Re: J4 - presentation/discussion on "Future of the COBOL Standard"

On Mar 18, 5:53 pm, Robert <n...@[EMAIL PROTECTED]
> wrote:
> On Mon, 17 Mar 2008 20:09:47 -0700 (PDT), Richard <rip...@[EMAIL PROTECTED]
>
wrote:
> >On Mar 18, 4:53 pm, Robert <n...@[EMAIL PROTECTED]
> wrote:
> >> On Tue, 18 Mar 2008 11:33:00 +1300, "Pete Dashwood"
<dashw...@[EMAIL PROTECTED]
>
> >> wrote:
>
> >> >"Robert" <n...@[EMAIL PROTECTED]
> wrote in message
> >> >news:6dmrt3dn0hueihi3p996d5n96sts0pgtm1@[EMAIL PROTECTED]
> >> >> Here's a simple rule: every table that's SEARCHed should be
defined with
> >> >> ODO.
>
> >> >No. You MIGHT formulate a rule where every table that's SEARCHed ALL
should
> >> >be defined with ODO.
>
> >> If you SEARCH serially and the subject is not in the table, there's
no point in searching
> >> 100 high values at end end. You want to stop after the last entry
loaded.
>
> >No. You should stop after finding one higher than what you are
> >searching for (given the entries in order).
>
> If the table is in key order, you wouldn't be doing a serial search.

The table may be in sequence but not have unique keys. It may for
example be a list of processing to do where some get done with
multiple entries.


> There is a point
> below which serial search is faster. I would hope SEARCH ALL is smart
enough to switch to
> serial at that point.
>
> -----
>
> I just tested it with Fujitsu on Wintel. The crossover point is 7
entries. ALL does NOT
> automatically switch below that point. At 3 entries, ALL is 13% slower.
At 15, ALL is 25%
> faster.