Probably one could write a very flexible (= allows many formats) parser with only few options: - decimal separator: point or comma - space allowed inside number representation or stops parsing - default number base - acceptance of Unicode special characters (Unicode Minus, superscript digits, arabic-indic digits, \cdot) - optional: which special values and which formats: Inf, #-INF, *INF, NaN_0xDEADBEEF, "positive-infinity" Or even restrict (some of) those options further to reduce the number of needed options. Parsing the exponent for special variants: - Exponent started with x10 or \cdot10 (\cdot the centered dot character) - Exponent with p/P for hexadecimal numbers (where e has another meaning as digit) - Exponent as E(13) with parantheses What is missing?