skip to navigation
skip to content

fastnumbers 2.0.1

Super-fast and clean conversions to numbers.

Super-fast and clean conversions to numbers.

fastnumbers is a module with the following three objectives:

  1. Provide drop-in replacements for the Python built-in int and float that on average are up to 2x faster. These functions should behave identically to the Python built-ins except for a few specific corner-cases as mentioned in the API documentation.
  2. Provide a set of convenience functions that wrap the above int and float replacements and provides easy, concise, powerful, fast and flexible error handling.
  3. Provide a set of functions that can be used to rapidly identify if an input could be converted to int or float.


The below examples showcase the fast_float function, which is a fast conversion function with error-handling. Please see the API Documentation for other functions that are available from fastnumbers.

>>> from fastnumbers import fast_float, float as fnfloat
>>> # Convert string to a float
>>> fast_float('56.07')
>>> # Unconvertable string returned as-is by default
>>> fast_float('bad input')
'bad input'
>>> # Unconvertable strings can trigger a default value
>>> fast_float('bad input', default=0)
>>> # 'default' is also the first optional positional arg
>>> fast_float('bad input', 0)
>>> # Integers are converted to floats
>>> fast_float(54)
>>> # One can ask inf or nan to be substituted with another value
>>> fast_float('nan')
>>> fast_float('nan', nan=0.0)
>>> fast_float(float('nan'), nan=0.0)
>>> fast_float('56.07', nan=0.0)
>>> # The default built-in float behavior can be triggered with
>>> # "raise_on_invalid" set to True.
>>> fast_float('bad input', raise_on_invalid=True) #doctest: +IGNORE_EXCEPTION_DETAIL
Traceback (most recent call last):
ValueError: invalid literal for float(): bad input
>>> # A key function can be used to return an alternate value for invalid input
>>> fast_float('bad input', key=len)
>>> fast_float(54, key=len)
>>> # Single unicode characters can be converted.
>>> fast_float(u'\u2164')  # Roman numeral 5 (V)
>>> fast_float(u'\u2466')  # 7 enclosed in a circle

NOTE: If you need locale-dependent conversions, supply the fastnumbers function of your choice to locale.atof.

import locale
locale.setlocale(locale.LC_ALL, 'de_DE.UTF-8')
print(atof('468,5', func=fast_float))  # Prints 468.5


Just how much faster is fastnumbers than a pure python implementation? Below are the timing results for the fast_float and float function on Python 2.7; please see the Timing Documentation for details into all timing results.

from timeit import timeit
float_try = '''\
def float_try(input):
    """Typical approach to this problem."""
        return float(input)
    except ValueError:
        return input

float_re = '''\
import re
float_match = re.compile(r'[-+]?\d*\.?\d+(?:[eE][-+]?\d+)?$').match
def float_re(input):
    """Alternate approach to this problem."""
        if float_match(input):
            return float(input)
            return input
    except TypeError:
        return float(input)

print('Invalid input:')
print("try:", timeit('float_try("invalid")', float_try))
print("re:", timeit('float_re("invalid")', float_re))
print("fast", timeit('fast_float("invalid")', 'from fastnumbers import fast_float'))
print('Valid input:')
print("try:", timeit('float_try("56.07e14")', float_try))
print("re:", timeit('float_re("56.07e14")', float_re))
print("fast", timeit('fast_float("56.07e14")', 'from fastnumbers import fast_float'))
print('Built-in float compared to fastnumbers.float:')
print("Built-in:", timeit('float("56.07e14")'))
print("fastnumbers:", timeit('float("56.07e14")', 'from fastnumbers import float'))

The results will be similar to below, but vary based on your system:

Invalid input:
try: 2.09141492844
re: 0.724852085114
fast 0.181249141693

Valid input:
try: 0.365114927292
re: 1.42145609856
fast 0.228940963745

Built-in float compared to fastnumbers.float:
Built-in: 0.234441041946
fastnumbers: 0.228511810303

As you can see, in all cases fastnumbers beats the pure python implementations (although not always significant).


Seth M. Morton


These are the last three entries of the changelog. See the package documentation for the complete changelog.

04-30-2017 v. 2.0.1

  • Fixed bug in decimal digit limit on GCC.

