Re: processing character entities

[Brandon Ibach <bibach@xxxxxxxxxxxxxx>]

Quoting Chris Maden <crism@xxxxxxxxxxx>:
> This is what I do.  It's really not that bad, just verbose.
> 
> (define (process-text #!optional (snl (current-node)))
>   ;; this part is inefficient; I need to rewrite this to carry a node
>   ;; index instead of an actual list of nodes
>   (let p-t-loop ((this-node (node-list-first (children snl)))
> 		   (other-nodes (node-list-rest (children snl))))
>        (if (node-list-empty? this-node)
> 	   (empty-sosofo)
> 	   (sosofo-append (case (node-property 'class-name
> 					       this-node)
> 			    ;; case clauses edited for brevity -BI
> 			  )
> 			  (if (node-list-empty? other-nodes)
> 			      (empty-sosofo)
> 			      (p-t-loop (node-list-first other-nodes)
> 					(node-list-rest other-nodes)))))))
> 
   In the interest of good practice (and possibly performance), you
may want to try this variation:

(define (process-text #!optional (snl (current-node)))
  (let p-t-loop ((this-node (node-list-first (children snl)))
		 (other-nodes (node-list-rest (children snl)))
		 (result (empty-sosofo)))
       (if (node-list-empty? this-node)
	   result
	   (p-t-loop (node-list-first other-nodes)
		     (node-list-rest other-nodes)
		     (sosofo-append result
				    (case (node-property 'class-name
							 this-node)
			    ;; case clauses edited for brevity -BI
				    ))))))

   This not only makes it tail-recursive, but also eliminates the
double check (at start and end) for the empty list.
   On this note, a question for the more Scheme-knowledgeable among us.
We know that a non-tail-recursive function will take up more space in
memory as the stack grows with each call, but what is the performance
impact?  Does the fix I've supplied above really have the potential to
make a significant impact?  If it saves even a little time on each
iteration of the loop, I'd bet it could make a significant difference
in the overall running time of Chris' application, given that this
routine probably processes just about every data character and sdata
entity in his document.

-Brandon :)


 DSSSList info and archive:  http://www.mulberrytech.com/dsssl/dssslist