How to test if a string only contains a number

> Have you measured the overhead when using regex in

> this case?

Can't be bothered - but the most efficient way regex could work on it would be to check the value range of each char, once. Which is what the "manual" solution does as well. Now considering that patterns need to be compiled before used, this compilation time can be considered overhead, although, admittedly, it will hardly matter if it's a one-time action on user input.

Which solution would be the best one IMO strongly depends on the allowed value range and the further use of that String, which the OP never specified.

CeciNEstPasUnProgrammeura at 2007-7-12 19:29:53 >

> > > sabre> it's true that parseInt won't take into

> > > considerations all kind of number, that's why i

> > also

> > > suggested regex

> >

> > You edited your post after I quoted it!

>

> actually i was editing it while you posted your

> answer, i think

>

> i'm not sure it's possible to edit your own post once

> someone else has answered behind :)

I had 'quoted' it and was in the process of answering it!

sabre150a at 2007-7-12 19:29:53 >

> > Have you measured the overhead when using regex in

> > this case?

>

> Can't be bothered - but the most efficient way regex

> could work on it would be to check the value range of

> each char, once. Which is what the "manual" solution

> does as well. Now considering that patterns need to

> be compiled before used, this compilation time can be

> considered overhead, although, admittedly, it will

> hardly matter if it's a one-time action on user

> input.

>

> Which solution would be the best one IMO strongly

> depends on the allowed value range and the further

> use of that String, which the OP never specified.

I have just compared the two approaches. On my average machine using JDK1.6 using test code

public static void testUsingRegex() throws Exception

{

java.util.regex.Matcher m = java.util.regex.Pattern.compile("\\d+").matcher("0");

String[] targets = {"1234","12sx"};

long start = 0;

for (int i = -1000; i < NUMBER_OF_ITERATIONS; i++)

{

if (i == 0)

start = System.currentTimeMillis();

for (String target : targets)

{

m.reset(target);

boolean valid = m.matches();

}

}

long end = System.currentTimeMillis();

double delta = 1000.0 * (end-start) / NUMBER_OF_ITERATIONS;

System.out.println("Regex " + delta + " us");

}

public static void testUsingCharArray() throws Exception

{

java.util.regex.Matcher m = java.util.regex.Pattern.compile("\\d+").matcher("0");

String[] targets = {"1234","12sx"};

long start = 0;

for (int i = -1000; i < NUMBER_OF_ITERATIONS; i++)

{

if (i == 0)

start = System.currentTimeMillis();

for (String target : targets)

{

boolean valid = true;

char[] chars = target.toCharArray();

for (char ch : chars)

{

if (!Character.isDigit(ch))

{

valid = false;

break;

}

}

}

}

long end = System.currentTimeMillis();

double delta = 1000.0 * (end-start) / NUMBER_OF_ITERATIONS;

System.out.println("Char array " + delta + " us");

}

I get as result

Regex 0.44248 us

Char array 0.25672 us

averaged over 100,000,000 iterations.

So yes, as expected, converting to a char array is faster but considering the actual values and the simplicity of the regex code I know which I prefer.

Message was edited by:

sabre150

sabre150a at 2007-7-12 19:29:53 >

'corlettk' - you need to show how you use the regex to get your timings because they do not stack up with my results.

The only way I can get a timing ratio comparable with yours is if I use boolean valid = target.matches("\\d+");

which is definitely not the way to use a regex when you want to compare performance.

Message was edited by:

sabre150

sabre150a at 2007-7-21 22:21:59 >