>I'm working on a project where I need to shred a string (ala shredding and XML string) into a set of tables. The string is passed to me by a call to an *.OCX. The typical length of the string is 43K+ bytes. The string structure looks like this:
>
>Data-Pair /pair delimiter/ Data-Pair /pair delimiter/ etc.
>There are additional delimiters inside each data-pair. Essentially a data-pair consists of a delimited string that defines the schema and also separates the schema from the actual value. A Data-Pair ends up looking something like this:
>Grand-ParentField/Delimiter/ParentField/Delimiter/Field/Value Delimiter/Value.
>
>My "parsing" engine works, but it's very slow. (About 2-3 seconds per string on a PII 300 Mhz notebook.) I need to get down to sub-second times without simply buying a faster notebook. I'm working on enhancements to make the string shorter. But I'm also looking for a way to more efficiently pass through the string as I shred it. That's what this post is for. So far the most efficient method of parsing the string is this:
>
>Do While .T.
>DataPair = SubString( LongString, 1, At( , LongString )- 1 )
>* Parsing code goes here
>LongString = SubString( LongString, At( , LongString ) + 1 )
>If Len( LongString ) = 0
> Exit
>EndIf
>EndDo
>
>I've tried NextWord from foxtools, ( I'm running VFP 6 ) but it's actually slower than this approach is.
>
>Ideas anyone on a better way to work my way through the string?



*------------------------------------------------------------------
i haven't actually tried this solution, but intuition tells me
that the string copy you do (LongString = SUBSTR(LongString...)
is the bottleneck. i'm guessing that behind the scene, a character
by character move is occuring.

it would really be nice if AT had a fourth parameter that let
you specify the starting position from which to begin searching,
but it doesn't.

instead of copying the remaining substring into LongString and
getting your new piece by using SUBSTR with a beginning index
of 1 ...

keep the original string intact and keep track of a front pointer
and a locater pointer, thus eliminating the copy operation on 43Kb.

Total_Length = LEN(ALLTRIM(Long_String)
Front_Pointer = 1
Locater_Pointer = AT(,Long_String
WHILE ( Locater_Pointer > 0 )
This_Piece = SUBSTR ( Long_String , Front_Pointer , Locater_Pointer )

Front_Pointer = Locater_Pointer + 1
Locater_Pointer = AT(,Long_String
ENDDO


interesting... think i'll put a timer around this and try it both
ways.

good luck.

patrick