Skip to content

KNOWN_ISSUES.md

Latest commit

 

History

History
143 lines (88 loc) · 5.25 KB

KNOWN_ISSUES.md

File metadata and controls

143 lines (88 loc) · 5.25 KB

Known Issues

This document tracks known issues and limitations in the DataOps Toolkit.

Last updated: January 2025

Test Issues

Quality Score Test Threshold Mismatch

File: tests/test_data_quality_tools.py::TestQualityScore::test_poor_quality_detection

Issue: The test expects a quality score < 70 for poor quality data, but the actual score calculated is 78.14. This indicates the quality scoring algorithm is slightly more lenient than expected.

Impact: Test-only issue, does not affect functionality. The quality score tool works correctly but may rate data quality slightly higher than the test expectations.

Workaround: Skip this specific test when running the test suite:

pytest -k "not test_poor_quality_detection"

Status: Low priority - The quality score calculation is working as designed, the test expectations may need adjustment.

Flaky ULID Generation Test

File: tests/test_utils.py::TestIds::test_generate_ulid

Issue: The test that verifies ULID ordering (assert id2 > id1) occasionally fails due to timing issues. ULIDs generated in the same millisecond may not have guaranteed ordering due to the random component.

Impact: Test-only issue, does not affect functionality. The ULID generation itself works correctly.

Workaround: The test can be skipped or the assertion can be relaxed to only check uniqueness rather than ordering.

Status: Low priority - ULIDs are unique which is the primary requirement. Ordering within the same millisecond is not critical for the audit trail functionality.

Pandas Compatibility Warnings

DataFrame dtype FutureWarnings

Issue: Some operations may trigger pandas FutureWarnings about dtype compatibility when filling NaN values with strings in numeric columns.

Affected Tools:

  • csv_merge (fixed in latest version)
  • Other CSV manipulation tools may show similar warnings

Resolution: Convert columns to object dtype before filling with string values to avoid the warning.

Date Format Inference Warnings

Issue: Multiple tools show warnings about date format inference:

UserWarning: Could not infer format, so each element will be parsed individually,
falling back to `dateutil`. To ensure parsing is consistent and as-expected,
please specify a format.

Affected Tools:

  • csv_profile - When detecting datetime columns
  • quality_score - When analyzing timeliness dimension
  • csv_clean - When standardizing date formats
  • map_suggest - When analyzing data patterns
  • schema_infer - When inferring column types

Impact: Cosmetic warning only, functionality is not affected. Date parsing still works correctly but may be slightly slower.

Workaround: These warnings can be suppressed or ignored. To fix, tools would need to explicitly specify date formats when calling pd.to_datetime().

Status: Low priority - Functionality works correctly, warnings are cosmetic.

Data Type Handling

Empty String vs NaN in CSV Files

Issue: When reading CSV files, pandas converts empty strings to NaN by default. This can cause issues when testing for empty values.

Affected Areas: Test assertions, data validation

Workaround: Use pd.read_csv(..., keep_default_na=False) when empty strings need to be preserved, or use .isna() checks instead of string comparisons.

Performance Considerations

Large File Processing

Issue: Some tools may be slow or memory-intensive when processing very large CSV files (>1GB).

Affected Tools:

  • All tools that load entire CSV into memory
  • Particularly: csv_profile, quality_score, dedupe_er

Workaround:

  • Use the --sample option where available to process a subset of data
  • Use csv_split to break large files into chunks first
  • Consider using csv_sql with DuckDB for better large file handling

Installation Issues

Python-Levenshtein Dependency

Issue: The dedupe_er tool requires python-Levenshtein for optimal fuzzy matching performance, which may require compilation on some systems.

Workaround: If installation fails, the tool will fall back to pure Python implementation (slower but functional). To install with C extensions:

# macOS
brew install python-Levenshtein

# Ubuntu/Debian
sudo apt-get install python3-levenshtein

# Or via pip with build tools installed
pip install python-Levenshtein

MCP Server Issues

Auto-discovery Limitations

Issue: The MCP auto-discovery server may not immediately reflect new tools added while the server is running.

Workaround: Restart the MCP server after adding new tools to ensure they are discovered.

Platform-Specific Issues

Windows Path Handling

Issue: Some path-related operations may behave differently on Windows due to path separator differences.

Affected Areas: File path arguments, glob patterns

Workaround: Use forward slashes (/) in paths even on Windows, or use Path objects from pathlib.

Reporting New Issues

If you encounter an issue not listed here, please report it at: https://github.com/yourusername/dataops-toolkit/issues

When reporting, please include:

  1. Tool name and command used
  2. Error message or unexpected behavior
  3. Sample data to reproduce (if possible)
  4. Python version and OS
  5. Output of pip list for dependency versions