Utilities

confetti.utils.array_to_string(arr: ndarray) → str

Convert a NumPy array into a single-line space-separated string.

Useful for serializing arrays into CSV files where nested structures are not supported.

Parameters:

arr (ndarray) – The array to be flattened and serialized.

Returns:

A flattened, space-separated representation of the array.

Return type:

str

confetti.utils.convert_string_to_array(data_string: str, timesteps: int, channels: int) → ndarray

Convert a serialized array string into a 2D NumPy array.

This function reconstructs a flattened, space-separated string representation of an array into a 2D array of shape (timesteps, channels). Brackets and line breaks are removed automatically.

Parameters:

  • data_string (str) – The flattened array stored as a whitespace-separated string.

  • timesteps (int) – Number of expected time steps in the reconstructed array.

  • channels (int) – Number of expected channels in the reconstructed array.

Returns:

The reconstructed numeric array.

Return type:

ndarray of shape (timesteps, channels)

Raises:

CONFETTIConfigurationError – If the number of parsed elements does not match timesteps * channels.

confetti.utils.load_multivariate_ts_from_csv(file_path: str | Path) → Tuple[ndarray, ndarray]

Load a multivariate time-series dataset saved with save_multivariate_ts_as_csv.

This function reconstructs the original X array by reshaping the long-format CSV and retrieves the sample-level labels from the companion *_labels.csv file.

Parameters:

file_path (str or Path) – Path to the main feature CSV file.

Returns:

  • X (ndarray of shape (n_samples, n_time_steps, n_features)) – The reconstructed multivariate time series.

  • y (ndarray of shape (n_samples,)) – The label associated with each sample.

Note

The function expects two files in the same directory:

  • <file>.csv

  • <file>_labels.csv

confetti.utils.save_multivariate_ts_as_csv(file_path: str | Path, x: ndarray, y: ndarray) → None

Save a multivariate time-series dataset to CSV in long format.

The time-series array x is converted into a long-format table where each row corresponds to one sample–time-step pair. The function also writes a companion *_labels.csv file storing a single label per sample.

Parameters:

  • file_path (str or Path) – Destination path for the main CSV file.

  • x (ndarray of shape (n_samples, n_time_steps, n_features)) – The multivariate time series data.

  • y (ndarray of shape (n_samples,)) – The label associated with each sample.

Note

Two files are written:

  • <file>.csv: long-format feature table

  • <file>_labels.csv: labels indexed by sample_id