未翻訳のページ

このページはまだ日本語に翻訳されていないため、英語の原文を表示しています。翻訳に協力する

sunau Module Complexity¶

The sunau module reads and writes Sun AU audio files: a short header followed by the samples. It is pure Python and works a call at a time: opening a file reads only the header, readframes() returns a block of frames, and writeframes() passes your bytes to the file, encoding them first for u-law. Nothing in the module holds the audio unless you read it all at once.

Removed in Python 3.13

Deprecated in Python 3.11 and removed in Python 3.13 (PEP 594). The page covers the module as it is on Python 3.11 and 3.12, and its examples run only there; there, import sunau emits a DeprecationWarning.

Frames are the unit throughout. k is the frames read or written in one call and w is the bytes in one frame as your code sees it, nchannels × sampwidth (for A-law, nchannels), so one call moves k·w bytes. Uncompressed samples are stored big-endian and are returned and written as stored, on every host; u-law samples are converted from and to native byte order. The official documentation calls the two classes AU_read and AU_write; the module names them Au_read and Au_write.

Complexity Reference¶

open and Error¶

Operation	Time	Space	Notes
`sunau.open(f, mode=None)`	O(1)	O(1)	`'r'` or `'rb'` returns an `Au_read`, `'w'` or `'wb'` an `Au_write`; without a mode, `f.mode` decides, else `'rb'`
`sunau.Error`	O(1)	O(1)	Raised for a file that is not AU or has an unsupported encoding, and for a parameter that is invalid, missing or set too late. A file that ends inside the header's fixed fields raises `EOFError` instead

Au_read¶

Operation	Time	Space	Notes
`sunau.Au_read(f)`, `sunau.open(f, 'rb')`	O(1)	O(1)	Reads the header and nothing else. Needs only `read()`: on a stream that cannot report its position, frames still read in order, but `setpos()` and `rewind()` raise `OSError`
`Au_read.readframes(nframes)`	O(k·w)	O(k·w), O(nframes·w) from a file on disk	k = frames returned, which is fewer than `nframes` at the end of the data; a file on disk sizes its read buffer for the request, so asking far past the end costs that much memory. u-law is decoded into a new buffer of 16-bit samples; A-law is returned undecoded, one byte per sample, though `getsampwidth()` reports 2
`Au_read.setpos(pos)`, `Au_read.rewind()`	O(1)	O(1)	Seek at once, reading nothing; `setpos()` raises `sunau.Error` for a position past the end
`Au_read.tell()`	O(1)	O(1)	Frames read so far, or the position set
`Au_read.getnchannels()`, `Au_read.getsampwidth()`, `Au_read.getframerate()`, `Au_read.getnframes()`	O(1)	O(1)	Read from the header when the file was opened. `getnframes()` returns `sunau.AUDIO_UNKNOWN_SIZE` when the header leaves the length unknown, and `readframes()` of that reads to the end
`Au_read.getcomptype()`, `Au_read.getcompname()`	O(1)	O(1)	`'NONE'`, `'ULAW'` or `'ALAW'`, and `'not compressed'`, `'CCITT G.711 u-law'` or `'CCITT G.711 A-law'`
`Au_read.getparams()`	O(1)	O(1)	A named tuple of the six values above
`Au_read.getmarkers()`, `Au_read.getmark(id)`	O(1)	O(1)	AU files have no markers: `getmarkers()` returns `None` and `getmark()` raises `sunau.Error`
`Au_read.close()`	O(1)	O(1)	Closes the file only if it was opened from a file name

