Allow index in replace(to_replace)
amueller opened this issue · comments
Andreas Mueller commented
Hey. I ran series.replace(some_things, "other")
which works with an index in pandas, but not in Koalas.
Is there a plan to support it?
In particular, this was n Int64Index that came out of a value_count operation.
My pandas code was
series = pd.Series(np.random.randint(0, 10, 40))
counts = series.value_counts()
series.replace(counts.index[counts < 4], -1)
You can't subscript an index, but I expected
series = ks.Series(np.random.randint(0, 10, 40))
counts = series.value_counts()
series.replace(counts[counts < 4].index, -1)
but it gives "to_replace should be one of str, list, dict, int, float".
Koalas version is 1.4.0.
I wanted to use .tolist()
as a workaround but that is not available in 1.4.0, in which version was it added?
Potentially related to #1516.
Hyukjin Kwon commented
Looks like it's added from 1.5.0 (https://github.com/databricks/koalas/releases/tag/v1.5.0). But yeah we should ideally make it working, I agree.