Re: text compare takes very long...

Tech-Archive recommends: Repair Windows Errors & Optimize Windows Performance

"Co" <vonclausowitz@xxxxxxxxx> wrote in message news:1192270828.787010.50550@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

Mike, Great stuff, just one remark. Could you add
another boolean so the user can choose if he wants
to run it case sensitive or not?

I've been wondering how long it would be before someone asked that question ;-) It is of course not quite so simple as "adding another Boolean", as I'm sure you are aware, but it can certainly be done. In the original code (the code that does not include the SAFEARRAY stuff) you could very easily make the comparison case insensitive by simply changing the following:

b1 = StrConv(String1, vbFromUnicode)
b2 = StrConv(String2, vbFromUnicode)

to:

b1 = StrConv(LCase(String1), vbFromUnicode)
b2 = StrConv(LCase(String2), vbFromUnicode)

However, in the modified faster code it is not quite so easy. The simplest way would be to perform an OR of the data words with the value 32, but before doing that it would be wise to investigate which is the best place to actually do that. In the earlier versions (where the loops in the main algorithm were equal to the product of the two string lengths) it would be best to do it two the two Integer arrays themselves, making sure that you revert the contents of those arrays back to their original condition at the end of the routine to avoid altering the String data in the calling procedure (because the original strings will respond to any changes you make to the Integer data). However, in the latest and fastest routine (where the loops in the new algorithm are much less than the product of the two string lengths) it might be best to instead do it in the algorithm itself. I think before I wrote any amendments I might be inclined to place a loop counter inside the new algorithm and test it on various different input strings to see how many loops occur and how many ORs are required in each loop, and then make the decision regarding the placement of the amendment on the basis of whether or not the number of required OR's in the main algorithm (on average) exceeded twice the sum of the two string lengths.

I haven't actually got time to look into any of that stuff at the moment (Saturday and the weekend and all that, and of course the all important England versus France rugby on TV here in the UK tonight) but perhaps others here might have a go at it. It would certainly be interesting to investigate both methods I have suggested, and perhaps other alternative methods, so that we end up with a flexible and yet still fast routine. It's quite interesting stuff, this :-)

Mike


.

Relevant Pages

  • Re: text compare takes very long...
    ... b1 = StrConv(String1, vbFromUnicode) ... In the earlier versions (where the loops in the main ... algorithm were equal to the product of the two string lengths) it would be ... loops in the new algorithm are much less than the product of the two string ...
    (microsoft.public.vb.general.discussion)
  • Re: A Symbol- Abstracted Al
    ... What did the algorithm you used look like. ... Except the loops must be optimized ... I was making a DES cracker example and had Visual C6.0 default to ... The basic code is the correct and functional code, ...
    (sci.crypt)
  • Re: Spatial Optimization Algorithm Help
    ... > Are you using a formal pseudocode or your own notation? ... > which you correctly addressed in your algorithm). ... subset" loops, which enumerate combinations. ... an algorithm with your favourite search engine (search for something like ...
    (comp.programming)
  • Re: APL Efficiency (Project Euler #12)
    ... and especially ones that do not need loops that are fast an effecient. ... place to obtain the same information faster in APL? ... number of divisors algorithm. ...
    (comp.lang.apl)
  • APL Efficiency (Project Euler #12)
    ... I am feeling more comfortable in APL now, and most things are beginning to make sense, though some of the best practices still elude me. ... Ignoring the issue of composition at the moment, which I will address in another message if I ever formulate things well enough, I have found that it is easy to write both inefficient looping code and very memory intensive/inefficient code without loops. ... At first, I could not do a brute force algorithm because of space, so I settled for a loop. ... This basically did a naive prime factorization and number of divisors algorithm. ...
    (comp.lang.apl)