---

name: visualize_fields description: 多次元フィールドデータ（Field4D/Matrix等）の抽出ロジックと描画APIを、物理的整合性を保ちながら実装する

Visualize Fields Best Practices

When implementing visualization for multi-dimensional data (3D, 4D, Matrix, etc.), design using the following three-layer structure to ensure maintainability and physical accuracy.

1. Infrastructure (Coordinate & Utility Layer)

Separate the conversion logic between physical coordinates and array indices from the class body, and implement it as reusable functions.

• nearest_index(axis, value): A nearest-neighbor search that ensures mutual conversion between unit systems (Astropy Units).
• select_value(data, mode): Extract real, abs, power, etc., from complex data.
• Location: gwexpy/plot/utils.py or _coord.py. This prevents the data type classes (Types layer) from directly depending on plotting libraries (such as Matplotlib).

2. Extraction API Layer

Add methods to data type classes that generate "subsets" for visualization.

• extract_points / slice_map2d: Returns data formatted for immediate use in drawing.
• Convention: Maintain the original class (e.g., Field4D) where possible, or return standard TimeSeries/FrequencySeries.

3. Plotting API Layer

Drawing methods called directly by the user.

• Naming Convention: plot_map2d, plot_profile, plot_timeseries_points, etc.
• Spectral Visualization:
  • freq_space_map (Waterfall, etc.) is a 2D map where the time axis is replaced by a frequency axis.
  • When plotting spectral density (PSD), consider using a log scale (norm=LogNorm or set_yscale('log')) by default.

Physical Consistency Checklist

• Unit Propagation: Does the unit become unit^2 in power mode? Is angle in rad?
• Coordinate Accuracy: Do the values of the plotted axis correspond correctly to the physical axis of the original data via nearest_index?
• Dimensional Maintenance: Are dimensions unexpectedly lost (Squeeze) through slicing operations, breaking the drawing logic?
• Memory Efficiency: For large-scale data, are unnecessary copies (copy=True) avoided?