IMAGE is the primary reason that the
HP 3000
was
a success. It is a reliable, simple, fast database for
MPE.
After about a decade of enhancements it was renamed
TurboIMAGE, then TurboIMAGE/XL after the PA-RISC
migration.
Programs can access and update records using
IMAGE intrinsics.
Today it is known as IMAGE/SQL
because it includes an
SQL
(Structured Query Language) frontend and
client-server support for PC accessors.
Database Design
Use masters for unique data items like customer numbers.
Use details for repeated data like customer purchases.
If a master record needs an alternate key, make it a detail and use an automatic master.
If the information in a detail dataset is volatile,
avoid more than two paths into
it, and avoid sorted paths,
except where entries are put in chain order.
Remember that IMAGE now allows any data-type as
a sort field (sort fields used to be limited to unsigned data-types).
If the information is relatively static and frequently
accessed by on-line users,
add more search paths into it.
Get
Suprtool
for fast serial scans and eliminate infrequently used search keys.
Polishing Database Design
For numeric fields, use J1 for fewer than five digits; use J2 for fewer than ten digits; otherwise, use a P-field (packed-decimal).
Types X, U, P, and Z give the best hashing results, especially if the key length is greater than six bytes.
Avoid keys of type I, J, K, and R.
Store dates in yyyymmdd format, either as a J2 (32-bit integer) or as an Z8 (do not leave out the century).
Assign a primary path to every detail dataset; select the most frequently used path with more than one entry per chain.
Try to keep master datasets less than 80 percent full.
Never use a power of 2 as a master capacity (for example, 32, 1024, and so on).
Avoid tricky data structures that you cannot use in all your tools.