wemoloh / frepr

Fast float repr() for CPython using https://github.com/google/double-conversion.

frepr

Fast float repr() for CPython using https://github.com/google/double-conversion. Approximately 8X faster than the built-in function.

Quick Start

>>> import frepr
>>> frepr.frepr(-1.8397892359791307e+134)
'-1.8397892359791307e+134'
>>> from timeit import timeit
>>> timeit('repr(-1.8397892359791307e+134)')
2.7296602272476758
>>> frepr.install()
>>> timeit('repr(-1.8397892359791307e+134)')
0.35723431229251545
>>> frepr.uninstall()

API

frepr.frepr(value)

Returns shortest string representation of a floating point value. The argument doesn't have to be a float, as long as it is coercible to one (i.e. it has a __float__() method).

frepr.install()

Monkey-patches the Python runtime. Subsequent calls to repr() on floats will use the faster algorithm. This is safe to call multiple times.

frepr.uninstall()

Reverses the monkey-patching done by install(). Subsequent calls to repr() will use the original built-in algorithm. This is safe to call multiple times, even if install() was not called first.

Features

• On average, approximately 8X faster than the standard function. (This result obtained on Python 3.6.0 64-bit/Windows 7/Intel Core i3 @ 2.93GHz. Test your own system with tests/perf.py).
• Speeds up the standard json package (and any other library that relies on repr()).
• Output is guaranteed to contain the shortest string of decimal digits necessary to uniquely identify the original IEEE 754 floating-point value.
• Guaranteed to satisfy value == ast.literal_eval(frepr(value)).
• Tested on CPython 2.6 through 3.6; probably works on other versions too.

Limitations

• Currently, there is no support for alternative Python implementations (e.g. PyPy).
• Output can be slightly different from the native function, due to floating-point rounding (but see accuracy guarantees above).
• Has no effect on other built-in methods of converting floats to strings (e.g. str(), str.format(), "%f", "%g").

See Also

• https://github.com/google/double-conversion - Core algorithm implementation. This is the same implementation used by Chrome's V8 JavaScript engine, among others.
• https://bugs.python.org/issue12450 - CPython issue from 2011 about this idea.