datagraph/wikidata_ext.py

import ratelimit
import urllib.parse
import wikidata.entity
import requests
import json
import logging

REQUEST_SESSION = None # TODO?

def concept_uri(obj):
    assert isinstance(obj, wikidata.entity.Entity), obj
    if obj.id.startswith('P'):
        return urllib.parse.urlparse('http://www.wikidata.org/prop/direct/{}'.format(obj.id))
    elif obj.id.startswith('Q'):
        return urllib.parse.urlparse('http://www.wikidata.org/entity/{}'.format(obj.id))
    else:
        assert False, "TODO: " + ojb.id

def fmt_triple_value(obj, prefer_obj = False):
    if obj is None:
        return ''
    if isinstance(obj, str):
        return '"{}"'.format(obj)
    elif isinstance(obj, urllib.parse.ParseResult):
        return obj.geturl() if prefer_obj else fmt_triple_value(obj.geturl())
    elif isinstance(obj, wikidata.entity.Entity):
        uri = concept_uri(obj)
        return fmt_triple_value(uri, True)
    else:
        assert False, type(obj)

@ratelimit.sleep_and_retry
def fetch_by_url(url, headers):
    logging.debug('Fetching: %s', url)
    assert REQUEST_SESSION is not None, 'REQUEST_SESSION must be set, before calling fetch_by_url'
    response = REQUEST_SESSION.get(url, headers = headers)
    if response.status_code != 200:
        logging.error('Got %s error message: %s', response.status_code, response.text)
        return None
    return response

ITEMS_PER_PAGE = "http://www.w3.org/ns/hydra/core#itemsPerPage"
TOTAL_ITEMS = "http://www.w3.org/ns/hydra/core#totalItems"

def fmt_params(subject, predicate, object):
    derp = [x for x in [subject, predicate, object] if x]
    assert len(derp) >= 1, str(derp)
    params = {
        'subject': fmt_triple_value(subject, prefer_obj = True),
        'predicate': fmt_triple_value(predicate, prefer_obj = True),
        'object': fmt_triple_value(object, prefer_obj = True),
        'page': 1,
    }
    return params

def get_triples_count(subject = None, predicate = None, object = None):
    '''
    Fetches first page in order to determine amount of items.
    '''
    params = fmt_params(subject, predicate, object)
    url = requests.Request(url = 'https://query.wikidata.org/bigdata/ldf', params = params).prepare().url
    response = fetch_by_url(url, headers = {'accept': 'application/ld+json'})
    if response is None:
        return {
            'items_per_page': 0,
            'items_total': 0,
            'num_pages': 0,
        }
    json_data = response.json()
    for item in json_data['@graph']:
        if TOTAL_ITEMS in item:
            return {
                'items_per_page': item[ITEMS_PER_PAGE],
                'items_total': item[TOTAL_ITEMS],
                'num_pages': int((item[TOTAL_ITEMS] - 1) / item[ITEMS_PER_PAGE] + 1),
            }
    assert False

def get_triples_internal(subject, predicate, object):
    params = fmt_params(subject, predicate, object)
    pagination_data = get_triples_count(subject, predicate, object)
    for current_page in range(1, pagination_data['num_pages']+1):
        params['page'] = current_page
        url = requests.Request(url = 'https://query.wikidata.org/bigdata/ldf', params = params).prepare().url
        response = fetch_by_url(url, headers = {'accept': 'application/ld+json'})
        json_data = response.json()

        for item in json_data['@graph']:
            if item['@id'].startswith('_:b'):
                continue
            if item['@id'].startswith('https://query.wikidata.org/bigdata/ldf'):
                continue
            if item['@id'].startswith('http://www.wikidata.org/.well-known/'):
                continue
            yield item

        # Bookkeeping
        del url, response, json_data

SCHEMA_ABOUT = urllib.parse.urlparse('http://schema.org/about')

def get_wikidata_concept_for_wikipedia_page(client, wikipage):
    triples = get_triples_internal(wikipage, SCHEMA_ABOUT, None);
    triples = list(triples)
    for item in triples:
        s = item['about'][3:]
        return client.get(s, load = False)

def get_triples(client, subject = None, predicate = None, object = None):
    triples = []
    iterator = get_triples_internal(subject, predicate, object)
    for item in iterator:
        is_looking_for = item['@id'].startswith('wd:') and predicate.id in item
        if is_looking_for :
            s = subject
            if s is None:
                s = client.get(item['@id'][3:], load = False)
            o = object or item[predicate.id]
            yield (s, predicate, o)
        del item, is_looking_for

def get_backlinks(client, predicate, object):
    for subject, _, _ in get_triples(client, subject = None, predicate = predicate, object = object):
        yield subject