Reading¶

For reading, instantiate a mmnpz.NpzReader, also available under the alias mmnpz.load().

mmnpz.load[source]¶: alias of NpzReader

class mmnpz.NpzReader(fn, mmap_mode='r', *, cache=True, preload=False)[source]¶

A dictionary-like object with lazy-loading of numpy arrays in the given uncompressed .npz file. Upon construction, creates a memory map of the full .npz file, returning views for the arrays within on request. Serves as a drop-in replacement for numpy.load().

Parameters:

fn (str or Path) – The zipped archive to open.
mmap_mode (str, optional) – Provided for compatibility with numpy. Only supports ‘r’, indicating that the file is loaded as a readonly memory map.
cache (bool, optional) – Whether to cache array objects in case they are requested again.
preload (bool, optional) – Whether to precreate all array objects upon opening. Enforces caching.

Raises:

OSError – If the input file does not exist or cannot be read.
zipfile.BadZipFile – If the input file cannot be interpreted as a zip file.

Examples

Store data to disk, and load it again as a memory map:

>>> np.savez("/tmp/123.npz", a=np.array([[1, 2], [3, 4]]))
>>> data = mmnpz.NpzReader("/tmp/123.npz")
>>> data["a"]
memmap([[1, 2],
        [3, 4]])

files¶

List of all uncompressed files in the archive with a .npy extension (listed without the extension). These are supported as dictionary keys.

Type:: list of str

mmap¶

The memory map of the full .npz file.

Type:: numpy.memmap

arrays¶

Preloaded or cached arrays.

Type:: dict

load(name)[source]¶

Creates a view for the specified array, disregarding the cache.

Parameters:: name (str) – File name in the archive to load, without its .npy extension.
Returns:: A view into the global memory map with correct shape and dtype.
Return type:: numpy.memmap