Problem with word boundaries in Valentina 2

[Erne]

On 13-nov-06, at 19:47, Ruslan Zasukhin wrote:

> On 11/13/06 6:02 PM, "Martin S." <blackfin at elfenherz.de> wrote:
>
> Hi Martin,
>
>> Hello,
>>
>> In Valentina 1.xx, I used to do a case-insensitive search for a word,
>> that e.g. begins with "bo" in that way:
>>
>> "SELECT * FROM xxx WHERE yyy LIKE '\<bo' no_case"
>>
>> I've read trough your documentation for the new Valentina 2 SQL  
>> Syntax
>> and found that no_case is deprecated and that I have to use RegEx.
>> No problem I thought, so I've further read through the ICU  
>> documentation
>> to build a Regular Expression that matches my needs.
>>
>> The problem that I have with V2 are word boundaries. I want the  
>> user to
>> type an expression which will be searched as a BEGIN of a word and  
>> only
>> WITHIN a word.
>> e.g.: User types "bo", Valentina should find words like "Book",
>> "Borrow", but not "unborn" or "rebound".
>>
>> I tried to use a ICU RegEx which contains word boundary matching:
>>
>> "SELECT * FROM xxx WHERE yyy REGEX '\b(?i)bo(.*)'"
>>
>> ...but sadly that does not work. The cursor's record count is  
>> always zero.
>> When I leave off "\b", then I get results, but it finds every text  
>> that
>> CONTAINS the typed string, and not only words that begin with it.
>> With \b on, there are never any results.
>>
>> I also tried the other way and set the collation atrribute  
>> #kstrength of
>> my database to #kprimary, sent RegEx to hell and tried it with  
>> good old
>> "LIKE" statements.
>> Unfortunately, when I use the wildcard "%", the results are the  
>> same as
>> with RegEx. (but a lot faster...).
>> As I'm rather sure that I won't be able to achieve the right results
>> with the use of a LIKE-statement, I'm switching back to RegEx now.
>>
>> But.....can you explain what I have to do with it to achieve my  
>> wanted
>> results?
>> Are ICU word boundaries not yet implemented in V2 or did I  
>> something wrong?
>
> In short -- the must work.
>
> Note, that you can make db strength to be primary,
> And use then simpler REGEX
>
>      fld REGEX '\bbo(.*)'"
>
> \b really is START OF WORD ?
>

maybe the search string needs to be escaped?


Cool Runnings,
Erne.

/ |\ |\  | /
– |/ | \ | –
\ |\ |  \| \