Skip to main content

Memory map WAVE or raw audio files

Project description

WaveMap logo

Manipulate huge WAVE or RAW files as numpy matrices - even if they are too large to fit into memory.

Memory mapping is a technique where files on disk are directly mapped to locations in memory and use the same logic as swap space does.

Samples from a WAVE or RAW audio file are directly memory mapped to entries in a numpy array, letting you manipulate very large audio files as if they all fit into memory at one time, and even directly change samples on disk.

Typical usage:

import wavemap

wm = wavemap('test.wav', 'r+')  # r+ means read/write
# Now you have a numpy matrix that you can use like any other

wm /= 2
# Each sample in the file is scaled by half.

API

wavemap()

wavemap(
     filename: str,
     mode: str='r',
     order: Union[str, NoneType]=None,
     always_2d: bool=False,
     dtype: Union[numpy.dtype, NoneType]=None,
     shape: Union[NoneType, int, tuple]=None,
     sample_rate: int=0,
     roffset: int=0,
     warn: Union[Callable, NoneType]='<function warn:  print to stderr>',
)

(wavemap/__init__.py, 56-121)

Memory map a WAVE file to a numpy array

Return an instance of ReadMap or WriteMap, depending on mode.

ARGUMENTS
filename

The name of the file being mapped

mode

The file is opened in this mode. Must be one of 'r', 'r+', 'c', 'w+'

In mode 'r', the default, the file is opened read-only and the numpy.darray is immutable.

In mode 'r+', the file is opened read-write and changes to the numpy.darray are automatically applied to the file.

In mode 'c', “copy-on-write”, the file is opened read-only, but the numpy.darray is not immutable: changes to the array are instead stored in memory.

In mode 'w+', “write”, the file is opened for write, and overwrites whatever else is there.

order

Samples usually get laid out in into a numpy.darray with`` shape=(N, C)`` where N is the number of audio frames, and C is the number of channels.

This is called column major order, but this can be toggled by setting the order parameter to F for Fortan or row-major row.

See https://stackoverflow.com/questions/27266338/

always_2d

If False, the default, mono WAVE files with only one channel get special treatment and are mapped to a one-dimensional vector with size=(N,).

If True, mono WAVE files are treated the same as any other file and are mapped to a two-dimensional matrix with size=(N, 1).

dtype

The numpy datatype of the samples in the file.

shape

The shape of the resulting numpy.darray. Can be a tuple, or a positive integer, or None.

sample_rate

The sample rate in Hz (cycles per second)

roffset

How many bytes in the file after the WAV data

warn

Programmers are sloppy so quite a lot of real-world WAVE files have recoverable errors in their format. warn is the function used to report those recoverable errors. By default, it’s set to print to sys.stderr but setting it to None disables errors entirely, or you can pass your own callback in

Class wavemap.RawMap

(wavemap/raw.py, 14-67)

“Memory map raw audio data from a disk file into a numpy matrix

wavemap.RawMap.__new__()

wavemap.RawMap.__new__(
     cls,
     filename: str,
     dtype: numpy.dtype,
     shape: Union[tuple, int, NoneType]=None,
     mode: str='r',
     offset: int=0,
     roffset: int=0,
     order: Union[str, NoneType]=None,
     always_2d: bool=False,
     warn: Union[Callable, NoneType]='<function warn:  print to stderr>',
)

(wavemap/raw.py, 17-67)

Memory map raw audio data from a disk file into a numpy matrix

ARGUMENTS
cls

Think of this as self. (This is because you need to implement __new__ and not __init__ when deriving from np.darray.)

filename

The name of the file being mapped

dtype

The numpy datatype of the samples in the file.

shape

The shape of the resulting numpy.darray. Can be a tuple, or a positive integer, or None.

mode

The file is opened in this mode. Must be one of 'r', 'r+', 'c', 'w+'

In mode 'r', the default, the file is opened read-only and the numpy.darray is immutable.

In mode 'r+', the file is opened read-write and changes to the numpy.darray are automatically applied to the file.

In mode 'c', “copy-on-write”, the file is opened read-only, but the numpy.darray is not immutable: changes to the array are instead stored in memory.

In mode 'w+', “write”, the file is opened for write, and overwrites whatever else is there.

offset

How many bytes in the file before the WAV data

roffset

How many bytes in the file after the WAV data

order

Samples usually get laid out in into a numpy.darray with`` shape=(N, C)`` where N is the number of audio frames, and C is the number of channels.

