This overview complements the introduction in the API documentation.
A Key consists of a name, a value and metadata.
It is the atomic unit in the key database. Its main purpose is that it
can be serialized to be written out to permanent storage. It can be
added to several aggregates using reference counting.
Putting Key objects into other data structures of supported
programming languages presents no problem.
The central data structure in Elektra is a KeySet. It aggregates Key
objects in order to describe configuration in an easy but complete way.
As the name "set" already implies every Key in a KeySet has a
unique name. A user can iterate over the Key objects of a KeySet.
KeySet sorts the keys by their names. This yields a deterministic
order advantage. So, independent of the appending sequence and, in
particular, the number of fetches and updates, KeySet guarantees the
same order of the Key objects. Some configuration storage systems
need this property, because they cannot remember a specific order.
On the other hand, any particular order can easily be introduced
(See order).
On the one side backends generate or store a KeySet object and, on the
other side, elektrified applications receive and send a KeySet object.
Both sides, as well as the core in between, have the possibility to
iterate, update, modify, extend and reduce the key set. Appending of
new or existing Key objects extends the key set. Otherwise it can be
reduced if keys are popped out. The Key object becomes independent of
the KeySet afterwards. The user can still change such a key or append
it into another key set. The affiliation to a key set is not exclusive.
Every key in a KeySet object has a unique name. Appending Key objects with
the same name will override the already existing Key object.
While objects of Key and KeySet only reside in memory,
Elektra’s third class
KDB actually provides access to the global key database. KDB,
an abbreviation of key database, is responsible for
actually storing and receiving configuration. KeySet represents the
configuration when communicating with KDB.
The typical elektrified application collects its configuration by one or
many calls of kdbGet().
As soon as the program finishes its work with the
KeySet,
kdbSet() is in charge of writing all changes back to the key
database.
This technique has some advantages. First, applications have full
control over modifying
Key and KeySet objects without touching the key database.
Second,
the decision how many KeySet objects the application
administrates is left to the application.
It can choose how to split up the KeySet objects.
The main reason for this technique is that for backend development the
same data structure is used, and
as we will see, the borderline between application
and backend development becomes blurred.
The application adapts the configuration between kdbGet()
and kdbSet() in memory.
The modifications are not only
faster, they also allow large atomic
configuration upgrades, robust merging of settings and handling of
complicated inter-relationships between keys without problematic
intermediate steps.
Elektrified applications, however, should be aware of conflicts.
It can happen that the key database is
changed while working with a KeySet.
Then, attempts to use kdbSet() lead to a conflict.
KDB detects such situations gracefully and lets the application decide
which configuration should be used.
For details and background read more about elektra data structures. For further information see the API documentation.