There are countless blog posts and stack overflow explanations about this, figured I need to write this little note for myself. To start things off, check this example of computing PHPKRW using hypothetical gold as a go-between:
require "bigdecimal"
# Float
php_gold_float = 0.00001144792038
gold_krw_float = 1962165.41
php_krw_float = php_gold_float * gold_krw_float
puts "#{php_krw_float.class} #{php_krw_float}"
# Big Decimal
php_gold_bd = BigDecimal.new("0.00001144792038")
gold_krw_bd = BigDecimal.new("1962165.41")
php_krw_bd = php_gold_bd * gold_krw_bd
puts "#{php_krw_bd.class} #{php_krw_bd.to_s('F')}"
# Output
Float 22.462713386070057
BigDecimal 22.4627133860700558
That’s odd, it’s rounded up? Or something? Why does this happen?
Long version, read up on IEEE specification to define floats. Short version: it stores a sign, a fraction and an exponent to represent a Float. It’s a scientific notation for binary, because of that it’s impossible to store exact numbers and decimal in Float.
If you’re a bank, or if you’re supporting accounting software, or just computing compounding interests, you could potentially miscalculate thousands down the line when using Float for computing. So, make sure to use BigDecimal when you’re dealing with anything that requires a high-degree of “sureness”. With Ruby/Rails, you can use :decimal for columns containing this type of information.
Money, accounting, computing compounding interests = needs high precision, then use BigDecimal. Other than that, just use float.
Some gotcha’s: With Ruby/Rails, when rendering an object with BigDecimal attributes as JSON, it will render the numbers as a String, instead of a Float. Part of the reasoning behind this change could be found in this discussion