Miscellaneous¶

Logging Integration¶

The fixedpoint package uses standard logging constructs to generate warnings and debug logs. The following specifications allow you to customize the logging schemes, handlers, formats, etc. to your liking.

Warnings¶

fixedpoint warnings are generated using the logging package.

fixedpoint.logging.WARNER_CONSOLE_HANDLER¶

Type

logging.StreamHandler

Value

stream is set to sys.stderr
level is set to logging.DEBUG.

fixedpoint.logging.WARNER¶

Type

logging.Logger

Value

name is FP.CONSOLE
level defaults to logging.WARNING

You can retrieve this logger with logging.getLogger("FP.CONSOLE").

The WARNER_CONSOLE_HANDLER is added to WARNER.

Additionally, each FixedPoint object has a unique serial number associated with it that is available in the LogRecord instance as the key sn. This is described in more detail here as the extra keyword.

Debug Logging¶

Logging is also used for debug purposes.

fixedpoint.logging.LOGGER¶

Type

logging.Logger

Value

name is FP
level defaults to logging.CRITICAL

fixedpoint.logging.DEFAULT_FILE_HANDLER¶

Type

logging.FileHandler

Value

filename is set to fixedpoint.log located in the same directory as the source code
mode is set to ‘w’
delay is set to True, thus no file is generated (or overwritten) until logging is enabled with fixedpoint.FixedPoint.enable_logging()
level defaults to logging.DEBUG

The DEFAULT_FILE_HANDLER is added to the LOGGER.

On initial import, LOGGER’s level is set to logging.CRITICAL. Since no critical level logs are made within fixedpoint, it essentially disables debug logging.

When FixedPoint.enable_logging() is called, LOGGER’s level is set to logging.DEBUG.

When FixedPoint.disable_logging() is called, LOGGER’s level is set back to logging.CRITICAL.

Typing¶

The fixedpoint package is typed (see PEP 484) and supported by mypy, PyCharm, etc.

Subclassing FixedPoint is also supported.

Numpy Integration¶

While not specifically tested, integration with numpy should be possible as long as unsupported operators (like @, /, //, %, etc.) are not used.

Examples taken from the numpy.convolve documentation:

>>> import numpy as np
>>> a = [FixedPoint(1), FixedPoint(2), FixedPoint(3)]
>>> b = [FixedPoint(0), FixedPoint(1), FixedPoint(0.5)]
>>> x = np.convolve(a, b)
>>> [float(fp) for fp in x]
[0.0, 1.0, 2.5, 4.0, 1.5]

>>> y = np.convolve(a, b, 'same')
>>> [float(fp) for fp in y]
[1.0, 2.5, 4.0]

>>> z = np.convolve(a, b, 'valid')
>>> [float(fp) for fp in z]
[2.5]

Serialization¶

JSON¶

class fixedpoint.json.JSONEncoder¶

class fixedpoint.json.JSONDecoder¶

>>> from fixedpoint.json import FixedPointEncoder, FixedPointDecoder
>>> import random, json
>>> L = 52
>>> signed = random.randrange(2)
>>> m = random.randrange(L)
>>> n = L - m
>>> original = FixedPoint(hex(random.getrandbits(L)), signed, m, n)
>>> serialized = json.dumps(original, cls=FixedPointEncoder)
>>> deserialized = json.loads(serialized, cls=FixedPointDecoder)
>>> original == deserialized
True

Pickle¶

The pickle scheme works out of the box:

>>> import random, pickle
>>> L = 52
>>> signed = random.randrange(2)
>>> m = random.randrange(signed, L)
>>> n = L - m
>>> original = FixedPoint(hex(random.getrandbits(L)), signed, m, n)
>>> pickled = pickle.dumps(original)
>>> unpickled = pickle.loads(pickled)
>>> original == unpickled
True