some questions

[Russ Tyndall]

On 7/18/06 7:37 AM, "Ruslan Zasukhin" <sunshine at public.kherson.ua> wrote:

> On 7/18/06 2:06 PM, "Anna Kritselis" <akritselis at earthlink.net> wrote:
> 
> Hi Anna,
> 
>> In my application, I give my users a place to enter student comments.  These
>> comments can be 3 sentences or 3-4 small paragraphs.  I keep going back and
>> forth between varchars and vText fields.  I had changed to vText when you
>> were
>> rewriting the varchar code a couple of months ago.  I did have db corruption
>> on the vText field and figured that you rewrote the varchar code, so I
>> switched back.  Then I see on the beta-list that vText fields are faster when
>> opening a database. I'm primarily concerned about corruption and still don't
>> feel like I know the best way to go.
> 
> I think is next:
> 
> * VarChar slow down OPEN only if you have e.g. 100 VarChars in db.
>   if you have 5-10, no significant affect.
> 

My "slow open" problem, discussed at length yesterday, went away when I
dropped a table that only had 2 varChar fields.  But even with those two
varChar fields, my db only had 6 varChar fields and only 1 was indexed.

*BUT* my db has 40+ String fields that are indexed.  So --- to work around
my problem -- should I concentrate on converting those varChar fields to
some other type, or spend time getting rid of indexes?

So far, trying to use String fields causes the db to balloon mightily in
size, but I've just barely started experimenting yet.

> * yesterday I have fix 1492 bug. Nasty.
>   could cause corruptions IF you delete some fields
>   then create new fields. New fields have reuse segments of old
>   fields without zeroing them
> 
>   report was exactly on TEXT fields. Although it could affect any.
>   May happens that it is reason of Dave report about 100K VarChar test.
> 
> * may happens that I will be able in 2.4 RELEASE also improve OPEN for
> VarChar...I have one old idea how to redo this. Idea is
>       
>   Keep ALL headers of all pagefiles (VarChar + Indexes)
>   in the single special file. So it must work much faster.
> 

Do you think (it may be too early to say) that we will keep compatibility
with 2.3 created dbs?

-- 
Russ Tyndall
Wake Forest, NC