feat: support pd.Series.explode with ExtensionArray._explode method

[douglasdavis]

Fix for #38

Now that pandas 2.2.0 is out we can implement this

[martindurant]

When done on a series, what's the return type of explode()? Just curious.

[douglasdavis]

Just a Series again. The two returns here are consumed by Series.explode, it uses the second return to build a new Index object and then passes the array and the index objects to Series._constructor:

https://github.com/pandas-dev/pandas/pull/54834/files#diff-a5257444a1b322d619680fc77361cc6ea11ef36b363b4bb2289fdef0f41feb70R4403

[CrfzdPQM6]

@martindurant suggested I paste this snippet here as a useful test:

import awkward as ak
values = [ak.Array([1.0,2.0,3.0]), ak.Array([4.0,5.0]), ak.Array([6.0])]
df = pd.DataFrame({'pt':values})

because the output of df.dtypes is that pt is object

I'm running into issues with pandas 2.2.1 that df.explode() does not work for a column with dtype awkward

[martindurant]

df = pd.DataFrame({'pt':values})

needs to have dtype specified (or astype called) to make it awkward, else you are simply calling pandas' default iteration, which would also work for python lists.