Re: general design issue
From: Bob Badour <bbadour_at_pei.sympatico.ca>
Date: Thu, 23 Apr 2009 16:54:29 -0300
Message-ID: <49f0c777$0$5493$9a566e8b_at_news.aliant.net>
>>Btw, Sqlite did use an index for 'select a from text where a > 'a190'
>>and a < 'a195'. Lots has been written about algebraic optimal
>>conversions of relational operators but very little about formal
>>conversions of type manipulations like this and many other situations.
>>Too bad, this would have been more worthwhile for the SQL committee to
>>have spent their time on than arguing about keywords and trying to make
>>their informal definiitions precise. .It's pretty bad when a single
>>implementation can't even make consistent optimizations for a
>>single-table query. No slur against sqlite per se, there are things I
>>like about it. I'll bet all the popular sql implementations are similar
>>when it comes to optimizations/separating logical from physical.
Date: Thu, 23 Apr 2009 16:54:29 -0300
Message-ID: <49f0c777$0$5493$9a566e8b_at_news.aliant.net>
skink wrote:
> On 23 Kwi, 17:13, paul c wrote: >
>>Btw, Sqlite did use an index for 'select a from text where a > 'a190'
>>and a < 'a195'. Lots has been written about algebraic optimal
>>conversions of relational operators but very little about formal
>>conversions of type manipulations like this and many other situations.
>>Too bad, this would have been more worthwhile for the SQL committee to
>>have spent their time on than arguing about keywords and trying to make
>>their informal definiitions precise. .It's pretty bad when a single
>>implementation can't even make consistent optimizations for a
>>single-table query. No slur against sqlite per se, there are things I
>>like about it. I'll bet all the popular sql implementations are similar
>>when it comes to optimizations/separating logical from physical.
> > well, if i cannot use substr/like with sqlite > because of bad performance what about my original solution with second > table keeping prefixes and integer with one bit per letter. > > how to make it fast and small? is it feasible at all?
Other than being non-standard, what is your objection to using GLOB 'prefix*' ? Received on Thu Apr 23 2009 - 21:54:29 CEST