如何在python中为任意预修复符号设置数字格式（k表示千等）？

我希望在传递到绘图注释时设置数字的格式，以便它们“漂亮地打印”，用“k”表示千等，并用精度四舍五入表示，例如3位精度：

13.1    -> 13.1
13.1676 -> 13.2
1246    -> 1.25k
560254  -> 560k
6.234e7 -> 62.3M

我已经编写了一个函数来实现这一点，但它似乎过于复杂：

import math


def num_fmt(num):
    i_offset = 15 # change this if you extend the symbols!!!
    prec = 3
    fmt = '.{p}g'.format(p=prec)
    symbols = ['Y', 'T', 'G', 'M', 'k', '', 'm', 'u', 'n']

    e = math.log10(abs(num))
    if e >= i_offset + 3:
        return '{:{fmt}}'.format(num, fmt=fmt)
    for i, sym in enumerate(symbols):
        e_thresh = i_offset - 3 * i
        if e >= e_thresh:
            return '{:{fmt}}{sym}'.format(num/10.**e_thresh, fmt=fmt, sym=sym)
    return '{:{fmt}}'.format(num, fmt=fmt)

---

类似的方法可能会奏效：

from bisect import bisect

# A mapping of breakpoints for suffixes.
suffixes = {
    1e-9: 'n',
    1e-6: 'u',
    1e-3: 'm',
    1: '',
    1e3: 'k',
    1e6: 'M',
    1e9: 'G',
    1e12: 'T',
    1e15: 'Y',
}

# List of sorted breakpoints' values.
suffix_breakpoints = sorted(suffixes.iterkeys())

def format_with_suffix(num):
    num_format = '{:.2f}{}'
    if not num:
       return num_format.format(num, '')

    if num in suffixes:
       return num_format.format(1.0, suffixes[num])

    # Find the index of first breakpoint `x`, that's greater than `num`
    # using binary search.
    breakpoint_idx = bisect(suffix_breakpoints, num)

    # Get the breakpoint's value. If `num` is lower than the first breakpoint, use
    # that instead.
    breakpoint = suffix_breakpoints[breakpoint_idx - 1 if breakpoint_idx else 0]

    # Get the suffix.
    suffix = suffixes[breakpoint]
    return num_format.format(float(num) / breakpoint, suffix)

>>> for x in xrange(-10, 20):
>>>     print format_with_suffix(10.0 ** x), 10.0 ** x
0.10n 1e-10
1.00n 1e-09
10.00n 1e-08
100.00n 1e-07
1.00u 1e-06
10.00u 1e-05
100.00u 0.0001
1.00m 0.001
10.00m 0.01
100.00m 0.1
1.00 1.0
10.00 10.0
100.00 100.0
1.00k 1000.0
10.00k 10000.0
100.00k 100000.0
1.00M 1000000.0
10.00M 10000000.0
100.00M 100000000.0
1.00G 1000000000.0
10.00G 10000000000.0
100.00G 1e+11
1.00T 1e+12
10.00T 1e+13
100.00T 1e+14
1.00Y 1e+15
10.00Y 1e+16
100.00Y 1e+17
1000.00Y 1e+18
10000.00Y 1e+19

---

@达诺，我不认为这是这个问题的重复