RAN Utils#

General utility functions for RAN algorithms.

Overview#

The RAN Utils module provides:

  • HDF5 I/O - Load and save data in HDF5 format

  • Data Transformation - Key mapping and data manipulation

  • Measurements - dB conversions and calculations

  • Timing Utilities - Performance measurement tools

API Reference#

Utils package.

ran.utils.db(x: numpy.typing.ArrayLike) → FloatArrayNP[source]#

Compute 10*log10(x) elementwise.

Parameters:

x (array-like) – Linear-scale input. Must be positive to avoid -inf results.

Returns:

Decibel-scale output as float64.

Return type:

np.ndarray

ran.utils.hdf5_load(filename: Path | str) → dict[str, Any][source]#

Load data from an HDF5 file.

Parameters:

filename (Path) – Path object pointing to the HDF5 file

Returns:

dict

Return type:

Dictionary containing datasets

ran.utils.hdf5_save(
filename: Path | str,
data: dict[str, Any],
) → None[source]#

Save a flat dict of numpy arrays/scalars to an HDF5 file.

Parameters:

  • filename (Path | str) – Output HDF5 file path

  • data (dict[str, Any]) – Dictionary mapping keys to numpy arrays or scalars

Notes

Behavior matches the inverse of hdf5_load:

  • Real arrays are saved transposed

  • Complex arrays are saved as structured dataset with re and im fields, each already transposed

  • Python scalars are saved as 0-d datasets

  • Only top-level datasets are supported (no groups), matching the loader

ran.utils.load_config(cfg_path: Path) → dict[source]#

Load configuration from YAML file.

Parameters:

cfg_path – Path to config file.

Returns:

Configuration dictionary

ran.utils.max_abs_diff(
x: Any,
y: Any,
eps: float = 1e-12,
) → tuple[float, float, float, float][source]#

Compare two arrays and optionally print the differences.

Parameters:

  • x – First array to compare.

  • y – Second array to compare.

  • eps – Small value to avoid division by zero.

Returns:

  • max_abs (Maximum absolute difference.)

  • mean_abs (Mean absolute difference.)

  • max_rel_pct (Maximum relative percentage difference.)

  • mean_rel_pct (Mean relative percentage difference.)