unnest is very slow .

[antonio-yu]

unnest takes very long time compared to the pandas methods. For example

import pandas as pd 
from datar.all import *
d ={'name': [['a','b','c','d','e','f']]*200000}
test = pd.DataFrame(d)
test >> unnest(f.name)

# it takes 1m 37.6s 

# pandas method
pd.DataFrame(np.concatenate(test.name.values))

#it takes only 2.2s

[pwwang]

How about unchop?

[antonio-yu]

@pwwang unchop is very fast. Only 2s taken

[pwwang]

You should use unchop in your case. See https://tidyr.tidyverse.org/reference/chop.html#ref-examples

In your case, you are unwrapping python lists not tibbles/dfs.

unnest has extra steps to parse tibbles.

[antonio-yu]

Thanks ,@pwwang. I was a little bit confused about unchop and unnest ,It seems they both could approach the same result. The unnest returns the same col name which I dont need to rename , but unchop returns a sequential col name such as name$x.

I would try more to feel the difference.

[pwwang]

Close it for now.