import re import sys import csv import json import time import itertools import datetime import requests from confluent_kafka import Producer, KafkaException from urllib.parse import urlparse, parse_qs from fatcat_tools.workers import most_recent_message from .harvest_common import HarvestState, requests_retry_session class HarvestCrossrefWorker: """ Crossref API date fields (and our interpretation):: - https://github.com/CrossRef/rest-api-doc#filter-names - *-index-date: "metadata indexed" is the API/index record update time - *-deposit-date: "metadata last (re)deposited" is the catalog record update time - *-update-date: "Metadata updated (Currently the same as *-deposit-date)" - *-created-date: "metadata first deposited" - *-pub-date (etc): publisher-supplied, not "meta-meta-data" https://api.crossref.org/works?filter=from-index-date:2018-11-14&rows=2 Also from the REST API: Notes on incremental metadata updates When using time filters to retrieve periodic, incremental metadata updates, the from-index-date filter should be used over from-update-date, from-deposit-date, from-created-date and from-pub-date. The timestamp that from-index-date filters on is guaranteed to be updated every time there is a change to metadata requiring a reindex. However, when Crossref re-indexes tens of millions of rows, using from-index-date can be very slow, taking several days to process a single day of updates. I think the design is going to have to be a cronjob or long-running job (with long sleeps) which publishes "success through" to a separate state queue, as simple YYYY-MM-DD strings. Within a day, will need to use a resumption token. Maybe should use a crossref library... meh. will want to have some mechanism in kafka consumer (pushing to fatcat) to group in batches as well. maybe even pass through as batches? or just use timeouts on iteration. logic of this worker: - on start, fetch latest date from state feed - in a function (unit-testable), decide which dates to ingest - for each date needing update: - start a loop for just that date, using resumption token for this query - when done, publish to state feed, with immediate sync TODO: what sort of parallelism? I guess multi-processing on dates, but need to be careful how state is serialized back into kafka. """ def __init__(self, kafka_hosts, produce_topic, state_topic, contact_email, api_host_url="https://api.crossref.org/works", start_date=None, end_date=None): self.api_host_url = api_host_url self.produce_topic = produce_topic self.state_topic = state_topic self.contact_email = contact_email self.kafka_config = { 'bootstrap.servers': kafka_hosts, 'message.max.bytes': 20000000, # ~20 MBytes; broker is ~50 MBytes } self.state = HarvestState(start_date, end_date) self.state.initialize_from_kafka(self.state_topic, self.kafka_config) self.loop_sleep = 60*60 # how long to wait, in seconds, between date checks self.api_batch_size = 50 self.name = "Crossref" self.producer = self._kafka_producer() def _kafka_producer(self): def fail_fast(err, msg): if err is not None: print("Kafka producer delivery error: {}".format(err), file=sys.stderr) print("Bailing out...", file=sys.stderr) # TODO: should it be sys.exit(-1)? raise KafkaException(err) self._kafka_fail_fast = fail_fast producer_conf = self.kafka_config.copy() producer_conf.update({ 'delivery.report.only.error': True, 'default.topic.config': { 'request.required.acks': -1, # all brokers must confirm }, }) return Producer(producer_conf) def params(self, date_str): filter_param = 'from-update-date:{},until-update-date:{}'.format( date_str, date_str) return { 'filter': filter_param, 'rows': self.api_batch_size, 'cursor': '*', } def update_params(self, params, resp): params['cursor'] = resp['message']['next-cursor'] return params def extract_key(self, obj): return obj['DOI'].encode('utf-8') def fetch_date(self, date): date_str = date.isoformat() params = self.params(date_str) http_session = requests_retry_session() http_session.headers.update({ 'User-Agent': 'fatcat_tools/0.1.0 (https://fatcat.wiki; mailto:{}) python-requests'.format( self.contact_email), }) count = 0 while True: http_resp = http_session.get(self.api_host_url, params=params) if http_resp.status_code == 503: # crude backoff; now redundant with session exponential # backoff, but allows for longer backoff/downtime on remote end print("got HTTP {}, pausing for 30 seconds".format(http_resp.status_code), file=sys.stderr) # keep kafka producer connection alive self.producer.poll(0) time.sleep(30.0) continue if http_resp.status_code == 400: print("skipping batch for {}, due to HTTP 400. Marking complete. Related: https://github.com/datacite/datacite/issues/897".format(date_str), file=sys.stderr) break http_resp.raise_for_status() resp = http_resp.json() items = self.extract_items(resp) count += len(items) print("... got {} ({} of {}), HTTP fetch took {}".format(len(items), count, self.extract_total(resp), http_resp.elapsed), file=sys.stderr) #print(json.dumps(resp)) for work in items: self.producer.produce( self.produce_topic, json.dumps(work).encode('utf-8'), key=self.extract_key(work), on_delivery=self._kafka_fail_fast) self.producer.poll(0) if len(items) < self.api_batch_size: break params = self.update_params(params, resp) self.producer.flush() def extract_items(self, resp): return resp['message']['items'] def extract_total(self, resp): return resp['message']['total-results'] def run(self, continuous=False): while True: current = self.state.next_span(continuous) if current: print("Fetching DOIs updated on {} (UTC)".format(current), file=sys.stderr) self.fetch_date(current) self.state.complete(current, kafka_topic=self.state_topic, kafka_config=self.kafka_config) continue if continuous: print("Sleeping {} seconds...".format(self.loop_sleep), file=sys.stderr) time.sleep(self.loop_sleep) else: break print("{} DOI ingest caught up".format(self.name), file=sys.stderr) class HarvestDataciteWorker(HarvestCrossrefWorker): """ datacite has a REST API as well as OAI-PMH endpoint. have about 8 million bulk export notes: https://github.com/datacite/datacite/issues/188 fundamentally, very similar to crossref. don't have a scrape... maybe could/should use this script for that, and dump to JSON? """ def __init__(self, kafka_hosts, produce_topic, state_topic, contact_email, api_host_url="https://api.datacite.org/dois", start_date=None, end_date=None): super().__init__(kafka_hosts=kafka_hosts, produce_topic=produce_topic, state_topic=state_topic, api_host_url=api_host_url, contact_email=contact_email, start_date=start_date, end_date=end_date) # for datecite, it's "from-update-date" self.name = "Datacite" def params(self, date_str): """ Dates have to be supplied in 2018-10-27T22:36:30.000Z format. """ return { 'query': 'updated:[{}T00:00:00.000Z TO {}T23:59:59.999Z]'.format(date_str, date_str), 'page[size]': self.api_batch_size, 'page[cursor]': 1, } def extract_items(self, resp): return resp['data'] def extract_total(self, resp): return resp['meta']['total'] def extract_key(self, obj): return obj['attributes']['doi'].encode('utf-8') def update_params(self, params, resp): """ Using cursor mechanism (https://support.datacite.org/docs/pagination#section-cursor). $ curl -sL https://is.gd/cLbE5h | jq -r .links.next Example: https://is.gd/cLbE5h Further API errors reported: https://github.com/datacite/datacite/issues/897 (HTTP 400) https://github.com/datacite/datacite/issues/898 (HTTP 500) """ parsed = urlparse(resp['links']['next']) page_cursor = parse_qs(parsed.query).get('page[cursor]') if not page_cursor: raise ValueError('no page[cursor] in .links.next') params['page[cursor]'] = page_cursor[0] return params