Randle Taylor

Formatting floats to a specific number of significant digits in Python

Often when you are writing scientific code you want to display numbers with a specific number of significant digits. This is easily achievable using Python's exponential format specifier: %e or %E. For example if you want to display the number 1.23 to 4 significant digits you can do "%.3E" % (1.23) and Python will correctly print 1.230E+00. However, sometimes you would rather have more "friendly" formatting for small numbers (e.g. 40.54 instead of 4.054E+01) and fall back to exponential notation when numbers are much greater than 1 or much smaller than 1. The "general" format specifiers %g and %G come very close to the correct behavior. The general format specifier is described in the docs as follows:

General format. For a given precision p >= 1, this rounds the number to p significant digits and then formats the result in either fixed-point format or in scientific notation, depending on its magnitude.

The precise rules are as follows: suppose that the result formatted with presentation type 'e' and precision p-1 would have exponent exp. Then if -4 <= exp < p, the number is formatted with presentation type 'f' and precision p-1-exp. Otherwise, the number is formatted with presentation type 'e' and precision p-1. In both cases insignificant trailing zeros are removed from the significand, and the decimal point is also removed if there are no remaining digits following it.

Note the bit I've emphasised in the last paragraph. Even though the docs say that only "insignficant trailing zeros's" are trimmed this is a little bit misleading. For example, if you tried "%.4G" % (1.230) you might expect Python to print "1.230" i.e. you would expect 4 significant figures to be included. What actually happens is that Python trims that trailing zero even though it is a significant digit in this case. So when you try "%.4G" % (1.230) what you get is "1.23" with only 3 significant figures.

I was playing around with some Javascript and noticed the Number.toPrecision method does precisely what we are looking for. If you enter (1.230).toPrecision(4) in your Javascript console, Javascript will correctly print "1.230", likewise if you try (123000000).toPrecision(4) you will correctly get "1.230e+8".

I wanted to duplicate this behaviour in Python so I dug around in the WebKit source code for the toPrecision method and ported it to Python as seen below:

def to_precision(x,p):
    """
    returns a string representation of x formatted with a precision of p

    Based on the webkit javascript implementation taken from here:
    https://code.google.com/p/webkit-mirror/source/browse/JavaScriptCore/kjs/number_object.cpp
    """

    x = float(x)

    if x == 0.:
        return "0." + "0"*(p-1)

    out = []

    if x < 0:
        out.append("-")
        x = -x

    e = int(math.log10(x))
    tens = math.pow(10, e - p + 1)
    n = math.floor(x/tens)

    if n < math.pow(10, p - 1):
        e = e -1
        tens = math.pow(10, e - p+1)
        n = math.floor(x / tens)

    if abs((n + 1.) * tens - x) <= abs(n * tens -x):
        n = n + 1

    if n >= math.pow(10,p):
        n = n / 10.
        e = e + 1

    m = "%.*g" % (p, n)

    if e < -2 or e >= p:
        out.append(m[0])
        if p > 1:
            out.append(".")
            out.extend(m[1:p])
        out.append('e')
        if e > 0:
            out.append("+")
        out.append(str(e))
    elif e == (p -1):
        out.append(m)
    elif e >= 0:
        out.append(m[:e+1])
        if e+1 < len(m):
            out.append(".")
            out.extend(m[e+1:])
    else:
        out.append("0.")
        out.extend(["0"]*-(e+1))
        out.append(m)

    return "".join(out)

Now you can have nicely formatted numbers with the correct number of significant digits always preserved!

>>> to_precision(1.23,4)
'1.230'
>>> to_precision(123000000,4)
'1.230e+8'
>>> to_precision(0.00000123,7)
'1.230000e-6'
>>>

The code is on GitHub and hopefully someone else finds this function useful.

Update: BebeSparkelSparkel has added more functionality to my original implementation. His version can be found here: https://github.com/BebeSparkelSparkel/to-precision.