Best Practices in Rev development

[J. Landman Gay]

Brian Yennie wrote:
 >
 > I think what you are discovering is that "it depends". I noticed a
 > couple of things with your example:
 >
 > 1) Your switch conditions are all expressions that need to be
 > reevaluated in every iteration (as opposed to constant value). This
 > effectively kills the switch statement's ability to "jump" to the
 > correct result - since it has to reevaluate the "jumps" every time
 > through. Try throwing in string constants and switch should start 
gaining.
 >
 > 2) 3-4 possible results is probably around the break-even point (in
 > general). But every extra condition will typically cause a gain on the
 > switch side.

Brian, you are absolutely right. I changed the test to string constants 
with four conditions and it was break-even. I knew there had to be 
something wrong with my test the results were too different.

on mouseUp
   put 100000 into tRepeats
   put the milliseconds into tstart
   switchTest tRepeats
   put the milliseconds - tstart into item 1 of tTimes
   put the milliseconds into tstart
   ifTest tRepeats
   put the milliseconds - tstart into item 2 of tTimes
   put tTimes
end mouseUp

on switchTest pRepeats
   put "a" into tRep
   repeat pRepeats
     switch tRep
     case tRep = "aaaaa"
       put tRep into x
       break
     case tRep = "aaaaaaaaaaa"
       put tRep into x
       break
     case tRep = "aaaaaaaaaaaaaaaaaaaaaaaaaaaaa"
       put tRep into x
       break
     default
       put tRep into x
     end switch
     put "a" after tRep
   end repeat
end switchTest

on ifTest pRepeats
   put "a" into tRep
   repeat pRepeats
     if tRep = "aaaaa" then
       put tRep into x
     else if tRep ="aaaaaaaaaaa" then
       put tRep into x
     else if tRep = "aaaaaaaaaaaaaaaaaaaaaaaaaaaaa" then
       put tRep into x
     else
       put tRep into x
     end if
     put "a" after tRep
   end repeat
end ifTest



> I think the basic lesson though is that there is no general answer. It 
> depends a lot on your data. For example, even the same if statement can 
> be written to be 10 times faster or slower depending on the data. 
> Suppose your first if statement passes 99% of the time, as opposed to 
> failing all tests 99% of the time.
> 
> My general thoughts:
> 
> * Use switch statements when you have 3 or more constant expressions (or 
> may need to add more in the future)
> * Use switch statements if you need the "fall through" logic (by 
> omitting break statements). Sometimes this leads to more elegant looking 
> code.
> * Use if statements when you have complex expressions and can predict to 
> some degree which conditions will pass/fail most of the time
> * If speed is not an issue, use whichever one makes the most sense to you
> 
>> Scott Kane wrote:
>>> ----- Original Message ----- From: "J. Landman Gay" 
>>> <jacque at hyperactivesw.com>
>>>> I read somewhere a long time ago that switch statements run faster 
>>>> than if/else. I think it was a general comment rather than 
>>>> specifically related to Revolution programming but it was so long 
>>>> ago I can't recall. Does anyone know if it is true for Rev scripts?
>>> I haven't timed it - but if it's read by the engine in the same 
>>> manner other programming languages handle it then it would be faster 
>>> as any subsequent conditions are skipped because once the case 
>>> statement (switch statement) is met the compiler knows not to proceed 
>>> - as opposed to if/then/else etc where the entire handler is 
>>> evaluated in full.
>>
>> I remember reading that Rev doesn't evaluate the entire if/else, so 
>> I'd think it would be about the same as switch. But lookee here, I did 
>> a test, with surprising results:
>>
>> on mouseUp
>>   put 100000 into tRepeats
>>   put the milliseconds into tstart
>>   switchTest tRepeats
>>   put the milliseconds - tstart into item 1 of tTimes
>>   put the milliseconds into tstart
>>   ifTest tRepeats
>>   put the milliseconds - tstart into item 2 of tTimes
>>   put tTimes
>> end mouseUp
>>
>> on switchTest pRepeats
>>   put 1 into tRep
>>   repeat pRepeats
>>     switch tRep
>>     case tRep > 10000
>>       put tRep into x
>>       break
>>     case tRep > 5000
>>       put tRep into x
>>       break
>>     case tRep > 2000
>>       put tRep into x
>>       break
>>     default
>>       put tRep into x
>>     end switch
>>     add 1 to tRep
>>   end repeat
>> end switchTest
>>
>> on ifTest pRepeats
>>   put 1 into tRep
>>   repeat pRepeats
>>     if tRep > 10000 then
>>       put tRep into x
>>     else if tRep > 5000 then
>>       put tRep into x
>>     else if tRep > 2000 then
>>       put tRep into x
>>     else
>>       put tRep into x
>>     end if
>>     add 1 to tRep
>>   end repeat
>> end ifTest
>>
>>
>> The "if/else" is 4 to 5 times faster. <blink> Is my test accurate?
>>
>> -- 
>> Jacqueline Landman Gay         |     jacque at hyperactivesw.com
>> HyperActive Software           |     http://www.hyperactivesw.com
>> _______________________________________________
>> use-revolution mailing list
>> use-revolution at lists.runrev.com
>> Please visit this url to subscribe, unsubscribe and manage your 
>> subscription preferences:
>> http://lists.runrev.com/mailman/listinfo/use-revolution
> 
> _______________________________________________
> use-revolution mailing list
> use-revolution at lists.runrev.com
> Please visit this url to subscribe, unsubscribe and manage your 
> subscription preferences:
> http://lists.runrev.com/mailman/listinfo/use-revolution
> 


-- 
Jacqueline Landman Gay         |     jacque at hyperactivesw.com
HyperActive Software           |     http://www.hyperactivesw.com