support for xsd:duration

[wrobell]

The script below gives output

[(rdflib.term.URIRef('file:///home/wrobell/projects/uddo/#c1'), None),
 (rdflib.term.URIRef('file:///home/wrobell/projects/uddo/#c2'), None)]

It would be nice to get time difference as xsd:duration instead of nulls.

The script

from pprint import pprint
import rdflib
import io

data = """
@prefix : <#>.
@prefix rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#>.
@prefix rdfs: <http://www.w3.org/2000/01/rdf-schema#>.
@prefix xsd: <http://www.w3.org/2001/XMLSchema#>.

:C a rdfs:Class.

:start a rdf:Property;
    rdfs:domain :C; 
    rdfs:range xsd:dateTime.

:end a rdf:Property;
    rdfs:domain :C; 
    rdfs:range xsd:dateTime.

:c1 a :C;
    :start "2016-01-01 20:00:00"^^xsd:dateTime;
    :end "2016-01-02 20:01:00"^^xsd:dateTime.

:c2 a :C;
    :start "2016-01-01 20:05:00"^^xsd:dateTime;
    :end "2016-01-01 20:30:00"^^xsd:dateTime.
"""

graph = rdflib.Graph()

f = io.StringIO(data)
graph.parse(f, format='n3')

result = graph.query("""
SELECT ?c ?duration
WHERE {
    BIND(?end - ?start AS ?duration)
    ?c :start ?start;
        :end ?end.
}
""")

pprint(list(result))

[gromgull]

SPARQL only defines - + etc for numeric types: https://www.w3.org/TR/sparql11-query/#OperatorMapping

Note that raw literals in RDFLib also do not support arithmetic:

In [110]: l1=rdflib.Literal(datetime.now())

In [111]: l2=rdflib.Literal(datetime.now())

In [112]: l1-l2
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-112-fe8dcc4af6cc> in <module>()
----> 1 l1-l2

TypeError: unsupported operand type(s) for -: 'Literal' and 'Literal'

(Although we support +, but in lexical space:

In [113]: l1+l2
Out[113]: rdflib.term.Literal(u'2017-01-19T15:11:29.4567162017-01-19T15:11:32.352804', datatype=rdflib.term.URIRef(u'http://www.w3.org/2001/XMLSchema#dateTime'))


In [114]: rdflib.Literal(1) + rdflib.Literal(2)
Out[114]: rdflib.term.Literal(u'12', datatype=rdflib.term.URIRef(u'http://www.w3.org/2001/XMLSchema#integer'))

🙈
)

BUT! That doesn't mean we cannot extend it! Ideally we implement the operators for all typed literals (see #451), and then we expose it in SPARQL.

This would be very nice, but is not really a blocker for 4.2.2, moving to 5.0

[Akash-Sharma-1]

Hi! Can I work on this issue? Seems interesting to me. Just waiting for the green light from the maintainers and contributors.

[JervenBolleman]

@Akash-Sharma-1 yes you can work on this! I suspect the area you need to add code to is

rdflib/rdflib/plugins/sparql/operators.py

Line 755 in bd82150

def AdditiveExpression(e, ctx):

But I am not sure.

Also see https://github.com/w3c/sparql-12/blob/master/SEP/SEP-0002/sep-0002.md

[Akash-Sharma-1]

Thanks for the information @JervenBolleman. 👍
Yes, I was also expecting to work in this part of the code.

[Akash-Sharma-1]

Hi!
So, I have been working on this issue for quite some time. I have extended the handling of arithmetic operations (addition/subtraction) of dates, times and durations for SPARQL queries.

So far, I have implemented and tested out these operations on my local code which are also mentioned here - (https://github.com/w3c/sparql-12/blob/master/SEP/SEP-0002/sep-0002.md) (@JervenBolleman thanks again for the link)
The above SPARQL query of the issue also works well with the fix.

I wanted to discuss some points regarding these operations before finalizing the PR:

1. Would it be beneficial, to handle multiple operands for these operations in the query similar to numeric types?
   For example, (datetime1 - datetime2 - dateTime3) or (dateTime1 - dayMonthDuration1 - dayMonthDuration2)
   These type of multiple operands operations are however not discussed in w3c SPARQL.
2. In cases of subtraction of dates or dateTimes, if the first date is earlier to the second date then the resulting duration comes out to be negative. Since negative xsd:duration is valid, I am allowing this case in my code. Would it be helpful if I notify the user about the subtraction from smaller date?
   An example of this case, "2016-01-01T20:00:00"^^xsd:dateTime - "2016-01-02T20:01:00"^^xsd:dateTime would result in "-P1DT1M"^^xsd:duration

[nicholascar]

@Akash-Sharma-1

My personal answers to your questions (i.e. not anything official from all of RDFlib!):

1. Yes! Just as normal maths allow multiple operands, I think this is expected, but certainly not required, behaviour

2. I wouldn't bother warning the user, they'll see the negative result. If they want to be warned, they could write a check in their application code. As you say, negative xsd:duration is valid