Au_write¶

Operation	Time	Space	Notes
`sunau.Au_write(f)`, `sunau.open(f, 'wb')`	O(1)	O(1)	Writes nothing; the header goes out with the first frames or at `close()`. The compression type starts as `'ULAW'`, not `'NONE'`
`Au_write.setnchannels(n)`, `Au_write.setsampwidth(n)`, `Au_write.setframerate(n)`, `Au_write.setnframes(n)`	O(1)	O(1)	Raise `sunau.Error` once frames have been written; `setnchannels()` accepts 1, 2 or 4, `setsampwidth()` 1 to 4, and `setnframes()` no negative count
`Au_write.setcomptype(type, name)`	O(1)	O(1)	`'NONE'` or `'ULAW'`; any other type raises `sunau.Error`, and `name` is ignored. Not refused once frames have been written, though the header already names the first type
`Au_write.setparams(params)`	O(1)	O(1)	The five setters in one call
`Au_write.writeframes(data)`	O(k·w)	O(1), O(k·w) u-law	Uncompressed bytes are written as given, not copied; u-law is encoded into a new buffer. After every call the header's count is patched in place, unless the frames written so far match both the count declared with `setnframes()` and the count the header holds
`Au_write.writeframesraw(data)`	O(k·w)	O(1), O(k·w) u-law	The same without the patch, which is left to `close()`
`Au_write.tell()`, `Au_write.getnframes()`	O(1)	O(1)	Frames written so far
`Au_write.getnchannels()`, `Au_write.getsampwidth()`, `Au_write.getframerate()`	O(1)	O(1)	`getnchannels()` and `getframerate()` raise `sunau.Error` until set, and `getsampwidth()` until the frame rate is
`Au_write.getcomptype()`, `Au_write.getcompname()`	O(1)	O(1)	`'ULAW'` and `'CCITT G.711 u-law'` until `setcomptype()`
`Au_write.getparams()`	O(1)	O(1)	Raises `sunau.Error` until channels and frame rate are set
`Au_write.close()`	O(1)	O(1)	Writes the header if no frames were, patches it on the same condition as `writeframes()`, then flushes; closes the file only if it was opened from a file name

Constants¶

Operation	Time	Space	Notes
`sunau.AUDIO_FILE_MAGIC`	O(1)	O(1)	`.snd` read as a big-endian integer; every AU file starts with it
`sunau.AUDIO_FILE_ENCODING_MULAW_8`, `sunau.AUDIO_FILE_ENCODING_LINEAR_8`, `sunau.AUDIO_FILE_ENCODING_LINEAR_16`, `sunau.AUDIO_FILE_ENCODING_LINEAR_24`, `sunau.AUDIO_FILE_ENCODING_LINEAR_32`, `sunau.AUDIO_FILE_ENCODING_ALAW_8`	O(1)	O(1)	Header encodings the module reads; it writes all but A-law
`sunau.AUDIO_FILE_ENCODING_FLOAT`, `sunau.AUDIO_FILE_ENCODING_DOUBLE`, `sunau.AUDIO_FILE_ENCODING_ADPCM_G721`, `sunau.AUDIO_FILE_ENCODING_ADPCM_G722`, `sunau.AUDIO_FILE_ENCODING_ADPCM_G723_3`, `sunau.AUDIO_FILE_ENCODING_ADPCM_G723_5`	O(1)	O(1)	Known encodings the module does not support; opening such a file raises `sunau.Error`
`sunau.AUDIO_UNKNOWN_SIZE`	O(1)	O(1)	`0xFFFFFFFF`: the header's length when it is unknown, and what `getnframes()` then returns

Reading AU Files¶

Opening Reads Only the Header¶

The header is a fixed set of fields plus a short description, so opening a file costs the same however long the audio is. The frame count comes from the header's data size.

import io
import sunau

buffer = io.BytesIO()
with sunau.open(buffer, 'wb') as out:
    out.setparams((2, 2, 44100, 0, 'NONE', ''))
    out.writeframes(bytes(4 * 44100))  # one second of stereo silence

buffer.seek(0)
with sunau.open(buffer, 'rb') as audio:  # O(1) - the samples are not read
    assert audio.getnframes() == 44100  # O(1)
    assert audio.getparams() == (2, 2, 44100, 44100, 'NONE', 'not compressed')
    assert audio.getmarkers() is None

Reading in Blocks¶

readframes() returns what you ask for as one bytes object, so memory follows the request, not the file. Reading a fixed block at a time holds one block; readframes(getnframes()) holds the whole file.

