This is a question related specifically to D3....

I have a large file, and desire to index the 5% of the items in it that have
a value in a particular field. The other 95% of the records with the field
being empty do not need to be indexed.

Universe has a "no nulls" option, used when creating an index. Nothing in
the RD Online Technical Manual suggests that they have anything equivilant.

Before I head off and index all the records in the file, I thought I'd ask
if any of you are aware of some way to build a "no nulls" index in D3?

Thanks.

No can do, Brian. Note that under some circumstances (D3/NT I think)
there is a limit on the maximum number of items in a single index
value (in this case, null). I think that max is 64K. I would check with
RD support.

Why don't you post D3 stuff to CDP or the RD forums? (Disregarding the
fact that I answered your question.)

/Scott Ballinger
Pareto Corporation
Edmonds WA USA
206 713 6006

Ok, skip the CDP part and what I meant was the RD forum part.... didn't
mean to be so snarky.

/Scott

This group is more fun.


"Scott Ballinger" <scott.ballinger (AT) gmail (DOT) com> wrote

Scott,

Please see my question below: You may ignore if you desire!

"Scott Ballinger" <scott.ballinger (AT) gmail (DOT) com> wrote

When you ask "Why don't you post D3 stuff to CDP...?", is there a CDP other
than this one???

Larry Hazel

Darn,

Should have read the entire thread! 8>(

By the way, I who have several clients for whom I have purchased D3
from Raining Data could not get access to post questions on the
Raining Data forums.

Larry

"Scott Ballinger" <scott.ballinger (AT) gmail (DOT) com> wrote

Simple matter to register for the forums

As Scott said, no way around "nulls" issue ... and you may want to
concat the key to the index to make more unique.

Did I say there was "no way around" the previous problem? OK, perhaps
there are if you wanted to do a little bit of work.

One technique would be to have a trigger in your main file that simply
maintained a secondary file - @ID = whatever you are indexing (and you
would add logic to exclude nulls), with @RECORD = @ID of main file ...
other variations to the theme in terms of structure, but you could now
use the index on the associated file to build a select list that could
then be used against the primary ... this is possibly neatest solution

As previously mentioned, would be to amend index from, say create-index
A16, to create-index A16:"*":A0 --> not necessarily useful with a
SELECT, but if you were accessing using a BASIC KEY routine this would
be fine.

There are other alternatives, like using an alternate BTree product
(like our SuperB [which we haven't sold for many a long year]) that
would allow you to exclude nulls (or any other repeated value), but I
wouldn't see these as viable

HTH

Ross Ferris
Stamina Software
Visage > Better by Design!

Ross:

Aren't "kludges" great? :-)

Bill

"Ross Ferris" <rossf (AT) stamina (DOT) com.au> wrote

It just seems really odd that a D3 index cannot actually be used in cases
where an index might come in handy.

'nuff said.

You have all the indexes you need and can go ahead and use them can't
you? I thought the only issue here was that there will be a "spare"
index. Did I miss something?

Mike.