ENH: pd.series.sample when n > len

[marctorsoc]

Feature Type

• Adding new functionality to pandas

• Changing existing functionality in pandas

• Removing existing functionality in pandas

Problem Description

Oftentimes, I have some code that samples from a df / series, as in

df.sample(sample_size, random_state=88)

and given new data, sample_size > len(df), getting

...Cannot take a larger sample than population...

I'd like a way to specify that if sample_size > len(df), I just want all elements back. This is already the case with .head(). So I don't really understand why the behaviour is not the same here.

Feature Description

I can see two possible solutions

1. As in here, "If n is larger than the number of rows, this function returns all rows."

2. If keping back-compatibility is a must, then adding a parameter errors, with possible values ignore or raise (default being raise, again to keep back-compatibility).

I'd lean towards (1), but I'd be content with (2)

Alternative Solutions

I proposed two solutions. Of course, one can always add a line

sample_size = min(len(df), sample_size)

before the call, but I honestly think pandas should provide support for this common case + being consistent with other methods e.g. head.

Additional Context

Happy to contribute with this feature, but first checking here. Just making sure owners would consider it, and nobody is working on this atm.

[lithomas1]

IIRC the error comes from numpy.

This is what I get

  File "/Users/thomasli/pandas/pandas/core/sample.py", line 153, in sample
    return random_state.choice(obj_len, size=size, replace=replace, p=weights).astype(
  File "numpy/random/mtrand.pyx", line 1001, in numpy.random.mtrand.RandomState.choice
ValueError: Cannot take a larger sample than population when 'replace=False

In that case, in my opinion, I think it's probably best for you to add an if check in your code before the call to sample to restrict the sample size to the length of the dataframe.

[marctorsoc]

yeah, for the cases of production code, of course I can add an if. But when exploring on a notebook with constructions like this (imagine this is not static code, but iterating with many filters on it):

(
         df
         .loc[lambda df: df.price.gt(10)]
         .loc[lambda df: df.date.lt("2023-04")
         .sample(10)
)

Here you cannot do an if before because there's no before. I happen to do this very very often. I guess one solution would be ok, create a pipe and do

(
         df
         .loc[lambda df: df.price.gt(10)]
         .loc[lambda df: df.date.lt("2023-04")
         .pipe(sample_pipe, n=10)
)

containing the if... but I continue thinking that pandas should support my use case 🤷‍♂️

[mroeschke]

Yeah especially since this is an error from numpy, I don't think it's appropriate for pandas to work around this so closing