BUG/API: interpolate with limit=0 should mean no interpolation

[shoyer]

Current behavior:

In [28]: s = pd.Series([1, 2, np.nan, np.nan, 5])

In [29]: s
Out[29]:
0     1
1     2
2   NaN
3   NaN
4     5
dtype: float64

In [30]: s.interpolate(limit=0)
Out[30]:
0    1
1    2
2    3
3    4
4    5
dtype: float64

But in fact, limit=0 in contrast to limit=None should mean no interpolation.

[jreback]

very confusing to have None and 0 mean different things

why would you do this?

[shoyer]

I'm testing the results for different interpolation limits and thought a simple way to trigger no interpolation would be use limit=0, e.g., df = pd.DataFrame({lim: s.interpolate(limit=lim) for lim in range(5)}).

I can see why limit=None makes sense as a sentinel value. But I interpreted limit=0 as "interpolate at most zero values".

[shoyer]

Another sane API design would be to disallow limit=0 entirely, e.g., assert limit > 0.

[jreback]

ok I understand the reason and it makes sense

though maybe a bit confusing (I get for testing this is useful) but from a user perspective u r like why have an option that makes the function do nothing?

[jreback]

yep maybe limit > 0 makes more sense (but raise a ValueError)