What this package do?
IMS VDEX is a standard for exchanging vocabularies.
collective.vdexvocabulary
bridges between vdex vocabularies and zope
vocabularies (zope.schema.vocabularies
), so you can easily use it in
systems like Plone/ Zope, Pyramid or any other Zope-Toolkit supporting system.
Contents
Whats so special about it?
collective.vdexvocabulary
contains two different types of vocabularies:
- If you have big vocabularies with a lot of relations, like +10.000
terms with +30.000 relations, so this would be perfect use case to use
the
VdexVocabulary
type ofcollective.vdexvocabulary
. - If you have tree-like vocabularies this is perfect too.
TreeVocabulary
supports nested/ hierachical vocabularies.
Also there is other stuff which not supported by other vocabulary packages (i.e. for Plone/Zope):
- i18n support. IMS VDEX supports translations within the VDEX-XML-File, both vocabulary types are supporting this way of translations.
- proper order also with unicode characters (if zope.ucol is installed, vdexvocabulary only). If VDEX is order-sigificant the order given by vdex file is taken (supported by both vocabulary types).
- easy registration using zcml
- relations as it specified in IMS VDEX standard (for now only VdexVocabulary)
How do I use it?
Configuration
In your configure.zcml add:
<configure ... xmlns:vdex="http://namespaces.zope.org/vdex" ...> <include package="collective.vdexvocabulary" file="meta.zcml" /> <include package="collective.vdexvocabulary" />
And to register a vdex vocabulary simply add line bellow pointing to file containing vdex vocabulary:
<configure ... xmlns:vdex="http://namespaces.zope.org/vdex" ...> <vdex:vocabulary file="path-to/very-interesting.xml" />
or for tree vocabularies:
<vdex:treevocabulary file="path-to/very-interesting-tree.xml" />
To make registration of vocabularies even easier you can also register several vocabularies and just point to directory:
<configure ... xmlns:vdex="http://namespaces.zope.org/vdex" ...> <vdex:vocabulary directory="path-to/my-vdex-vocabularies" />
or for tree vocabularies:
<vdex:treevocabulary directory="path-to/my-vdex-vocabularies" />
vdex files in path-to/my-vdex-vocabularies
directory should have ending
.vdex
to be recognized by vdex:vocabulary
ZCML directive.
Sometimes you dont want VDEX-files inside your code tree. Therefore an environment variable can be given defining the base directory:
<vdex:vocabulary file="my-vocabulary.vdex: environment="VDEX_BASE_DIR" />
or for tree vocabularies:
<vdex:treevocabulary file="my-vocabulary.vdex: environment="VDEX_BASE_DIR" />
Before running the code with environment variable relative filenames/directories
one has to set the environmant variable, i.e. do an
export VDEX_BASE_DIR=/home/joe/vdex/
in order to make it look for the vdex
at /home/joe/vdex/my-vocabulary.vdex
.
Usage in Code
Vocabularies are named utiltiies in a zope-toolkit manner. The name is taken
from the vdex file, its the vocabIdentifier
.
Given a vocabulary with the name beeeurope
(see tree example below) one has
to get the utility using the zope component architecture way:
>>> from zope.component import getUtility >>> from zope.schema.interfaces import IVocabularyFactory >>> factory = zope.component.getUtility(IVocabularyFactory, 'beeeurope')
The factory returns on call a vocabulary. It expects a context, which can be
None
in our case. If you are in an application server pass here your current
context. In case of flat vocabularies this is used to detect the language, for
tree vocabularies it is ignored, here an more advanced method is used to support
i18n:
>>> context = None >>> vocabulary = factory(context)
Now you can use the vocabulary:
>>> for term in vocabulary: ... print term.value ... print term.token ... print term.title ... print term.description
How to use tree-vocabularies
Once looked up as shown above traversing the tree is easy. It works as defined
in zope.schema.interfaces.ITreeVocabulary
. The term
is also the key for
the sublevel:
>>> def printlevel(leveldict, ident=0): ... for term in leveldict: ... print indent * ' ' + term.title ... printlevel(leveldict[term], indent+1)
Hint: collective.dynatree
uses this kind of vocabularies and can be used as
an example for own implementations too.
How to access relations (from code)
Relations are defined by ISO2788.
