (pretty sure this used to do the right thing)

index1 = ['d', 'b', 'a', 'c']
index2 = sorted(index1)
s1 = Series([4, 7, -5, 3], index=index1)
s2 = Series(s1, index=index2)

cc @hughdbrown

[manuteleco]

Hi,

I had some code that was relaying on the Series constructor the way it was before this bugfix. I was able to adapt it to the new version very easily, but I just wonder if this is the intended behavior now:

In [2]: Series(Series(["a"]), index=[2])
Out[2]: 2    NaN

when before you could do:

In [2]: Series(Series(["a"]), index=[2])
Out[2]: 2    a

Of course, you can still get the same result as before just by using the .values property:

In [3]: Series(Series(["a"]).values, index=[2])
Out[3]: 2    a

so, again, I'm just wondering if this is the intended behavior or maybe an unforeseen side-effect.

Thanks and regards.
Manu.

[wesm]

Yes. I never intended for this behavior:

In [2]: Series(Series(["a"]), index=[2])
Out[2]: 2    a

It's not consistent with how Series handles other data types, like dicts. It's been brought up to me by others, but I think unfortunately a lot of people have written code relying on the "bug" here.