This is the use case I am looking at:

This example failes, becuase some of the text is the words "is" and "try" which can not be columns.

from sklearn.feature_extraction.text import TfidfVectorizer
import vaex
ds = vaex.from_arrays(x=['I try this','but this is hard','I should buy a cat'])
tdfidf = TfidfVectorizer()
s = tdfidf.fit_transform(ds.x)
ds.add_columns(tdfidf.get_feature_names(), s)

Another example which will crash also because columns can't start with a digit:

ds = vaex.from_arrays(x=['I try this','this 1and and 2and','I should buy a cat'])

Yes, those a new bugs, i have fixes for them.

ds = vaex.from_arrays(x=['I try this','this 1and and 2and','I should buy a cat'])

This just works right?