DOC/TST: pandas.Series.str.get with dict elements supports passing hashable label

[BeRT2me]

Pandas version checks

• I have checked that the issue still exists on the latest versions of the docs on main here

Location of the documentation

https://pandas.pydata.org/docs/dev/reference/api/pandas.Series.str.get.html

Documentation problem

The fact that Series.str.get also works with Non-Numerical dict types isn't clear in the documentation.

Suggested fix for documentation

Parameters should be int or str and an example using a dictionary should be added.

[mroeschke]

I find it odd that arguments other than int are allowed in get since the value should indicate the position. I think instead, a ValueError should be raised if the value is not an int.

[phofl]

This does not work anyway, this returns an all nan series

s = pd.Series(["String",
              (1, 2, 3),
              ["a", "b", "c"],
              123,
              -456,
              {1: "Hello", "2": "World"}])

print(s.str.get({1: 1}))

[xr-chen]

I find it odd that arguments other than int are allowed in get since the value should indicate the position. I think instead, a ValueError should be raised if the value is not an int.

Would TypeError be more suitable than ValueError?

[mroeschke]

Sure, TypeError makes sense here

[BeRT2me]

This does not work anyway, this returns an all nan series

s = pd.Series(["String",
              (1, 2, 3),
              ["a", "b", "c"],
              123,
              -456,
              {1: "Hello", "2": "World"}])

print(s.str.get({1: 1}))

I'm not sure where you managed to get that passing a dictionary is what I was referring to...

This is what works just fine:

s = pd.Series([{"name": "Hello", "value": "World"},
              {"name": "Goodbye", "value": "Planet"}])

print(s.str.get('name'))

Raising an error would destroy functionality...

[phofl]

You should have made this significantly clearer in your initial post.

Anyway, I don’t think this should be supported

[BeRT2me]

There's literally an example of it being used with a dictionary already in the existing documentation.
And I was clear, you just didn't read the whole post.

Parameters should be int or str...

What is one good reason for NOT supporting this functionality? Would you prefer apply and lambda have to be used every time you want to access dict values in a series?? Intuitively, having .get work this way makes perfect sense, it's mirroring the existing dict.get functionality.

[xr-chen]

s = pd.Series([{"name": "Hello", "value": "World"},
              {"name": "Goodbye", "value": "Planet"}])

print(s.str.get('name'))

Raising an error would destroy functionality...

I just tried s.str['name'] works well even if we raise an error when non-integer type passed in Series.str.get.

[BeRT2me]

What is the point of Series.str.get then? s.str[0] works just as well.
If anything, I think it should better mirror dict.get and allow a default value to be specified.

[BeRT2me]

This is the current implementation, it's obviously not meant to be int only...

    def _str_get(self, i):
        def f(x):
            if isinstance(x, dict):
                return x.get(i)
            elif len(x) > i >= -len(x):
                return x[i]
            return self._str_na_value

        return self._str_map(f)

---

If anything, it should be:

    def _str_get(self, i, default=None):
        def f(x):
            if isinstance(x, dict):
                return x.get(i, default)
            elif len(x) > i >= -len(x):
                return x[i]
            return default if default else self._str_na_value

        return self._str_map(f)

[mroeschke]

Thanks @BeRT2me. Although I find this "odd usage" being under str.get, I suppose this should be supported and have better documentation.

[BeRT2me]

I agree that it's odd usage, but how Series.str works for anything (list, tuple, dict, set, etc.) that isn't a string hasn't made much sense logically for a while.