Anatoly Myachev

I found another place that block async execution (dtypes computing if there isn't cache for that): https://github.com/modin-project/modin/blob/adb16a17f721048005520388080627975c6852d8/modin/core/io/file_dispatcher.py#L167

@jbrockmendel I tried this approach, it became slower to work on our script (because of Ray's work with object store as @mvashishtha mentioned). In addition, for some operations, for example...

This should help us for `to_pandas` function because we most likely already have index in the main process.

Current state (for dd27013 in master): - ~ 50 min - 24 parallel jobs

Further improvements should be for: - `test_series.py` - `test_groupby.py` - `test+default.py`

Hi @wqh17101! Can you reproduce `TypeError: 'LocalFileOpener' object is not iterable` problem with following reproducer? ```python import fsspec file_path = r"./test.csv" file = fsspec.open(file_path).open() for line in file: print(line) ```

For reference: problem with Ray is reported [here.](https://github.com/ray-project/ray/issues/23046)

Thanks @wqh17101! This looks like a problem with Windows 11 support on `fsspec` side (which `Modin` use for reading files, just like `pandas`). Can you create an issue [there](https://github.com/fsspec/filesystem_spec)?

@wqh17101 thanks for working on that! We should think about upgrading minimum version of `fsspec` thinking about compatibility with pandas. @jbrockmendel doesn't the problem affect pandas too? Maybe you faced...

thanks @jbrockmendel! @wqh17101 could you check fsspec v2021.5.0 with Windows 11 (I haven't the OS)? If this works, then a future release of Modin that supports pandas 1.5.0 will also...