Skip to content
forked from thegetty/crom

Python library to make creation of CIDOC CRM easier by mapping classes/predicates to python objects

License

Notifications You must be signed in to change notification settings

linked-art/crom

 
 

Repository files navigation

Build Status Coverage Status

Cromulent

A Python library to make creation of CIDOC CRM easier by mapping classes/predicates to python objects/properties, thereby making the CRM "CRoMulent", a Simpsons neologism for "acceptable" or "fine".

Status: Beta

The core vocabulary loading functionality is reasonably stable. The vocabulary section is expanding as we find new, useful terms to include and will likely constantly change.

The code is actively being developed and compability breaking changes are thus to be expected as we use it in various projects across The J Paul Getty Trust, and beyond.

How to Use It

Basic Usage

Import the classes from the model module. As the classes are dynamically generated, they're not in the code but will be there once the build_classes function has been called.

from cromulent.model import factory, Group
g1 = Group(ident="Organization")
g2 = Group(ident="Department")
g1.member = g2
print factory.toString(g1, compact=False)

The constructor for the classes takes the following parameters:

  • ident - an identifier to use for this instance. If specified, it should be a URI represented as a string. If it is the empty string, it will result in no identifier. If not specified, or specified as None, then it will be auto-generated by the factory if auto_assign_id is true, or if auto_assign_id is false, then it will result in no identifier.
  • label - a human readable label for the resource, to act as internal documentation for the data
  • value or content - a data value for the class. Dimensions and MonetaryAmounts use value which must be a number, and Name, Identifier, LinguisticObject and similar use content which must be a string.
  • Additional keywords may be passed in, and will be sent to class-specific initialization code.

Vocabulary

from cromulent.model import factory
from cromulent.vocab import Height
h = Height()
h.value = 6
print factory.toString(h, compact=False)

Tricks and Gotchas

  • Assigning to the same property repeatedly does NOT overwrite the value, instead it appends. To overwrite a value, instead set it to a false value first.

Factory settings

There are quite a few settings for how the module works, which are managed by a factory object.

URI and File System Configuration:

  • base_url The base url on to which to append any slug given when an object is created
  • base_dir The base directory into which to write files, via factory.toFile()
  • filename_extension The extension to use on files written via toFile(), defaults to ".json"
  • default_lang The code for the default language to use on text values
  • context_uri The URI to use for @context in the JSON-LD serialization
  • context_json The parsed JSON object of the context from which the prefixes are derived
  • full_names Should the serialization use the full CRM names for classes and properties instead of the more readable ones defined in the mapping, defaults to False
  • prefixes A dictionary of prefix to URI for URIs to compress down to prefix:slug format
  • prefixes_rev The reverse of the prefixes dictionary
  • pipe_scoped_contexts A convenience setting for generating documentation, where properties that map to the same JSON output are represented as short_name|full_name to be post-processed.
  • json_indent How many spaces should each level of indentation be when serializing to a human readable form, defaults to 2
  • id_type_label Should the id, type and label properties all be used when serializing resources that have already been processed, defaults to True
  • elasticsearch_compatible Despite JSON-LD 1.0 compaction rules, should a single URI be represented as {"@id": "URI"} rather than just "URI", to make the resulting JSON compatible with elasticsearch and similar JSON processing engines. Defaults to False.
  • serialize_all_resources NOT YET IMPLEMENTED. If true, then all resources will be serialized separately, not just the top level resource.

Model Validation and Generation:

  • materialize_inverses Should the inverse relationships be set automatically, defaults to False
  • validate_properties Should the model be validated at run time when setting properties, defaults to True (this allows you to save processing time once you're certain your code does the right thing)
  • validate_properties Should the properties be validated as being part of the model at all
  • validate_profile Should the profile of which terms should be used be validated
  • process_multiplicity Should properties that allow multiple values always be an array
  • validate_range Should the object be validated that it is legal to be the value of the property
  • auto_assign_id Should a URI be autogenerated and assigned, defaults to True
  • auto_id_type The method by which the URI is generated, taken from the following values:
    • "int" (just increment an integer in a single value space)
    • "int-per-type" (increment an integer, with a separate value space per class)
    • "int-per-segment" (increment an integer, with a separate value space per URI segment associated with a class)
    • "uuid" (just use UUIDs everywhere)

Internal:

  • debug_level Settings for debugging errors and warnings, defaults to "warn"
  • log_stream An object implementing the stream API to write log messages to, defaults to sys.stderr

How it Works

At import time, the library parses the vocabulary data file (data/crm_vocab.tsv) and creates Python classes in the module's global scope from each of the defined RDF classes. The names of the classes are intended to be easy to use and remember, not necessarily identical to the CRM ontology's names. It also records the properties that can be used with that class, and at run time checks whether the property is defined and that the value fits the defined range.

Hacking

You can change the mapping by tweaking utils/vocab_reader.py and rerunning it to build a new TSV input file. See also the experimental code for loading completely different ontologies.

About

Python library to make creation of CIDOC CRM easier by mapping classes/predicates to python objects

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 100.0%