Skip to content

Windows Compatibility Issue: DataFrame Materialization and File Permission Errors When Writing Parquet Files #713

@drockparashar

Description

@drockparashar

Description

When running the SDK on Windows, DataFrame writes to Parquet files (using pandas/pyarrow) frequently fail due to file permission errors and file locking issues. This does not occur on Unix-like systems.

Reproduction Repo/Script (if any)

N/A (Issue occurs with any workflow in the SDK that writes DataFrames to Parquet on Windows)

Reproduction Steps

1. Import SDK
2. Call method that writes a DataFrame to Parquet (e.g., dataframe.to_parquet(...))
3. Observe error: PermissionError: [Errno 13] Permission denied or file locking issues

Logs / Tracebacks

PermissionError: [Errno 13] Permission denied: 'output_file.parquet'
# Or: FileNotFoundError, zero-byte files, access denied errors

Expected vs Actual

Expected:

  • DataFrame writes to Parquet should succeed on Windows as they do on Unix-like systems.
  • No file locking or permission errors should occur.

Actual:

  • Frequent permission errors and file locking issues.
  • Sometimes files are not created, are zero bytes, or require manual cleanup.

Environment

Windows

SDK Version

0.1.1rc38

Metadata

Metadata

Assignees

No one assigned

    Labels

    bugSomething isn't working

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions