Cache Class
The Cache class
contains methods used to retrieve Record<class> objects from within
a process’ local memory, calling the Db.getRecord
method itself if a record does not exist in the cache. The intent of cache
is to provide database optimization for references to records that either
are not expected to change during the life of the cache and/or process,
or where database updates can randomly be visible or not within the cache
without affecting the operation of the program. Only use the Cache class
in processes that contain read-only references to frequently accessed
tables that exhibit repeated references to the same records (e.g., CUVAR,
TRN, PRODCTL, PRODFT*).
The Cache class does not use the vobj() array for data storage. This
limits the implementation to force a consistent identifier structure for
all procedures that reference the cache. That is, the cache identifier
(referencing the same cache) cannot be a simple variable in one process
and an array in another process. This implementation enhances performance.
Using a Cache object and the Cache.getRecord
method does not have a significant performance impact in an M database
implementation. Benchmarks produce either negligible improvements or degradation
(+/- 10%), depending on the size of the cache and the structure of the
identifier (i.e., array identifiers are slower than named variables, large
caches are slower than small ones). However, significant performance improvements
are expected on an ORACLE database. Separate caches per table seem to
be more efficient (and easier to manage) than one large array of caches
with the table or process name as the subscript key.
Implement the Cache class in transaction processing and long-running
batches, such as accruals. Cache the Institution Variables table (CUVAR)
at the driver level (call the cache object %CACHE), making it available
for all processes. Use the NEW command and KILL command on the identifier
to manage the scope of a cache object.
Example
The following code provides an example of how you can cache the CUVAR
and TRN tables at the driver level in a cache object called %CACHE.
TAG(String ETC) //
type public
Cache %CUVAR
type Public
Cache %CACHE()
type RecordCUVAR cuvar = %CUVAR.getRecord("CUVAR")
type RecordTRN trn = %CACHE("TRN").getRecord("TRN",":ETC")
set tjd = cuvar.tjd
set conam = cuvar.conam
set x = trn.etc
set x = trn.pgm
quit
|
|
Although the CUVAR table should be cached, evaluate source
code for optimization through the #IF and #ELSE compiler commands in combination with compile
time CUVAR lookups. This optimization can substantially reduce the amount
of database IO and code that executes at run-time. |
|
|
Cache is not under transaction control and may, in fact,
not represent the current state of the database or any known prior state.
The state of any RecordClass object within the cache is as of the instant
it was loaded into the cache. |
This class is available in Profile v7.0
and greater. There are future plans to make this class available
in Profile v6.4 as well.
|
Methods |
Description |
|
getRecord |
Retrieves a single record from within a process’ local memory. |