At 11:17 PM 27/4/95 -0400, Denys Beauchemin wrote:
>I respectfully disagree that your solution is efficient.  The other solutions
>provided for this problem all involved re-reading the master entry location
>after the DBDELETE.  If a migrating secondary situation is involved, the
>secondary chain will be deleted in its entirety at that location.  Secondary
>chains are usually not very long unless you are dealing with an Integer key
>type and you have sampled the data to get a capacity instead of calculating
>the exact capacity which will yield no secondaries.
>
>However, in the re-read technique, you only go through the dataset once,
>catching everything.  You advocate going through the dataset twice.  I find
>this to be inefficient.
 
I think Gilles was correct...almost.  In fact I read misread it first time
and assumed the best solution was there, so didn't consider it further.
 
I thought the solution was:
----------------------------------------------------------------------
dbget(mode 2)            << get first >>
while (status.element(1) <> 0) do
  begin
  while (status.element(5-6) <> 0) do
        begin           << synonyms exist >>
        dbdelete(..)    << delete primary >>
        dbget(mode 1)   << reread new primary >>
        end
  dbdelete(..)          << synonym or primary with no synonyms >>
  dbget(mode 2)         << get next >>
  end
if status.element(1) <> 11 then error(..)
-----------------------------------------------------------------------
The only possible improvement this offers over some of the previous
solutions is that it avoids a dbget(mode 1) when it is known it would fail.
 
----
Jim "seMPEr" Wowchuk           Internet:    [log in to unmask]
Vanguard Computer Services     Compu$erve:  100036,106
 _--_|\                        Post:        PO Box 18, North Ryde, NSW 2113
/      \                       Phone:       +61 (2) 888-9688
\.--.__/ <---Sydney NSW        Fax:         +61 (2) 888-3056
      v      Australia