To get listing of BMW car models from above VDEX example you have to:
>>> from zope.schema.vocabulary import getVocabularyRegistry >>> vr = getVocabularyRegistry() >>> car_manufacturers = vr.get(self.context, 'your.package.car_manufacturers') >>> car_models = vr.get(self.context, 'your.package.car_models') >>> bmw = car_manufacturers.getTerm('bmw') >>> bmw_car_models = bmw.related.get('NT', [])
Example VDEX file
Flat with with relations
Example of car manufacturers list (car_manufacturers.vdex
).:
<?xml version="1.0" encoding="utf-8"?> <vdex xmlns="http://www.imsglobal.org/xsd/imsvdex_v1p0" orderSignificant="false" language="en"> <vocabIdentifier>your.package.car_manufacturers</vocabIdentifier> <term> <termIdentifier>ford</termIdentifier> <caption> <langstring language="en">Ford</langstring> <langstring language="es">Una miedra de coche</langstring> </caption> </term> <term> <termIdentifier>bmw</termIdentifier> <caption> <langstring language="en">BMW</langstring> <langstring language="es">Be-eMe-uWe, mierda</langstring> </caption> </term> <relationship> <sourceTerm>bmw</sourceTerm> <targetTerm vocabIdentifier="your.package.car_models">very-special-bmw-model</targetTerm> <relationshipType source="http://www.imsglobal.org/vocabularies/iso2788_relations.xml">NT</relationshipType> </relationship> ... </vdex>
List of car models (car_models.vdex
).:
<?xml version="1.0" encoding="utf-8"?> <vdex xmlns="http://www.imsglobal.org/xsd/imsvdex_v1p0" orderSignificant="false" language="en"> <vocabIdentifier>your.package.car_models</vocabIdentifier> <term> <termIdentifier>very-special-bmw-model</termIdentifier> <caption> <langstring language="en">Very special BMW model</langstring> <langstring language="es">Un modelo de Be-eMe-uWe</langstring> </caption> </term> <relationship> <sourceTerm>very-special-bmw-model</sourceTerm> <targetTerm vocabIdentifier="your.package.car_manufacturers">bmw</targetTerm> <relationshipType source="http://www.imsglobal.org/vocabularies/iso2788_relations.xml">BT</relationshipType> </relationship> ... </vdex>
Hierachical Tree
example of a tree vocabulary:
<vdex xmlns="http://www.imsglobal.org/xsd/imsvdex_v1p0" orderSignificant="true"> <vocabIdentifier>beeeurope</vocabIdentifier> <vocabName> <langstring language="en">European Honey Bees</langstring> </vocabName> <term> <termIdentifier>nwe</termIdentifier> <caption> <langstring language="en">North-west of Europe</langstring> </caption> <term> <termIdentifier>nwe.1</termIdentifier> <caption> <langstring language="en">A. m. iberica</langstring> </caption> </term> <term> <termIdentifier>nwe.2</termIdentifier> <caption> <langstring language="en">A. m. intermissa</langstring> </caption> </term> <term> <termIdentifier>nwe.3</termIdentifier> <caption> <langstring language="en">A. m. lihzeni</langstring> </caption> </term> <term> <termIdentifier>nwe.4</termIdentifier> <caption> <langstring language="en">A. m. mellifera</langstring> </caption> </term> <term> <termIdentifier>nwe.5</termIdentifier> <caption> <langstring language="en">A. m. sahariensis</langstring> </caption> </term> </term> <term> <termIdentifier>swe</termIdentifier> <caption> <langstring language="en">South-west of Europe</langstring> </caption> <term> <termIdentifier>swe.1</termIdentifier> <caption> <langstring language="en">A. m. carnica</langstring> </caption> </term> <term> <term> <termIdentifier>swe.2</termIdentifier> <caption> <langstring language="en">A. m. cecropia</langstring> </caption> </term> <term> <termIdentifier>swe.3</termIdentifier> <caption> <langstring language="en">A. m. ligustica</langstring> </caption> </term> <term> <termIdentifier>swe.4</termIdentifier> <caption> <langstring language="en">A. m. macedonica</langstring> </caption> </term> <term> <termIdentifier>swe.5</termIdentifier> <caption> <langstring language="en">A. m. ruttneri</langstring> </caption> </term> <term> <termIdentifier>swe.6</termIdentifier> <caption> <langstring language="en">A. m. sicula</langstring> </caption> </term> </term> </vdex>
Where can I complain / help / send rum?
Home + Source: | https://github.com/collective/collective.vdexvocabulary |
---|---|
Report Issues: | https://github.com/collective/collective.vdexvocabulary/issues |
Send rum: | contact rok@garbas.si for more info |
Credit
- Rok Garbas, http://garbas.si, <rok@garbas.si>, Author
- Seantis gmbh, http://www.seantis.ch Thank you for initial idea with seantis.vdex where got the idea and then reimplement and extend it.
- Jens W Klein, http://kleinundpartner.at, <jens@bluedynamics.com>, Cleanup, Treevocabulary/ i18n-support
TODO
- fetch vocab(s) via url (new directive)
- load vocabs view entry_points
- store vocabs (or changed vocabs in zodb), will probably also need diff and merge option
- write test and get decent test coverage
- write documentation
- make ZCML optional
- make through the web vdex editor (this would probably need sponsoring)
- add relation support to TreeVocabulary