Subject: | Re: Like
| Date: | Fri, 22 Dec 2017 12:05:19 +0000
| From: | Michael Kennedy <Info@KennedySoftware.ie>
| Newsgroups: | pnews.paradox-dos
|
On 22/12/2017 00:38, Steven Green wrote:
> don't recall all of the testing that was done, but it was generally
> concluded that it was skewed towards the first 5 characters of the
> string, and was less predictable than most dot-dot options.. don't know
> how you'd want to reproduce it, if you can't predict it
AFAIR, I NEVER used "like", because I never knew how it worked.
But, with that 5-char skew that Steve remembered, perhaps it was an
implementation of (standard?) "Soundex"? If so, it would apply only to
alphabetic (maybe alphanumeric?) strings:
- always do an exact match on the very first char,
- then match on the (english?) "sound" of the next 3/4/5/6/.. chars.
Eg:
If QBE has: Like paradox, this would translate to a Soundex code of P632
Then, DB records with Paradox, Poridex, Piridicks, etc (which also
translate to P632) would be matches.
(I've not run the above example).
- Mike
> OK Steve, I'll take the bait!
>
> What is known of the traps/bugs/features/poo in 'LIKE'?
>
> Anyone wanna do some tests?
>
> I'll be quiet for a while trying to sensibly parse what we have so far,
> but will be lurking ...
>
> - Bernie
>
> ---
> This email has been checked for viruses by Avast antivirus software.
> https://www.avast.com/antivirus
>
|