perf(python): Construct Series for bytes/binary data 10x faster when dtype not explicitly set

[alexander-beedie]

Can still detect/use the fast-path binary constructor, rather than the generic object constructor, even when the dtype was not explicitly set as pl.Binary.

Not a bad return on a one-line update ;)

Example

from codetiming import Timer
import polars as pl

binary_data = [str(n).encode("utf8") for n in range(50_000_000)]

with Timer():
    s = pl.Series(binary_data)

Timings

Before: 5.6219 secs  (new_object)
 After: 0.5609 secs  (new_binary) 🚀

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 79.58%. Comparing base (3fb10e8) to head (5cc49b9).
Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #20157      +/-   ##
==========================================
- Coverage   79.58%   79.58%   -0.01%     
==========================================
  Files        1563     1563              
  Lines      217417   217417              
  Branches     2472     2472              
==========================================
- Hits       173023   173021       -2     
- Misses      43826    43828       +2     
  Partials      568      568              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.