Remove zero-len arrays from test dataset

[coreyjadams]

PhysicsNeMo Pull Request

Fixes errors in the release candidate branch.

Description

Checklist

• I am familiar with the Contributing Guidelines.
• New or existing tests cover these changes.
• The documentation is up to date with these changes.
• The CHANGELOG.md is up to date with these changes.
• An issue is linked to this pull request.

Dependencies

[greptile-apps]

Greptile Overview

Greptile Summary

Removed scalar fields air_density and stream_velocity from the synthetic_domino_data() function. These scalar values were causing issues when saved to zarr format as they were being treated as zero-length arrays.

The fix is correct because:

• The test suite already provides these values via the keys_to_read_if_available parameter (line 289-292) with hardcoded tensors [1.225, 10.0]
• CAEDataset has a built-in mechanism to fill missing keys from keys_to_read_if_available (see cae_dataset.py:116-120)
• The synthetic test data should only contain array data, not scalar parameters
• This aligns with the proper separation of concerns: test data files contain mesh/field arrays, while global parameters come from configuration

Confidence Score: 5/5

• This PR is safe to merge - it's a targeted bug fix that removes problematic scalar values from test data generation
• The change correctly addresses the zero-length array issue by removing scalar values that were improperly included in the synthetic test dataset. The existing keys_to_read_if_available fallback mechanism already handles these values properly, making the removal safe and correct. The fix is minimal, well-targeted, and doesn't affect production code or real data pipelines.
• No files require special attention

Important Files Changed

File Analysis

Filename	Score	Overview
test/datapipes/test_domino_datapipe.py	5/5	Removed scalar fields air_density and stream_velocity from synthetic test data generation - these values are already provided via keys_to_read_if_available fallback mechanism

Sequence Diagram

sequenceDiagram
    participant Test as Test Suite
    participant SynthData as synthetic_domino_data()
    participant Zarr as Zarr File
    participant CAEDataset as CAEDataset
    participant DataPipe as DoMINODataPipe
    
    Note over Test,DataPipe: Before: Scalar values causing issues
    Test->>SynthData: Generate test data
    SynthData->>SynthData: Create out_dict with arrays
    SynthData->>SynthData: Add air_density scalar (np.float32)
    SynthData->>SynthData: Add stream_velocity scalar (np.float32)
    SynthData->>Zarr: Save dict with scalars
    Note over Zarr: Scalars treated as zero-length arrays
    Test->>CAEDataset: Create dataset with keys_to_read_if_available
    CAEDataset->>Zarr: Read data files
    CAEDataset->>CAEDataset: Fill missing keys from keys_to_read_if_available
    CAEDataset->>DataPipe: Provide data
    
    Note over Test,DataPipe: After: Clean separation of concerns
    Test->>SynthData: Generate test data
    SynthData->>SynthData: Create out_dict with arrays only
    Note over SynthData: Scalars removed from synthetic data
    SynthData->>Zarr: Save dict (arrays only)
    Test->>CAEDataset: Create dataset with keys_to_read_if_available
    Note over CAEDataset: global_params from keys_to_read_if_available
    CAEDataset->>Zarr: Read data files
    CAEDataset->>CAEDataset: Fill global_params from fallback
    CAEDataset->>DataPipe: Provide complete data

Loading

_{1 file reviewed, no comments}

_{Edit Code Review Agent Settings | Greptile}

[coreyjadams]

/blossom-ci

[greptile-apps]

Greptile Overview

Greptile Summary

Removes two unused scalar fields (air_density and stream_velocity) from the synthetic test dataset generation in test_domino_datapipe.py. These fields were never read by the DoMINO datapipe or used in any tests.

• The datapipe uses keys_to_read_if_available with default values for global_params_values and global_params_reference (both [1.225, 10.0])
• The removed fields were only being written to test data files but never accessed
• This cleanup makes the test dataset more accurate by only including fields that are actually used

Confidence Score: 5/5

• This PR is safe to merge with no risk
• The change only removes unused test data fields that were never read by the datapipe or referenced in any tests. The removed fields (air_density and stream_velocity) were dead code that had no functional impact. All necessary data fields remain intact, and the datapipe uses default values from keys_to_read_if_available instead.
• No files require special attention

Important Files Changed

File Analysis

Filename	Score	Overview
test/datapipes/test_domino_datapipe.py	5/5	Removed unused scalar fields (air_density, stream_velocity) from synthetic test dataset generation - these fields were never read by the datapipe

Sequence Diagram

sequenceDiagram
    participant Test as Test Suite
    participant SyntheticData as synthetic_domino_data()
    participant DataFile as Test Data Files
    participant Dataset as CAEDataset
    participant DataPipe as DoMINODataPipe
    
    Test->>SyntheticData: Generate test data
    SyntheticData->>SyntheticData: Create mesh & field data
    Note over SyntheticData: Removed: air_density, stream_velocity<br/>(unused scalar fields)
    SyntheticData->>DataFile: Save to zarr/npz/npy
    Test->>Dataset: Create CAEDataset
    Note over Dataset: keys_to_read_if_available:<br/>global_params_values=[1.225, 10.0]<br/>global_params_reference=[1.225, 10.0]
    Dataset->>DataFile: Read keys_to_read
    DataFile-->>Dataset: Return data
    Dataset->>Dataset: Fill missing keys with defaults
    Test->>DataPipe: Create DoMINODataPipe
    DataPipe->>Dataset: Request data
    Dataset-->>DataPipe: Return complete dataset
    DataPipe-->>Test: Process and return sample

Loading

_{1 file reviewed, no comments}

_{Edit Code Review Agent Settings | Greptile}