Re: [JBoss-dev] so how fast is string comparison
I'm slightly of topic, anyway IMHO, Strings are dead slow when comparing to ints. This is not because of string math (i.e. comparison, etc.) but because of _new_ and _new_ Strings which takes part in many methods with Strings involved (and almost _every_ String finishes in GC) So I look at jdk1.4 as a (someway) solver with its nio.* and regex.* stuff LT Not too bad if the compared strings don't have the same length. Anyway I think switch will be always better than if...else. IMHO I use to work only with int and translate them to String only for printing and for input. I assume a lot as String manipulation is just dead slow, but I could be wrong, I really don't know. marcf ___ Jboss-development mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/jboss-development
RE: [JBoss-dev] so how fast is string comparison
Most of the strings we are talking about are dynamic (or at least run-time configurable), so we would have to do string.intern() to be able to use == and I don't think that you'd want that internal list of strings growing like that. Besides, when you have a new string and do an intern(), it looks up the common string in a hashtable using hashcode to return the string on which you can do == on. Michael -Original Message- From: Adam Heath [mailto:[EMAIL PROTECTED]] Sent: Thursday, February 07, 2002 1:17 PM To: marc fleury Cc: Jboss-Development@Lists. Sourceforge. Net Subject: Re: [JBoss-dev] so how fast is string comparison On Thu, 7 Feb 2002, marc fleury wrote: when we do if (string.equals(this)) else if (string.equals(that)) how much slower is it than switch (stringAsInt) case THIS: case THAT: If strings are final, then you can use ==, instead of the much slower .equals(). A switch stmt is really nothing more than a series of if/elseif blocks. So, making all strings static/final, and using ==, should make for a win, without making the code hard to read by using abstract ints. ___ Jboss-development mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/jboss-development ___ Jboss-development mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/jboss-development
RE: [JBoss-dev] so how fast is string comparison
This technique was also proposed in the JDC tech tip of this month, and has the caveat that a serialized/deserialized String would not work with ==, even if it was assigned from a static value. OTOH, the final code translated for a switch depend on the number of 'case' values. If they're many the compiler should use some kind of binary search (as a quick sort, but for finding) previously ordering the values. That is why they are required to be primitive types. When the options are a few, they're searched lineally to improve performance. Or it was so the last time I poke around that. -Mensaje original- De: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]En nombre de [EMAIL PROTECTED] Enviado el: jueves, 07 de febrero de 2002 18:44 Para: [EMAIL PROTECTED] Asunto: RE: [JBoss-dev] so how fast is string comparison Most of the strings we are talking about are dynamic (or at least run-time configurable), so we would have to do string.intern() to be able to use == and I don't think that you'd want that internal list of strings growing like that. Besides, when you have a new string and do an intern(), it looks up the common string in a hashtable using hashcode to return the string on which you can do == on. Michael -Original Message- From: Adam Heath [mailto:[EMAIL PROTECTED]] Sent: Thursday, February 07, 2002 1:17 PM To: marc fleury Cc: Jboss-Development@Lists. Sourceforge. Net Subject: Re: [JBoss-dev] so how fast is string comparison On Thu, 7 Feb 2002, marc fleury wrote: when we do if (string.equals(this)) else if (string.equals(that)) how much slower is it than switch (stringAsInt) case THIS: case THAT: If strings are final, then you can use ==, instead of the much slower .equals(). A switch stmt is really nothing more than a series of if/elseif blocks. So, making all strings static/final, and using ==, should make for a win, without making the code hard to read by using abstract ints. ___ Jboss-development mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/jboss-development ___ Jboss-development mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/jboss-development ___ Jboss-development mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/jboss-development