04-30-2017 v. 2.0.0

  • Dropped support for Python 2.6.
  • Added support for Python 3.6 underscores.
  • Added drop-in replacements for the built-in int() and float() functions.
  • Incorporated unit tests from Python’s testing library to ensure that any input that Python can handle will also be handled the same way by fastnumbers.
  • Added Appveyor testing to ensure no surprises on Windows.
  • Revamped documentation.
  • Refactored internal mechanism for assessing overflow to be faster in the most common cases.

04-23-2016 v. 1.0.0

  • “coerce” in fast_real now applies to any input, not just numeric; the default is now True instead of False.
  • Now all ASCII whitespace characters are stripped by fastnumbers
  • Typechecking is now more forgiving
  • fastnumbers now checks for errors when converting between numeric types
  • Fixed bug where very small numbers are not converted properly
  • Testing now includes Python 2.6.
  • Removed safe_* functions (which were deprecated since version 0.3.0)
  • Fixed unicode handling on Windows.
  • Fixed Python2.6 on Windows.

03-19-2016 v. 0.7.4

  • Added the “coerce” option to fast_real.
File Type Py Version Uploaded on Size
fastnumbers-2.0.1-cp27-cp27m-macosx_10_11_x86_64.whl (md5) Python Wheel cp27 2017-05-01 24KB
fastnumbers-2.0.1-cp27-cp27m-win32.whl (md5) Python Wheel cp27 2017-05-01 24KB
fastnumbers-2.0.1-cp27-cp27m-win_amd64.whl (md5) Python Wheel cp27 2017-05-01 24KB
fastnumbers-2.0.1-cp33-cp33m-macosx_10_9_x86_64.whl (md5) Python Wheel cp33 2017-05-01 23KB
fastnumbers-2.0.1-cp33-cp33m-win32.whl (md5) Python Wheel cp33 2017-05-01 23KB
fastnumbers-2.0.1-cp33-cp33m-win_amd64.whl (md5) Python Wheel cp33 2017-05-01 23KB
fastnumbers-2.0.1-cp34-cp34m-macosx_10_11_x86_64.whl (md5) Python Wheel cp34 2017-05-01 23KB
fastnumbers-2.0.1-cp34-cp34m-win32.whl (md5) Python Wheel cp34 2017-05-01 23KB
fastnumbers-2.0.1-cp34-cp34m-win_amd64.whl (md5) Python Wheel cp34 2017-05-01 23KB
fastnumbers-2.0.1-cp35-cp35m-macosx_10_11_x86_64.whl (md5) Python Wheel cp35 2017-05-01 23KB
fastnumbers-2.0.1-cp35-cp35m-win32.whl (md5) Python Wheel cp35 2017-05-01 25KB
fastnumbers-2.0.1-cp35-cp35m-win_amd64.whl (md5) Python Wheel cp35 2017-05-01 26KB
fastnumbers-2.0.1-cp36-cp36m-macosx_10_11_x86_64.whl (md5) Python Wheel cp36 2017-05-01 24KB
fastnumbers-2.0.1-cp36-cp36m-win32.whl (md5) Python Wheel cp36 2017-05-01 25KB
fastnumbers-2.0.1-cp36-cp36m-win_amd64.whl (md5) Python Wheel cp36 2017-05-01 26KB
fastnumbers-2.0.1.tar.gz (md5) Source 2017-05-01 292KB (md5) MS Windows installer 2.7 2017-05-01 247KB (md5) MS Windows installer 3.3 2017-05-01 246KB (md5) MS Windows installer 3.4 2017-05-01 246KB (md5) MS Windows installer 3.5 2017-05-01 605KB (md5) MS Windows installer 3.6 2017-05-01 605KB
fastnumbers-2.0.1.win32-py2.7.exe (md5) MS Windows installer 2.7 2017-05-01 220KB
fastnumbers-2.0.1.win32-py3.3.exe (md5) MS Windows installer 3.3 2017-05-01 215KB
fastnumbers-2.0.1.win32-py3.4.exe (md5) MS Windows installer 3.4 2017-05-01 215KB
fastnumbers-2.0.1.win32-py3.5.exe (md5) MS Windows installer 3.5 2017-05-01 478KB
fastnumbers-2.0.1.win32-py3.6.exe (md5) MS Windows installer 3.6 2017-05-01 478KB