Skip to content

Latest commit

 

History

History
52 lines (41 loc) · 3.48 KB

README.md

File metadata and controls

52 lines (41 loc) · 3.48 KB
Raw

Ballpark

When people think of human-readable numbers, they think of rounding to two decimal places and adding a thousands separator. 12,214.17 is already quite an improvement over 12214.16666667. But standard formats for human-readable numbers still have various flaws:

  • even with a thousands separator, at a glance you might easily mistake a billion for a trillion
  • even when rounding, an amount like 12,214.17 dollars is a lot of number noise for communicating 12.2K
  • scientific notation leads to exponents like 1.22e4 which are hard to interpret because we're used to working with thousands, millions and billions – orders of magnitudes that are multiples of three
  • when comparing multiple measurements of the same underlying variable, like the yearly sales numbers for 2010-2015, it's annoying to have some numbers in thousands and other numbers in millions – you want consistency so that digits in the same position are of the same magnitude

python-ballpark introduces business notation, an offshoot of engineering notation, for producing better human-readable numbers.

Install with pip install ballpark or pip3 install ballpark.

What it looks like

numbers rounded engineering notation business notation
11234.22, 233000.55, 1175125.2 11,234.22, 233,000.55, 1,175,125.2 11.2E+3, 233E+3, 1.18E+6 11K, 233K, 1,180K
111, 1111.23, 1175125.234 111, 1,111.23, 1,175,125.23 111, 1.11E+3, 1.18E+6 0.11K, 1.11K, 1,180.00K

How to use it

>>> from ballpark import human, scientific, engineering, business, ballpark
>>> business([11234.22, 233000.55, 1175125.2])
['11K', '233K', '1,180K']
>>>
>>> # business notation is also aliased as `ballpark`
>>> ballpark([11234.22, 233000.55, 1175125.2])
['11K', '233K', '1,180K']
>>>
>>> # or use the shortcut functions
>>> from ballpark import H, S, E, B
>>> B([11234.22, 233000.55, 1175125.2])
['11K', '233K', '1,180K']
>>>
>>> # all notations accept single numbers too, but then we can't guarantee
>>> # that all numbers will have the same prefix (kilo, mega etc.)
>>> [B(value) for value in [11234.22, 233000.55, 1175125.2]]
['11.2K', '233K', '1.18M']

How it works

business(values, precision=3, prefix=True, prefixes=SI, statistic=median)
  • precision: the amount of significant digits. When necessary, business will round beyond the decimal sign as well: in the example above, 1175125.2 was turned into 1,180K rather than 1,175K to retain only 3 significant digits.
  • prefix: whether to use SI prefixes like m (milli), K (kilo) and so on instead of scientific exponents like E+03.
  • prefixes: a mapping of orders of magnitude to prefixes, e.g. {-3: 'm', 3: 'K'}, allowing you to customize the prefixes, for example using B for billion instead of T for tera.
  • statistic: a function to produce the reference number. The reference number determines the order of magnitude and precision for the entire group of numbers, so that for example when the reference number is 23.3K, smaller numbers like 1.1K won't gain a decimal place and larger numbers like 1,180K won't jump an order of magnitude to 1.18M. The median often works well, but if you want more precision for small outliers, try ballpark.statistics.Q1 or even Python's builtin min.