Rethinking number formatting
Problem statement
A friend of mine, let’s call him Jack, asked me a seemingly innocent question (I’m paraphrasing):
How can I [in PHP1] format a number so that the number of decimals gets padded by spaces (not zeros) to a defined fixed length?
Which is another way of saying that Jack wants 12.345 and 123.4 formatted
as strings "12.345_" and "123.4___" (respectively)2.
We had a lively discussion about it… and in the end concluded there’s no simple3 way to do it.
I mean, there’s:
<?php
print_r preg_replace_callback("/0*$/",
function ($x) { return str_repeat("_", strlen($x[0])); },
sprintf("%.8f", 12.345)) . "\n";
# => "12.345_____\n"
and it’s equivalent in Ruby:
("%.8f" % 12.345).sub(/0*$/) { |x| "_" * x.length }
# => "12.345_____"
But the sad fact is, sprintf sucks.
Can we do better?
Discussion
I think there is a problem with documentation and sprintf-style format strings
in Jack’s question.
While you can right-pad the entire number with spaces
(sprintf("%-8.4f", 12.345)) the documentation isn’t very clear on the fact
that you can’t right/left pad individual parts of the number4. Which becomes
rather important when you’re going for this effect:
| Item | Price |
|---|---|
| A | 12.345_ |
| B | 123.4___ |
So we need a formatting function that allows to specify:
- Minimal length of the whole numbers part
- Padding character of the whole numbers part (and it’s direction)
- Maximal length of the decimal part (if any)
- Padding character of the decimal part (and it’s direction)
- Decimal separator character
- [Optional] thousands separator character
Solution
I think the above can be written as a wrapper around sprintf.
DEFAULT_FORMAT = {
whole_min_length: 0,
whole_padding: ' ',
whole_direction: 'l',
decimal_max_length: nil, # ≤ 15, tbh
decimal_padding: '0',
decimal_direction: 'r',
decimal_separator: '.',
thousands_separator: '\'',
}
def format(number, **opts)
fmt = DEFAULT_FORMAT.dup.update(opts)
whole, decimal = number.to_s.split(".")
if fmt[:thousands_separator]
whole = whole.reverse.scan(/.{1,3}/).join(fmt[:thousands_separator]).reverse
end
if whole.size < fmt[:whole_min_length]
case fmt[:whole_direction]
when 'l'
whole = whole.rjust(fmt[:whole_min_length], fmt[:whole_padding])
when 'r'
whole = whole.ljust(fmt[:whole_min_length], fmt[:whole_padding])
else
# we could throw an error but we'll just refuse to pad instead
end
end
if fmt[:decimal_max_length]
ml = fmt[:decimal_max_length]
if decimal.size > ml
# round
if (0..4).include?(decimal[ml,1])
decimal = decimal[0, ml]
else
decimal = decimal[0, ml].succ
end
elsif decimal.size < ml
# pad
case fmt[:decimal_direction]
when 'l'
decimal = decimal.rjust(ml, fmt[:decimal_padding])
when 'r'
decimal = decimal.ljust(ml, fmt[:decimal_padding])
else
# we could throw an error but we'll just refuse to pad instead
end
end
end
[whole, decimal].join(fmt[:decimal_separator])
end
if __FILE__ == $0
puts format(12.345, thousands_separator: ',') # => "12.345"
puts format(12312.345, whole_min_length: 7) # => " 12'312.345"
puts format(0.12345678901234567890) # => "0.12345678901234568"
puts format(0.12345678901234567890, decimal_max_length: 7) # => "0.1234568"
puts format(12.345, decimal_max_length: 8, decimal_padding: '_', decimal_direction: 'r')
# => "12.345_____"
puts format(12.345, decimal_max_length: 8, decimal_padding: '_', decimal_direction: 'l')
# => "12._____345"
puts format(12312.345, whole_min_length: 7,whole_direction: 'r', decimal_direction: 'l',
decimal_max_length: 8, decimal_padding: ' ') # => "12'312 . 345"
end
But don’t ask me to write it in PHP. That’d be the end of me.
Closing words
Seeing how many moving parts there are, I’m not sure it could be compressed into a sprintf-style format string. But I’m pretty sure someone will eventually try.
Just like someone already implemented a Minecraft-powered CPU @ 1Hz.