Insertion to arbitrary place of long sorted chains are very slow in Image,
because the system starts to search the insertion point at the end of the
sorted chain. IN our system we have a sorted chain in the biggest detail
data set, with sometimes 10000 entries/key value. This layout is optimized
for quick DBPUT of new entries: In normal case the new entries are
appended to the end of the chain. But (when we have to change an other
search item in some entries [because to customers are merged and got the
same code]) if we DBDELETE and DBPUT entries in the random place of the
chain, the process is time consuming.

Before we do such slow transaction,  we check the chain length
with QUERY. During the transaction the program reports the progress
done/remained and a time estimate.

                                             Frank

  |\  /~ ~~|~~~ Family : NAGY; first name : FERENC; title : Ph. D.
  | \ |   -+-   Institute of Isotopes of the Hungarian Academy of Sciences
  |  \|    |    1121 BUDAPEST Konkoly-Thege M. ut 29, HUNGARY (1525 Bp. POB 77)
`-'   '  `-'    E-mail : [log in to unmask]

Fax: (36)-1-395-9002, work phone: (36)-1-395-9123, home phone: (36)-1-277-4229.
            ^NEW!                        ^NEW!
Home address: H-1214 BUDAPEST Raketa u. 29. I. 3.

There are 3 kinds of programming errors:  syntactical, semantical and mystical.
The programmers have to suck up the users just as much as absolutely necessary.