otsrdflib 0.5.0

Ordered Turtle Serializer for rdflib

Ordered Turtle Serializer for rdflib

An extension to the rdflib Turtle serializer that adds order (at the price of speed). Useful when you need to generate diffs between Turtle files, or just to make it easier for human beings to inspect the files.

$ pip install otsrdflib

Usage:

from rdflib import graph
from otsrdflib import OrderedTurtleSerializer

my_graph = Graph()

out = open('out.ttl', 'wb')
serializer = OrderedTurtleSerializer(my_graph)
serializer.serialize(out)

Class order

By default, classes are ordered alphabetically by their URIS.

A custom order can be imposed by adding classes to the class_order attribute. For a SKOS vocabulary, for instance, you might want to sort the concept scheme first, followed by the other elements of the vocabulary:

serializer.class_order = [
    SKOS.ConceptScheme,
    SKOS.Concept,
    ISOTHES.ThesaurusArray,
]

Any class not included in the class_order list will be sorted alphabetically at the end, after the classes included in the list.

Instance order

By default, instances of a class are ordered alphabetically by their URIS.

A custom order can be imposed by defining functions that generate sort keys from the URIs. For instance, you could define a function that returns the numeric last part of an URI to be sorted numerically:

serializer.sorters = [
    ('.*?/[^0-9]*([0-9.]+)$', lambda x: float(x[0])),
]

The first element of the tuple (‘.*?/[^0-9]*([0-9.]+)$’) is the regexp pattern to be matched against the URIs, while the second element (lambda x: float(x[0])) is the sort key generating function. In this case, it returns the first backreference as a float.

The patterns in sorters will be attempted matched against instances of any class. You can also define patterns that will only be matched against instances of a specific class. Let’s say you only wanted to sort instances of SKOS.Concept this way:

from rdflib.namespace import SKOS

serializer.sorters_by_class = {
    SKOS.Concept: [
        ('.*?/[^0-9]*([0-9.]+)$', lambda x: float(x[0])),
    ]
}

For a slightly more complicated example, let’s look at Dewey. Classes in the main schedules are describes by URIs like http://dewey.info/class/001.433/e23/, and we will use the class number (001.433) for sorting. But there’s also table classes like http://dewey.info/class/1–0901/e23/. We want to sort these at the end, after the main schedules. To achieve this, we define two sorters, one that matches the table classes and one that matches the main schedule classes:

serializer.sorters = [
    ('/([0-9A-Z\-]+)\-\-([0-9.\-;:]+)/e', lambda x: 'T{}--{}'.format(x[0], x[1])),  # table numbers
    ('/([0-9.\-;:]+)/e', lambda x: 'A' + x[0]),  # main schedule numbers
]

By prefixing the table numbers with ‘T’ and the main schedule numbers with ‘A’, we ensure the table numbers are sorted after the main schedule numbers.

Changes in version 0.5

• The topClasses attribute was renamed to class_order to better reflect its content and comply with PEP8. It was also changed to be empty by default, since the previous default list was rather random.

• A sorters_by_class attribute was added to allow sorters to be defined per class.