'Fastest way to convert numeric chars to int

From looking over here and other websites I know there are two common ways to convert a numeric char value like '5' to an int value:

  1. Using Character.getNumericValue()

  2. Subtracting the number with the ASCII value for '0'; i.e. int number = num - '0', where num is a char value.

Which of these two approaches is the fastest and most efficient?

javachardata-conversion


Solution 1:[1]

  1. The two versions are not equivalent:

    • The Character.getNumericalValue(...) methods work for a variety of characters that represent digits or numbers, and it will return -1 or -2 in cases where the character doesn't represent a non-negative integer.
    • The num - '0' approach only gives the correct answer for the codepoints that correspond to the ASCII characters '0' through '9'. For all other codepoints or codeunits, it gives a meaningless value.

  2. The num - '0' version will be faster. This is clear from looking at the source code for getNumericalValue(...).

  3. While the difference is significant in relative terms, it is very small in absolute terms.

I concur with the comments that say that this is most likely a premature optimization.

It is also an incorrect optimization in some contexts.

I use it a lot so was wondering if I was using the most efficient one.

This is definitely premature optimization :-)

The number of times you write a particular code sequence is unrelated to performance of the code sequence when is executed. A section of code is only worth optimizing if the time spent executing it makes a significant difference to your entire application.

Solution 2:[2]

Well, Character.getNumericValue() takes Unicode, radix, case, culture into account:

 '0' ->  0 // Same as number - '0'
 '9' ->  9 // Same as number - '0'
 'A' -> 10 // Hexadecimal 0xA == 10
 'f' -> 15 // Hexadecimal 0xF == 15
 '³' ->  3 // Unicode superscript 3
 '?'-> 20 // Unicode enclosed alphanumeric 20 
 '?' ->  5 // Persian digit 5
 '*' -> -1 // Doesn't have any corresponding integer value
 '?' -> -2 // Even if 5/6 fraction Unicode character is a number, it's not integer

while

number - '0'

is just a subtraction of two ints. That's why Character.getNumericValue() is inevitably slower (some nanoseconds, is it worth optimizing?). But, please, note, that in 'A', 'f', '³', '*' etc. cases you are going to have wrong asnwers with number - '0' code.

Sources

This article follows the attribution requirements of Stack Overflow and is licensed under CC BY-SA 3.0.

Source: Stack Overflow

Solution Source
Solution 1
Solution 2

Related Questions