This is called column major order, but this can be toggled by setting the order parameter to F for Fortan or row-major row.

See https://stackoverflow.com/questions/27266338/

always_2d

If False, the default, mono WAVE files with only one channel get special treatment and are mapped to a one-dimensional vector with size=(N,).

If True, mono WAVE files are treated the same as any other file and are mapped to a two-dimensional matrix with size=(N, 1).

warn

Programmers are sloppy so quite a lot of real-world WAVE files have recoverable errors in their format. warn is the function used to report those recoverable errors. By default, it’s set to print to sys.stderr but setting it to None disables errors entirely, or you can pass your own callback in

Class wavemap.ReadMap

(wavemap/read.py, 18-84)

Memory-map an existing WAVE file into a numpy vector or matrix

wavemap.ReadMap.__new__()

wavemap.ReadMap.__new__(
     cls: Type,
     filename: str,
     mode: str='r',
     order: Union[str, NoneType]=None,
     always_2d: bool=False,
     warn: Union[Callable, NoneType]='<function warn:  print to stderr>',
)

(wavemap/read.py, 21-84)

Memory-map an existing WAVE file into a numpy matrix.

ARGUMENTS
cls

Think of this as self. (This is because you need to implement __new__ and not __init__ when deriving from np.darray.)

filename

The name of the file being mapped

mode

The file is opened in this mode. Must be one of 'r', 'r+' and 'c'.

In mode 'r', the default, the file is opened read-only and the numpy.darray is immutable.

In mode 'r+', the file is opened read-write and changes to the numpy.darray are automatically applied to the file.

In mode 'c', “copy-on-write”, the file is opened read-only, but the numpy.darray is not immutable: changes to the array are instead stored in memory.

order

Samples usually get laid out in into a numpy.darray with`` shape=(N, C)`` where N is the number of audio frames, and C is the number of channels.

This is called column major order, but this can be toggled by setting the order parameter to F for Fortan or row-major row.

See https://stackoverflow.com/questions/27266338/

always_2d

If False, the default, mono WAVE files with only one channel get special treatment and are mapped to a one-dimensional vector with size=(N,).

If True, mono WAVE files are treated the same as any other file and are mapped to a two-dimensional matrix with size=(N, 1).

warn

Programmers are sloppy so quite a lot of real-world WAVE files have recoverable errors in their format. warn is the function used to report those recoverable errors. By default, it’s set to print to sys.stderr but setting it to None disables errors entirely, or you can pass your own callback in

Class wavemap.WriteMap

(wavemap/write.py, 12-115)

“Memory-map a new wave file into a new numpy vector or matrix

wavemap.WriteMap.__new__()

wavemap.WriteMap.__new__(
     cls: Type,
     filename: str,
     dtype: numpy.dtype,
     shape: Union[NoneType, int, tuple],
     sample_rate: int,
     roffset: int=0,
     warn: Union[Callable, NoneType]='<function warn:  print to stderr>',
)

(wavemap/write.py, 15-85)

Open a memory-mapped WAVE file in write mode and overwrite any existing file.

ARGUMENTS
cls

Think of this as self. (This is because you need to implement __new__ and not __init__ when deriving from np.darray.)

filename

The name of the file being mapped

dtype

The numpy datatype of the samples in the file.

shape

The shape of the resulting numpy.darray. Can be a tuple, or a positive integer, or None.

sample_rate

The sample rate in Hz (cycles per second)

roffset

How many bytes in the file after the WAV data

warn

Programmers are sloppy so quite a lot of real-world WAVE files have recoverable errors in their format. warn is the function used to report those recoverable errors. By default, it’s set to print to sys.stderr but setting it to None disables errors entirely, or you can pass your own callback in

wavemap.convert()

wavemap.convert(
     arr: numpy.ndarray,
     dtype: Union[numpy.dtype, NoneType],
     must_copy: bool=False,
)

(wavemap/convert.py, 6-77)

Returns a copy of a numpy array or matrix that represents audio data in another type, scaling and shifting as necessary.

ARGUMENTS
arr

A numpy darray representing an audio signal

dtype

The numpy dtype to convert to - none means “no conversion”

must_copy

If true, arr is copied even if it is already the requested type

(automatically generated by doks on 2021-02-23T14:37:02.652534)

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

wavemap-2.0.1.tar.gz (19.7 kB view hashes)

Uploaded Source

Built Distribution

wavemap-2.0.1-py3-none-any.whl (18.7 kB view hashes)

Uploaded Python 3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page