Elektra  0.9.3
Plugin: spec

The spec plugin is a global plugin that copies metadata from the spec-namespace to other namespaces using their key names as globbing expressions. Globbing resembles regular expressions. They do not have the same expressive power, but are easier to use. The semantics are more suitable to match path names:

The plugin copies the metadata of the corresponding spec key to every matching (see below) key in the other namespaces. The copied metadata is removed again during kdbSet (if remained unchanged).

The spec plugin also provides basic validation and structural checking. Specifically it supports:

The matching of the spec (globbing) keys to the keys in the other namespaces is based on elektraKeyGlob(), which in turn is based on the well known fnmatch(3). However, there is special handling for array specifications (#) and wildcard specifications (_).

Keys which contain a part that is exactly # (e.g. my/#/key or my/#) are called array specifications. Instead of just matching the spec key to any existing keys in other namespaces, theses keys are "instantiated". This means we lookup the array size (defined by the array metakey) using a cascading ksLookup. This only looks at non-spec namespaces, if we don't find an array size their, we check the array parent in the spec namespace. If we still have no array size, the array is deemed to be empty. For empty arrays, we will simply validate that they are indeed empty.

For non-empty arrays "instantiation" takes place. This means that we replace each # part in the spec key by all the valid array elements for this specification, up to the array size (which was already checked to be less than the allowed maximum). In other words, instead of actually matching the globbing expression, we create a separate but identical specification for all valid array elements.

The purpose of "instantiation" is to allow specifying default values for array elements.

Keys which contain a part that is exactly _ (e.g. my/_/key or my/_) are called wildcard specifications. It would be nice to have an "instantiation" procedure for _ similar to the one for arrays. However, this is not possible with the current implementation, since there is no way of knowing in advance, which keys matching the globbing expression may be requested via ksLookup.

Instead _ is simply treated like * during matching. Afterwards we check that no array elements were matched.

The basic functionality of the plugin is to just copy (using keyCopyMeta() so we don't waste memory) the metadata of spec keys to all matching (as described above) keys in other namespaces. This ensures that other plugins can do their work as expected, without manually setting metadata on every key. If any metakey we want to copy already exists on the target key (with a different value), this causes a collision conflict.

In addition to the basic functionality, the plugin does some validation itself.

If a spec key has the metakey default set and the key does not exist in other namespaces, we create a special (cascading) key that will be found by ksLookup. This key has the default value as its value. We also copy over metadata as always.

A similar procedure takes place for assign/condition, but in this case, we don't set the value of the key. We only create it and copy metadata.

If a spec key has the metakey require (the value of this metakey is irrelevant and ignored), we ensure that this key exists in at least one other namespace, i.e. it can be found using a cascading ksLookup. If the key cannot be found, this causes a missing conflict.

As hinted to above, we validate array sizes. If a spec key x/# is given, and the spec key x has the metakey array/min or array/max set, we validate the array size (given as metakey array on x) is within the limits of array/min and array/max. Both array/min and array/max have to be valid array-elements similar to array. If the array size is out of bounds, this causes a range conflict.

We also check that the array only contains actual array elements. If this not the case, this causes a member conflict.

Note: We don't actually validate that the array doesn't contain elements above the given array size. This is because it doesn't have anything to do with the specification, whether the array contains additional elements. Note also that we only copy metadata onto elements within the bounds of the array size.

There are various ways in which a conflict can occur during the validation process. To handle these conflicts, we provide various configuration options.

The possible conflicts are:

To define what actions should be taken on on conflicts during kdbGet and kdbSet respectively use the keys conflict/set and conflict/get in the plugin configuration.

This base configuration can be overridden on a per-conflict basis to provide more granularity. The keys for this are (replace _ with get or set accordingly):

conflict/_/member
conflict/_/range
conflict/_/missing
conflict/_/invalid
conflict/_/collision

For even more granularity, the above per-conflict metakeys can also be specified on individual keys.

The allowed values for the conflict keys are always the same:

There is also the special key missing/log. If it is set to 1, the plugin will create the meta array logs/spec/missing/#. This array will contain a list of the missing keys. The key missing/log can only be part of the main plugin configuration, not individual keys' metakeys. It also applies to kdbGet and kdbSet calls.

Ini files can be found in /examples/spec which should be PWD so that the example works:

cd ~e/examples/spec
kdb global-mount # mounts spec plugin by default
kdb mount $PWD/spec.ini spec ni
kdb mount $PWD/spectest.ini /testkey ni
kdb export /testkey ni # note: spec can only applied on cascading access

With spec mount one can use (in this case battery.ini needs to be installed in kdb file spec (this should be preferred on non-development machines so that everything still works after the source is removed):

cp battery.ini $(dirname $(kdb file spec))
kdb mount battery.ini spec/example/battery ni
kdb spec-mount /example/battery
kdb meta-ls /example/battery/level # we see it has a check/enum
kdb meta-get /example/battery/level check/enum # now we know allowed values
kdb set /example/battery/level low # success, low is ok!
kdb set /example/battery/level x # fails, not one of the allowed values!
cp openicc.ini $(dirname $(kdb file spec))
kdb mount openicc.ini spec/freedesktop/openicc ni
kdb spec-mount /freedesktop/openicc
kdb ls /freedesktop/openicc # lets see the whole configuration
kdb export spec/freedesktop/openicc ni # give us details about the specification
kdb meta-ls /freedesktop/openicc/device/camera/#0/EXIF_serial # seems like there is a check/type
kdb set "/freedesktop/openicc/device/camera/#0/EXIF_serial" 203 # success, is a long
kdb set "/freedesktop/openicc/device/camera/#0/EXIF_serial" x # fails, not a long