BUG: df.setitem can be 10x slower than pd.concat(..., axis=1) (original) (raw)

I have checked that this issue has not already been reported.
- I think so, it's a bit hard to search for though. Nothing under concat and setindex
I have confirmed this bug exists on the latest version of pandas.
(optional) I have confirmed this bug exists on the master branch of pandas.

Note: Please read this guide detailing how to provide the necessary information for us to reproduce your bug.

Code Sample, a copy-pastable example

import pandas as pd, numpy as np from string import ascii_lowercase

def setitem(x, x_cols, df): new = pd.DataFrame(index=df.index) new[x_cols] = x new[df.columns] = df return new

def concat(x, x_cols, df): return pd.concat( [ pd.DataFrame(x, columns=x_cols, index=df.index), df, ], axis=1, )

x = np.ones((1000, 10)) x_col = list(ascii_lowercase[:10]) df = pd.DataFrame( { "str": np.random.choice(np.array(list(ascii_lowercase)), size=1000), "int": np.arange(1000, dtype=int), } )

%timeit setitem(x, x_col, df)

3.78 ms ± 193 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)

%timeit concat(x, x_col, df)

306 µs ± 9.29 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)

Problem description

This seems unintuitive from a performance perspective. I would assume that these would be close to equivalent. The setitem implementation might even be expected to have better performance due to less allocation.

While this is a stripped down example, the use case is building a dataframe to return from a function. Making a dataframe, then adding columns seemed like the natural idiom here.

Output of `pd.show_versions()`

Details