Trying to code the fastest algorithm...

[J. Landman Gay]

On 6/5/05 4:14 AM, jbv wrote:

> Hi list,
> 
> I'm trying to build the fastest possible algorithm for the
> following task :
> I have a variable containing reference words; each "word"
> in this variable is considered as an item.
> I also have a list of "sentences" (each sentence being a
> list of "words" separated by spaces), 1 per line. Each
> sentence can contain 1 or more words from the reference,
> as well as other words.
> 
> I need to determine for each sentence, which words from
> the reference are included in it, in which order, and output
> a list of offsets.
> For example :
>     reference variable :  W1,W2,W3,W4
>     sentence 1 : Wx W2 Wy Wz W3
>     output : 2,3
> 
> And last but not least, I need to keep only sentences that
> contain more than 1 word from the reference.

My stab at it:

function calcwords tReference,tSentence
   --  tReference should be the comma-delimited word list, i.e.: "w1,w2,w3"
   --  tSentence is the user's space-delimited entry, i.e.: "wx w2 wy wz w3"
   put tReference into tRef -- so we can manipulate a copy
   replace comma with comma & cr in tRef
   split tRef by cr and comma
   replace space with comma & cr in tSentence
   split tSentence by cr and comma
   intersect tSentence with tRef -- it now has the right keys

   repeat for each line l in keys(tSentence)
     put itemoffset(l,tReference) & comma after tOutput
   end repeat
   delete last char of tOutput -- the comma
   if comma is in tOutput then return tOutput
   else return empty
end calcwords

I didn't time it but it seems like it should be faster.

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