Structured arrays (arrays)

Provide structured arrays for cosmological data.

DataArray()	Abstract data array with save and load methods.
SphericalArray(disc)	Structured array for spherically decomposed cosmological data.
CartesianArray(orders, wavenumbers[, ...])	Structured array for Cartesian decomposition of cosmological data.

exception harmonia.algorithms.arrays.IndexingError

Exception raised for unsupported slicing or indexing in __getitem__ methods.

class harmonia.algorithms.arrays.DataArray

Abstract data array with save and load methods.

save(output_file, file_extension)

Save the structured array.

Parameters:

• output_file (str or pathlib.Path) – Output file path.

• extension ({‘.pkl’, ‘.npz’}) – Output file extension.

classmethod load(input_file)

Load the structured array from a .npz or .pkl file.

Parameters:

input_file (str or pathlib.Path) – Input file path.

class harmonia.algorithms.arrays.SphericalArray(disc)

Structured array for spherically decomposed cosmological data.

Array is initialised with a discrete spectrum of Fourier modes and consists of three fields: the ‘index’ field of \((\ell, m_\ell, n_\ell)\) triplets, the ‘wavenumber’ field of discrete \(k_{\ell n}\), and the ‘coefficient’ field of spherically decomposed data.

Parameters:

disc (DiscreteSpectrum) – Discrete spectrum associated with the structured array.

Variables:

• array (numpy.ndarray) – Structured NumPy array.

• size (int) – Total number of elements in the array. This should equal the sum of disc.mode_counts.

• attrs (dict) – Attributes of the structured array inherited from disc.

See also

DiscreteSpectrum

__getitem__(key)

Get the ‘coefficient’ field value(s).

The access key can be an integer, a slice expression, a tuple of index triplet or a string, e.g. [-1], [:], [(0, 0, 1)] or 'degree_0'.

Parameters:

key (int, slice, tuple(int, int, int) or str) – ‘coefficient’ field access key.

Returns:

‘coefficient’ field data entry.

Return type:

complex

__setitem__(key, value)

Set the ‘coefficient’ field value(s).

Parameters:

• key (int, tuple of int or slice) – ‘coefficient’ field access key.

• value (complex) – ‘coefficient’ field data entry.

See also

SphericalArray.__getitem__()

vectorise(pivot, collapse=None)

Returrn a data vector from the ‘coefficient’ field.

Vectorisation is performed by pivoting in either of the following orders of precedence—

• ‘natural’: ordered by \((\ell, m, n)\);

• ‘spectral’: ordered by \((k_{\ell n}, m)\).

Subarrays of equivalent \((\ell, n)\) may be further collapsed over spherical order \(m\) by simple averaging or averaging in quadrature.

Parameters:

• pivot ({‘natural’, ‘spectral’}) – Pivot order for vectorisation.

• collapse ({None, ‘mean’, ‘qaudrature’}, optional) – If not None (default), subarrays are collapsed over equivalent spherical order \(m\) by averaging (‘mean’) or averaging in quadrature (‘qaudrature’).

Returns:

vectorised_data – Vectorised coefficient data.

Return type:

numpy.ndarray

class harmonia.algorithms.arrays.CartesianArray(orders, wavenumbers, mode_counts=None, shot_noise=None)

Structured array for Cartesian decomposition of cosmological data.

Array is initialised with three fields: the ‘order’ field of the Legendre multipoles, the ‘wavenumber’ field of \(k\)-bin centres and the ‘power’ field for power spectrum multipole measurements.

Parameters:

• orders (list of tuple of int) – Orders of the power spectrum multipole.

• wavenumbers (float, array_like) – Wavenumbers of the multipole data.

• mode_counts (list of int or None, optional) – Mode counts in wavenumber bins (default is None).

• shot_noise (float or None, optional) – Shot noise level (default is None).

Variables:

• array (numpy.ndarray) – Structured NumPy array.

• size (int) – Total number of elements in the array. This should equal the product of the numbers of wavenumbers and multipoles.

• attrs (dict) – Initialisation parameters as attributes.

__getitem__(key)

Access the ‘power’ field.

The access key can be an integer positional index, a slice, a tuple of (order, wavenumber) or a string, e.g. [-1], [:], [(0, 0.04)] or 'power_0'.

Parameters:

key (int, slice, tuple(int, float) or str) – ‘power’ field access key.

Returns:

‘power’ field data entry.

Return type:

float

__setitem__(key, value)

Set the ‘power’ field value(s).

Parameters:

• key (int, tuple(int, float) or slice) – ‘power’ field access key.

• value (float) – ‘power’ field data entry.

See also

CartesianArray.__getitem__()

vectorise(pivot)

Return a data vector from the ‘power’ field.

Vectorisation is performed by pivoting in either of the following orders of precedence—

• ‘order’: ordered by multipole order \(\ell\);

• ‘wavenumber’: ordered by wavenumber \(k\).

Parameters:

pivot ({‘order’, ‘wavenumber’}) – Pivot order for vectorisation.

Returns:

vectorised_data – Vectorised power spectrum data.

Return type:

numpy.ndarray