import io
import sunau

buffer = io.BytesIO()
with sunau.open(buffer, 'wb') as out:
    out.setparams((1, 2, 8000, 0, 'NONE', ''))
    out.writeframes(bytes(2 * 10_000))

buffer.seek(0)
with sunau.open(buffer, 'rb') as audio:
    frames = 0
    while block := audio.readframes(4096):  # O(k·w) per block
        assert len(block) <= 4096 * 2
        frames += len(block) // 2
    assert frames == 10_000
    assert audio.readframes(4096) == b''  # at the end

Seeking¶

setpos() and rewind() seek the file at once, so the frames skipped are never read. They need a stream that can report its position; on one that cannot, both raise OSError.

import io
import sunau

buffer = io.BytesIO()
with sunau.open(buffer, 'wb') as out:
    out.setparams((1, 1, 8000, 0, 'NONE', ''))
    out.writeframes(bytes(range(256)) * 4)

buffer.seek(0)
with sunau.open(buffer, 'rb') as audio:
    audio.setpos(1000)  # O(1)
    assert audio.tell() == 1000
    assert audio.readframes(10) == bytes(range(232, 242))  # O(k·w)
    audio.rewind()  # O(1)
    assert audio.readframes(3) == bytes([0, 1, 2])

    try:
        audio.setpos(audio.getnframes() + 1)
    except sunau.Error as error:
        assert 'not in range' in str(error)
    else:
        raise AssertionError('a position past the end was accepted')

Reading from an Unseekable Stream¶

Opening reads the header straight through and readframes() reads on from there, so a pipe or socket can be read from start to end. Only setpos() and rewind() need to seek.

import io
import sunau

class Pipe:
    """Can be read from, but not sought or asked its position."""
    def __init__(self, data):
        self._data = io.BytesIO(data)
    def read(self, size=-1):
        return self._data.read(size)

buffer = io.BytesIO()
with sunau.open(buffer, 'wb') as out:
    out.setparams((1, 1, 8000, 0, 'NONE', ''))
    out.writeframes(bytes(range(100)))

with sunau.open(Pipe(buffer.getvalue()), 'rb') as audio:
    assert audio.readframes(60) == bytes(range(60))  # O(k·w)
    assert audio.readframes(60) == bytes(range(60, 100))
    try:
        audio.rewind()
    except OSError as error:
        assert 'cannot seek' in str(error)
    else:
        raise AssertionError('an unseekable stream was rewound')

Writing AU Files¶

The Default Compression Is u-law¶

A new writer encodes to u-law unless told otherwise, which halves 16-bit samples and loses precision. Call setcomptype('NONE', ...), or pass 'NONE' to setparams(), to store the bytes as given. A u-law writer encodes each call into a new buffer, and reading it back decodes into another.

import array
import io
import sunau

samples = array.array('h', [0, 1000, -1000, 32000])  # native byte order

buffer = io.BytesIO()
out = sunau.open(buffer, 'wb')  # O(1) - nothing written yet
out.setnchannels(1)
out.setsampwidth(2)
out.setframerate(8000)
assert out.getcomptype() == 'ULAW'
out.writeframes(samples)  # O(k·w), encoded to one byte per sample
out.close()
assert len(buffer.getvalue()) == 32 + len(samples)  # header + samples

buffer.seek(0)
with sunau.open(buffer, 'rb') as audio:
    assert audio.getsampwidth() == 2
    decoded = array.array('h', audio.readframes(4))  # O(k·w), decoded to 16 bits
    assert decoded[0] == 0 and decoded != samples  # close, but not the same
    assert all(abs(a - b) < 1000 for a, b in zip(decoded, samples))

Patching the Header¶

The header records the data size. After every call, writeframes() seeks back to patch it, a constant cost per call, unless the frames written so far match both the count declared with setnframes() and the count the header holds. Without a declared count, every call patches; declaring the total and writing it in one call needs none. writeframesraw() never patches, and close() then patches once on the same condition.

