def parse_binary_search(self, response, minimum=None, maximum=None):
offset = int(get_parameter_value(response.request.url, 'offset'))
first_offset = response.request.meta['first']
if minimum and maximum:
self.logger.info(f'Starting binary search for {first_offset:,} within [{minimum:,}, {maximum:,}]')
elif self.is_http_success(response):
minimum = response.request.meta['minimum']
maximum = offset
else:
minimum = offset + 1
maximum = response.request.meta['maximum']
# If the search succeeded, parse the response as usual. We use a threshold, because getting the exact
# millisecond requires 27 requests.
if minimum + THRESHOLD >= maximum:
self.logger.info(f'New offset found after {first_offset:,} at {maximum:,}!')
if offset == maximum:
# If the last request used the offset, we can reuse its response.
yield from self.parse(response)
else:
url = replace_parameters(response.request.url, offset=maximum)
yield self._build_request(url, self.parse, {})
else:
url = replace_parameters(response.request.url, offset=(minimum + maximum) // 2)
yield self._build_request(url, self.parse_binary_search, {'minimum': minimum, 'maximum': maximum,
'first': first_offset})
def parse_list(self, response):
data = response.json()
# The last page returns an empty JSON object.
if not data:
return
for item in data['data']:
url = replace_parameters(response.request.url, offset=None) + item['ocid']
yield self.build_request(url, formatter=components(-2))
url = replace_parameters(response.request.url, offset=data['offset'])
yield self.build_request(url, formatter=join(components(-1), parameters('offset')), callback=self.parse_list)
def parse_list(self, response):
base_url = 'http://public.eprocurement.systems/ocds/tenders/'
data = json.loads(response.text)
# The last page returns an empty JSON object.
if not data:
return
for item in data['data']:
yield self.build_request(base_url + item['ocid'], formatter=components(-1))
url = replace_parameters(response.request.url, offset=data['offset'])
yield self.build_request(url, formatter=parameters('offset'), callback=self.parse_list)
def parse_pages(self, response):
content = json.loads(response.text)
for url in self.get_files_to_download(content):
yield self.build_request(url,
formatter=components(-1),
dont_filter=True)
pagination = content['pagination']
if pagination['current_page'] < pagination['total_pages']:
page = pagination['current_page'] + 1
url = replace_parameters(response.request.url, page=page)
yield self.build_request(url,
formatter=parameters(
'fecha_desde', 'page'),
dont_filter=True,
callback=self.parse_pages)
def parse_date_range(self, response):
offset = int(get_parameter_value(response.request.url, 'offset'))
# Scrapy uses `datetime.datetime.utcnow()`, so we don't need to worry about time zones.
start_time = int(self.crawler.stats.get_value('start_time').timestamp() * 1000)
# We use the first offset to calculate the new offset, and in log lessages.
first_offset = response.request.meta.get('first', offset)
# The exponent for the exponential search.
exponent = response.request.meta.get('exponent', -1) + 1
# If this offset succeeded, do a binary search from the previous offset to this offset.
if self.is_http_success(response):
yield from self.parse_binary_search(response, response.request.meta['prev'], offset)
# If this offset failed and reached a limit, stop.
elif offset >= start_time or exponent > EXPONENT_LIMIT:
self.logger.info(f'No offset found after {first_offset:,} within {2 ** EXPONENT_LIMIT} days.')
yield self.build_file_error_from_response(response)
# Otherwise, continue.
else:
new_offset = min(first_offset + MILLISECONDS_PER_DAY * 2 ** exponent, start_time)
url = replace_parameters(response.request.url, offset=new_offset)
yield self._build_request(url, self.parse_date_range, {'prev': offset, 'exponent': exponent,
'first': first_offset})
def _build_url(self, params):
url_params = params.copy()
url_params.update(self.additional_params)
return util.replace_parameters(self.base_url, **url_params)
def _set_base_url(self, url):
self.base_url = util.replace_parameters(url, page=None, limit=None, offset=None)
def test_replace_parameters(url, value, expected):
assert replace_parameters(url, page=value) == expected
