virtuals fields / ViSQL

[Robert Brenstein]

>On 1/14/05 5:08 PM, "Olivier" <vidal_olivier at yahoo.fr> wrote:
>
>>  I agree.
>>
>>  olivier
>>
>>  Le 14 janv. 05, à 15:39, Robert Brenstein a écrit :
>>
>>>  But having index-by-words set but being not in effect IS
>>>  confusing/misleading.
>>>
>>>  This is why I am suggesting that it becomes an independent flag: one
>>>  flag (index) controls normal indexing and another flag
>>>  (index-by-words) control word indexing. When both are set, the latter
>>>  (higher functionality) wins.
>>>
>>>  I think this would make things clearer for developers.
>
>Think about this behavior.
>
>I have field (Indexed + ByWords)
>
>Now I want set index OFF for some time, then return it back.
>
>So I do
>
>         fld.Indexed = false.
>
>         ...
>
>         fld.Indexed = true.
>
>Now in 2.0, the first line DO NOT remove flag IndexByWords (1.x did).
>So the second line is able __correctly__ return back THE SAME index kind.

Then in 2, these lines would tackle both flags as 
needed. At low level,I mean in kernel, you should 
probably have a separate flag that tells kernel 
whether to index or not, so kernel does not have 
to touch what users set.

>Even do not have idea how todo this in 1.x.
>* You need self get and save flag IndexByWords ?
>* and flag Unique?
>* and for each field if you do this for several fields.

V1.x should stay as it is in my opinion. There is 
a discontinuity from going from 1 to 2 anyway, so 
I see no problem. You should really use this 
opportunity to clean up and improve all APIs.

Robert