import io
import sunau

class Recorder(io.BytesIO):
    """A BytesIO that counts its seeks."""
    seeks = 0
    def seek(self, *args):
        self.seeks += 1
        return super().seek(*args)

in_parts = Recorder()
with sunau.open(in_parts, 'wb') as out:
    out.setparams((1, 2, 8000, 300, 'NONE', ''))
    for _ in range(3):
        out.writeframes(bytes(2 * 100))  # O(k·w), header patched each call
assert in_parts.seeks == 6  # two seeks per patch

declared = Recorder()
with sunau.open(declared, 'wb') as out:
    out.setparams((1, 2, 8000, 300, 'NONE', ''))  # O(1)
    out.writeframes(bytes(2 * 300))  # O(k·w), the header is already right
assert declared.seeks == 0

# Parameters are fixed once frames are written
out = sunau.open(io.BytesIO(), 'wb')
out.setparams((1, 2, 8000, 0, 'NONE', ''))
out.writeframes(bytes(2))
try:
    out.setframerate(22050)
except sunau.Error as error:
    assert 'cannot change parameters' in str(error)
else:
    raise AssertionError('a parameter changed after writing')
out.close()

Writing to an Unseekable Stream¶

A pipe or socket cannot be patched. Declare the total, write with writeframesraw(), and close() finds nothing to patch; the target needs write() and flush(), and nothing else.

import sunau

class Pipe:
    """Accepts writes but cannot seek, like a pipe."""
    def __init__(self):
        self.chunks = []
    def write(self, data):
        self.chunks.append(bytes(data))
        return len(data)
    def flush(self):
        pass

pipe = Pipe()
with sunau.open(pipe, 'wb') as out:
    out.setparams((1, 2, 8000, 1000, 'NONE', ''))  # O(1)
    out.writeframesraw(bytes(2 * 500))  # O(k·w), no patch
    out.writeframesraw(bytes(2 * 500))

assert sum(map(len, pipe.chunks)) == 32 + 2 * 1000  # header + samples

Common Patterns¶

Converting AU to WAV a Block at a Time¶

Uncompressed AU samples are big-endian, while wave takes and returns samples in native byte order, so 16-bit frames are byte-swapped on a little-endian host. Doing it a block at a time keeps memory at one block however long the file is.

import array
import io
import sunau
import sys
import wave

samples = array.array('h', range(-5000, 5000))
if sys.byteorder == 'little':
    samples.byteswap()  # AU stores big-endian

source = io.BytesIO()
with sunau.open(source, 'wb') as out:
    out.setparams((1, 2, 8000, 0, 'NONE', ''))
    out.writeframes(samples)

source.seek(0)
target = io.BytesIO()
with sunau.open(source, 'rb') as audio, wave.open(target, 'wb') as out:
    out.setnchannels(audio.getnchannels())
    out.setsampwidth(audio.getsampwidth())
    out.setframerate(audio.getframerate())
    while block := audio.readframes(1024):  # O(k·w)
        values = array.array('h', block)
        if sys.byteorder == 'little':
            values.byteswap()  # big-endian to native, O(k·w)
        out.writeframes(values)  # O(k·w), wave takes native order

target.seek(0)
with wave.open(target, 'rb') as audio:
    converted = array.array('h', audio.readframes(audio.getnframes()))
assert converted[0] == -5000 and converted[-1] == 4999

Performance Best Practices¶

✅ Do:

Read with a fixed block size, so memory is one block rather than the file
Call setcomptype('NONE', ...) for uncompressed output; a new writer encodes to u-law
Write an unseekable output with setnframes() and writeframesraw(), so no call needs to seek

❌ Avoid:

readframes(getnframes()) on a long file - the one call here that holds every frame at once
setpos() or rewind() on a stream that cannot report its position - both raise OSError
Setting parameters once frames are written - all but setcomptype() raise sunau.Error, and that one silently mixes encodings
Expecting close() to close a file object you passed in - only a file opened by name is closed

Version Notes¶

Python 3.11+: import sunau emits DeprecationWarning
Python 3.13+: The module is removed; import sunau raises ModuleNotFoundError
All Python 3: Uncompressed samples are returned and written as stored, big-endian

wave - The same frame-at-a-time interface for WAV files, still in the standard library
array - Turn a block of frames into samples and back
aifc - AIFF files, removed in Python 3.13
audioop - The u-law conversions sunau uses, removed in Python 3.13