>Hmmm... great idea. Yeah that might work although a little involved if you need to make this more generic for size.
>
>This works:
>
>

>lnSecs = SECONDS()
>
>lnCount = 0
>n = 0
>lnBuffer = 3500
>FOR n = 0 TO 9999999999
>	j = SUBSTR(m.lcString,(n*lnBuffer)+1,lnBuffer)
>	lnLen = LEN(j)
>	IF (lnLen < 1)
>	    EXIT
>	ENDIF
>	For lnX = 1 To lnLen
>		lcVal = Substr(m.j,lnX,1)
>		* lnCount = lnCount + 1  && to make sure it's getting hit the right amount of times!
>	Endfor
>ENDFOR
>
>* 0.32 seconds  
>? "String Splitting: " + TRANSFORM( SECONDS() - lnSecs )  + " " + TRANSFORM(lnCount)
>

>
>Also playing around with the buffer size - looks like on my machine ~3500 chars seems to work best. Over 5000 it starts slowing down. Under 2,000 it slows down too although not much on that end.
>
>Also interesting - see the `lnCount` in there? If I leave the counter in there times nearly double!
>
>Great thinking. I'll update the blog post (with credit to you)... Thank you!
>
>+++ Rick ---
>
>>

>>* 1mb string
>>Local lnX, lcString
>>lcString = Replicate("1234567890",100000)
>>
>>lnLength = Len(lcString)
>>? "Iterations: " + Transform(lnLength,"9,999,999")
>>
>>	lnSecs = Seconds()
>>
>>FOR n = 0 TO 999
>>	j = SUBSTR(m.lcString,(n*1000)+1,1000)
>>	For lnX = 1 To 1000
>>		lcVal = Substr(m.j,lnX,1)
>>	Endfor
>>endfor
>>
>>	? "SUBSTR: " +  Transform(Seconds() - lnSecs )
>>
>>

Yes.. for this test any exec code you run will have to execute a million times..!

In my experience with nfJson , using alines() to split the string by control
characters ( ,;[] " ) and then work in chunks was the key to speed.. I think
any big string manipulation should be tuned for each case..

Anyway, it's always good to share this little findings. !