read_file_in_chunks fails when pandas not available

[f-hafner]

Bug

Reading a file in chunks fails with output format polars when pandas is not in the virtual environment.

To reproduce

# script.py
import polars as pl
import pyreadstat
import tempfile

if __name__ == "__main__":

    with tempfile.NamedTemporaryFile(suffix=".sav") as f:
        data_dict = {
            "a": [1,2,3,4],
            "b": [10,11,12,13]
        }
        data = pl.DataFrame(data_dict)
        pyreadstat.write_sav(data, f.name)

        for chunk, _ in pyreadstat.read_file_in_chunks(
               read_function=pyreadstat.read_sav,
               chunksize=2,
               file_path=f,
               output_format="polars",
               ):
            assert isinstance(chunk, pl.DataFrame)
            print("Success.")

Run with

uv run --with "polars,pandas,pyreadstat" script.py # works
uv run --with "polars,pyreadstat" script.py # fails

Error message:

Traceback (most recent call last):
  File "pyreadstat/_readstat_parser.pyx", line 1276, in pyreadstat._readstat_parser.run_conversion
    import pandas
ModuleNotFoundError: No module named 'pandas'
[...]
pyreadstat._readstat_parser.PyreadstatError: You requested pandas as output_format but cannot import pandas

File example

see above

Expected behavior

No error when pandas is not in the venv.

Setup Information:

How did you install pyreadstat? uv (pip, conda, directly from repo)
Platform (windows, macOS, linux, 32 or 64 bit) linux 64bit
Python Version 3.14.4
Python Distribution (System, plain python, Anaconda): via uv
Using Virtualenv or condaenv? with inline dependencies via uv

Solution

I think the the output_format argument is not passed when reading the metadata here?

pyreadstat/pyreadstat/pyreadstat.py

Line 1043 in e23b1d8

_, meta = read_function(file_path, metadataonly=True)