Sync main to release (v1.1.2)

[Erikpostt]

Release v1.1.2

Release date: 15/06/2026

Description

This PR synchronizes the main branch into release to prepare for the v1.1.2 release of ParaDigMa.

Changes

Breaking Changes

• Unified pipeline return format: All pipelines (run_gait_pipeline(), run_tremor_pipeline(), run_pulse_rate_pipeline()) now return a unified dictionary structure instead of tuples/DataFrames:
  • Previously:
    • Gait: (quantified_df_dict: dict, metadata_dict: dict) (from v1.1.1)
    • Tremor/Pulse Rate: DataFrame
  • Now: All return unified dict with keys:
    • 'quantification': Computed results (dict for gait, DataFrame for others)
    • 'metadata': Metadata dict (gait only)
    • 'preprocessing', 'classification': Intermediate results (if requested via return_intermediate)
    • '_steps_executed': List of steps that were actually executed
    • '_error': Error message (None if successful)

Features

• Selective pipeline step execution: New parameters run_steps and return_intermediate in all pipeline functions and orchestrator:

  • run_steps: Which processing steps to execute ('preprocessing', 'classification', 'quantification', 'aggregation')
  • return_intermediate: Which intermediate results to include in return dict (independent of disk storage)
  • Available in both individual pipeline functions and main run_paradigma() orchestrator
  • Enables parallel processing workflows where preprocessing/classification run separately from quantification
  • Example: run_steps=['preprocessing', 'classification'] to stop after classification without computing quantification

• Adaptive sampling frequency detection: orchestrator now automatically detects data sampling frequency and adjusts frequency-dependent parameters accordingly. It is no longer necessary to manually set the sampling frequency to 100 Hz

• Independent save/return control: save_intermediate and return_intermediate now work independently:

  • Save results to disk without returning them (memory-efficient for large datasets)
  • Return intermediate results without saving to disk (useful for debugging)
  • Previously, saving automatically returned results in memory

• Orchestrator updates: run_paradigma() now supports selective step execution and handles unified dict returns from all pipelines:

  • Updated to work with new pipeline return formats (unified dict structure)
  • Properly handles gait filtered/unfiltered results in combined output
  • Enables conditional execution paths (e.g., only run quantification for pre-processed data)

Improvements

• Added comprehensive error handling with _error field in all pipeline function returns.
• All pipeline functions now log which steps are being executed and returned.

Checklist

• Tests passed
• Documentation updated
• Version bumped in pyproject.toml (1.1.1 → 1.1.2)

[Erikpostt]

@KarsVeldkamp @nienketimmermans Note that this is a workaround for the merge conflicts in #299. To reproduce:

1. Pull release
2. Create a new branch from release
3. Merge main into this new branch (locally)
4. Push new branch and merge into release

This seems to solve the diverging histories. I suggest after the release to remove the main branch and re-create it from release, such that the histories align. Let's do a little bit of research this does not break stuff.

[nienketimmermans]

Looks good! @KarsVeldkamp let's discuss whether we want to delete the main branch or if this workaround is good enough for future releases.