ENH: dataframe.explode() function variant

[jablka]

Feature Type

• Adding new functionality to pandas

• Changing existing functionality in pandas

• Removing existing functionality in pandas

Problem Description

There is a Dataframe.explode function that explodes list-like data to rows... that is the long format.

I would like to have a similar function, that would expand (or explode) from list-like data to columns (wide format) instead.

Feature Description

new function, that would expand (or explode, or unpack) from list-like data to columns (wide format).

Alternative Solutions

the problem can be also formulated as "splitting column of list-like data".
There are solutions, but not straight-forward.
Split a Pandas column of lists into multiple columns (Stack Overflow)

Additional Context

Q:
in pandas dataframe,
in one column I have list of strings ... like ['str1','str2'].
How to expand (or unpack) those strings to separate columns ?
https://chat.openai.com/c/15a91b3a-c01c-468f-bb1f-19d20c0e1418

[BenjaminHelyer]

I'd be interested in seeing this as well. Just to clarify, does this example encapsulate what you're suggesting?

A
------
[1, 2]
[3, 4]
[5, 6]

is expanded to

A | B
--------
1 | 2
3 | 4
5 | 6

Following the behavior of explode(), this would seem to entail that all elements of list A must either (1) be of the same length or (2) be scalar, in which case they are repeated across columns.

Any further thoughts on this @jablka or anyone else?

[jablka]

yes, basically that's it...
but I leave the technical details up to others.

[rmhowe425]

take

[jbrockmendel]

Not exactly what you're looking for, but two workarounds come to mind:

vals = [[1, 2], [3, 4], [5, 6]]
ser = pd.Series(vals)

# option 1: use a pyarrow dtype
ser2 = ser.astype(pd.ArrowDtype(pa.list_(pa.int64())))
df = pd.concat([ser2.list[0], ser2.list[1]], axis=1)

# option 2: use the regular explode, then unstack a multiindex
flat = ser.explode().astype("int64")
flat_df = flat.to_frame()
flat_df[1] = list(range(len(vals[0]))) * len(vals)
ser3 = flat_df.set_index(1, append=True)[0]
df2 = ser3.unstack(1)
df2.columns.name=None