seq() converts pandas DataFrame into Sequence

[Arshaku]

from functional import seq
from pandas import DataFrame

df = DataFrame({'col1': [1,2,3], 'col2': [4,5,6]})
s = seq([df])
el = s.first()
print(type(el))

this code prints: "<class 'functional.pipeline.Sequence'>"
but the expected output is "<class 'pandas.core.frame.DataFrame'>"

[EntilZha]

This is related to #158, where the root issue is that for convenience I originally decided to wrap a little to aggressively. I think the fix would be in two steps: (1) add a configurable option to not wrap elements with default to wrap (2) bump version to 2.X and make default to not wrap to avoid breakage. I'd be open to a PR that does this.

[reklanirs]

Similar issue duing reduce with the lastest master.

Expected:

from functools import reduce

reduce(lambda x,y: x.add(y), [df,df])

Out[1]:
  A     B
0  24.0  14.0
1   8.0   4.0

In fact:

seq([df,df]).reduce(lambda x,y: x.add(y))

Out[2]:
[array([24., 14.]), array([8., 4.])]

to_pandas can give some help but the column names will be missing:

seq([df,df]).reduce(lambda x,y: x.add(y)).to_pandas()

Out[3]:
      0     1
0  24.0  14.0
1   8.0   4.0

Hope it can be fixed

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[Arshaku]

Hi, any chance this issue will be fixed soon?

[EntilZha]

I don't have the bandwidth to contribute fixes myself right now, I'd welcome/review pull requests that fix it roughly how previously outlined.

[swiergot]

@EntilZha Any reason why __getitem__() also wraps?

[EntilZha]

The reason I originally did it this way is the same, I wanted it to be easy to do something like:

In [1]: from functional import seq

In [2]: seq.range(10).grouped(3)[0].map(lambda x: x * 2)
Out[2]: [0, 2, 4]

In [3]: type(seq.range(10).grouped(3)[0].map(lambda x: x * 2))
Out[3]: functional.pipeline.Sequence

As I mentioned in my prior comments, in retrospect, this has three issues: (1) there is no way to configure the behavior, namely disable it, (2) even if it were configurable, I think its probably incorrect to make the default in most cases to wrap, it probably should do that more sparingly, and (3) changing this is a breaking change, likely requiring a move to 2.x.

I'd welcome/review PRs that would fix this, but don't have the time to do it myself right now. If you are interested, I can outline how I'd do this in a little more detail.

Thanks!

[swiergot]

@reklanirs

In fact:

seq([df,df]).reduce(lambda x,y: x.add(y))

Out[2]:
[array([24., 14.]), array([8., 4.])]

This is a different problem caused by the fact that PyFunctional has a special handling for DataFrame - for some reason it extracts values from it.

to_pandas can give some help but the column names will be missing:

seq([df,df]).reduce(lambda x,y: x.add(y)).to_pandas()

Out[3]:
      0     1
0  24.0  14.0
1   8.0   4.0

How about this:

>>> seq(reduce(lambda x,y: x.add(y), [df,df])).to_pandas(df.columns)
   col1  col2
0     2     8
1     4    10
2     6    12