PERF: 34% faster Series.to_dict

[staadecker]

Improve the performance of Series.to_dict() by not including a generator that is not necessary. Local tests showed significant performance improvements.

Note this builds off the changes of #46487 from @RogerThomas

[rhshadrach]

Nice! Can you post the results from your local test?

[staadecker]

Here is the test. Performance improves by 34%

import pandas as pd
import random
from timeit import timeit
df = pd.Series([random.random() for _ in range(1000)], dtype=float)
print(timeit("df.to_dict()", globals=globals(), number=10000))

Before change: 2.49s
After change: 1.64s

[staadecker]

Note the performance gain only applies when the series uses Python native types. Also out of curiosity, what's the release timeline for Pandas 2.0?

[RogerThomas]

LGTM, sorry, should have noticed that the generator was no longer necessary in my PR

[rhshadrach]

lgtm. No need for an entry in the whatsnew, the whatsnew from the PR linked in the OP here covers this as well.

[rhshadrach]

Thanks @staadecker - very nice!

[staadecker]

Thank you all!