usage.py failing: minimum polars version required?

[abubelinha]

Coming from #3 , where I tried installation on two different Windows machines. Both of them have now polars and polars-fuzzy-match v. 0.1.3

• On Python 3.9 machine (Windows 10 64 bits), I couldn't run usage.py example. This line raised error:

from polars_fuzzy_match import fuzzy_match_score

c:\python39\python polarex.py
Traceback (most recent call last):
  File "c:\python\polarex.py", line 78, in <module>
    polarsfm_usage(); sys.exit()
  File "c:\python\polarex.py", line 11, in polarsfm_usage
    from polars_fuzzy_match import fuzzy_match_score
  File "c:\python39\lib\site-packages\polars_fuzzy_match\__init__.py", line 4, in <module>
    from polars.type_aliases import IntoExpr
ModuleNotFoundError: No module named 'polars.type_aliases'

• On Python 3.8 machine (Windows 7 64 bits) the example runs without errors:

c:\Python38\python polarex.py
shape: (2, 2)
┌─────────────┬───────┐
│ strs        ┆ score │
│ ---         ┆ ---   │
│ str         ┆ u32   │
╞═════════════╪═══════╡
│ foo quz BAR ┆ 88    │
│ baaarfoo    ┆ 74    │
└─────────────┴───────┘

• The code:

  def polarsfm_usage():
  """ Taken from https://github.com/bnm3k/polars-fuzzy-match/blob/master/examples/usage.py """
  import polars as pl
  from polars_fuzzy_match import fuzzy_match_score
  
  df = pl.DataFrame(
      {
          'strs': ['foo', 'foo quz BAR', 'baaarfoo', 'quz'],
      }
  )
  pattern = 'bar'
  out = (
      df.with_columns(
          score=fuzzy_match_score(
              pl.col('strs'),
              pattern,
          )
      )
      .filter(pl.col('score').is_not_null())
      .sort(by='score', descending=True)
  )
  print(out)

---

What could be the reason for this difference?

Both machines can run this very basic polars script without errors:

import polars as pl
localcsv = "sample.csv"
df = pl.read_csv(localcsv, infer_schema_length=2, ignore_errors=True)
print(df.sample(25))
print(df.describe())

Thanks a lot in advance @abubelinha

[abubelinha]

Update:

Now I recall a problem when I installed polars yesterday in Python 3.8 / Windows 7. Although pip installation was completed without errors, importing polars raised an error (something I don't recall, related to CPU) and suggesting to try this package instead (and that's what I did): pip installpolars-lts-cpu After doing that, polars example worked (and polars-fuzzy-match usage.py example works too).

My installation of polars in Python 3.9 / Windows 10 did not raise such polars importing error, so that machine has the normal polars pip package, and my polars read_csv example script works good.

Oddly, the "normal" installation is the one where polars-fuzzy-match usage.py example fails.

[abubelinha]

I guess this was related to polars version, which was older in my Python 3.9 machine. Something around 0.16? ... sorry I didn't annotate it, but I upgraded polars to 0.20.13 and now polars-fuzzy-match usage.py example is working.

Perhaps polars-fuzzy-match installation could take care of this? (in case you can confirm a minimum polars version is required)

[bnm3k]

Yeah, I was about to add that if the issue is that polars.type_aliases cannot be imported then the problem's most likely with polars rather than the plugin. But seems you've figured it out :)