moviewyrm/bookwyrm/connectors/inventaire.py

220 lines
8.9 KiB
Python
Raw Normal View History

""" inventaire data connector """
2021-04-29 00:18:14 +00:00
import re
2021-04-07 01:34:55 +00:00
from bookwyrm import models
2021-09-16 18:30:04 +00:00
from bookwyrm.book_search import SearchResult
from .abstract_connector import AbstractConnector, Mapping
from .abstract_connector import get_data
2021-04-06 20:39:10 +00:00
from .connector_manager import ConnectorException
class Connector(AbstractConnector):
2021-09-29 19:38:31 +00:00
"""instantiate a connector for inventaire"""
def __init__(self, identifier):
super().__init__(identifier)
2021-04-07 00:46:06 +00:00
get_first = lambda a: a[0]
shared_mappings = [
Mapping("id", remote_field="uri", formatter=self.get_remote_id),
Mapping("bnfId", remote_field="wdt:P268", formatter=get_first),
Mapping("openlibraryKey", remote_field="wdt:P648", formatter=get_first),
]
self.book_mappings = [
2021-04-06 20:53:58 +00:00
Mapping("title", remote_field="wdt:P1476", formatter=get_first),
Mapping("title", remote_field="labels", formatter=get_language_code),
2021-04-06 20:53:58 +00:00
Mapping("subtitle", remote_field="wdt:P1680", formatter=get_first),
Mapping("inventaireId", remote_field="uri"),
2021-04-29 17:40:49 +00:00
Mapping(
"description", remote_field="sitelinks", formatter=self.get_description
),
Mapping("cover", remote_field="image", formatter=self.get_cover_url),
2021-04-06 20:53:58 +00:00
Mapping("isbn13", remote_field="wdt:P212", formatter=get_first),
Mapping("isbn10", remote_field="wdt:P957", formatter=get_first),
Mapping("oclcNumber", remote_field="wdt:P5331", formatter=get_first),
2021-04-07 00:46:06 +00:00
Mapping("goodreadsKey", remote_field="wdt:P2969", formatter=get_first),
Mapping("librarythingKey", remote_field="wdt:P1085", formatter=get_first),
2021-04-07 01:10:42 +00:00
Mapping("languages", remote_field="wdt:P407", formatter=self.resolve_keys),
Mapping("publishers", remote_field="wdt:P123", formatter=self.resolve_keys),
2021-04-06 20:53:58 +00:00
Mapping("publishedDate", remote_field="wdt:P577", formatter=get_first),
Mapping("pages", remote_field="wdt:P1104", formatter=get_first),
2021-04-07 01:10:42 +00:00
Mapping(
"subjectPlaces", remote_field="wdt:P840", formatter=self.resolve_keys
),
Mapping("subjects", remote_field="wdt:P921", formatter=self.resolve_keys),
2021-04-06 20:53:58 +00:00
Mapping("asin", remote_field="wdt:P5749", formatter=get_first),
2021-04-07 00:46:06 +00:00
] + shared_mappings
# TODO: P136: genre, P674 characters, P950 bne
2021-04-07 00:46:06 +00:00
self.author_mappings = [
Mapping("id", remote_field="uri", formatter=self.get_remote_id),
Mapping("name", remote_field="labels", formatter=get_language_code),
2021-04-29 17:30:38 +00:00
Mapping("bio", remote_field="sitelinks", formatter=self.get_description),
2021-04-07 00:46:06 +00:00
Mapping("goodreadsKey", remote_field="wdt:P2963", formatter=get_first),
Mapping("isni", remote_field="wdt:P213", formatter=get_first),
Mapping("viafId", remote_field="wdt:P214", formatter=get_first),
Mapping("gutenberg_id", remote_field="wdt:P1938", formatter=get_first),
Mapping("born", remote_field="wdt:P569", formatter=get_first),
Mapping("died", remote_field="wdt:P570", formatter=get_first),
] + shared_mappings
def get_remote_id(self, value):
"""convert an id/uri into a url"""
2021-09-18 18:32:00 +00:00
return f"{self.books_url}?action=by-uris&uris={value}"
2021-04-06 20:39:10 +00:00
def get_book_data(self, remote_id):
data = get_data(remote_id)
extracted = list(data.get("entities").values())
2021-04-06 20:53:58 +00:00
try:
data = extracted[0]
except KeyError:
raise ConnectorException("Invalid book data")
# flatten the data so that images, uri, and claims are on the same level
return {
2021-04-29 17:12:56 +00:00
**data.get("claims", {}),
2021-08-08 22:55:49 +00:00
**{k: data.get(k) for k in ["uri", "image", "labels", "sitelinks", "type"]},
2021-04-06 20:53:58 +00:00
}
2021-06-18 21:12:56 +00:00
def search(self, query, min_confidence=None): # pylint: disable=arguments-differ
"""overrides default search function with confidence ranking"""
results = super().search(query)
if min_confidence:
# filter the search results after the fact
return [r for r in results if r.confidence >= min_confidence]
return results
def parse_search_data(self, data):
2021-04-06 20:53:58 +00:00
return data.get("results")
def format_search_result(self, search_result):
images = search_result.get("image")
2021-09-18 18:33:43 +00:00
cover = f"{self.covers_url}/img/entities/{images[0]}" if images else None
# a deeply messy translation of inventaire's scores
confidence = float(search_result.get("_score", 0.1))
confidence = 0.1 if confidence < 150 else 0.999
return SearchResult(
title=search_result.get("label"),
2021-04-29 16:54:42 +00:00
key=self.get_remote_id(search_result.get("uri")),
author=search_result.get("description"),
2021-09-18 18:32:00 +00:00
view_link=f"{self.base_url}/entity/{search_result.get('uri')}",
cover=cover,
confidence=confidence,
connector=self,
)
def parse_isbn_search_data(self, data):
2021-04-29 16:54:42 +00:00
"""got some daaaata"""
2021-04-29 16:56:35 +00:00
results = data.get("entities")
2021-04-29 16:54:42 +00:00
if not results:
return []
return list(results.values())
def format_isbn_search_result(self, search_result):
2021-04-29 16:54:42 +00:00
"""totally different format than a regular search result"""
title = search_result.get("claims", {}).get("wdt:P1476", [])
if not title:
return None
return SearchResult(
title=title[0],
key=self.get_remote_id(search_result.get("uri")),
author=search_result.get("description"),
2021-09-18 18:32:00 +00:00
view_link=f"{self.base_url}/entity/{search_result.get('uri')}",
2021-04-29 16:54:42 +00:00
cover=self.get_cover_url(search_result.get("image")),
2021-04-29 16:56:35 +00:00
connector=self,
2021-04-29 16:54:42 +00:00
)
def is_work_data(self, data):
return data.get("type") == "work"
2021-04-07 01:34:55 +00:00
def load_edition_data(self, work_uri):
"""get a list of editions for a work"""
2021-09-18 18:33:43 +00:00
url = f"{self.books_url}?action=reverse-claims&property=wdt:P629&value={work_uri}&sort=true"
2021-04-07 01:34:55 +00:00
return get_data(url)
def get_edition_from_work_data(self, data):
data = self.load_edition_data(data.get("uri"))
2021-04-06 20:39:10 +00:00
try:
uri = data.get("uris", [])[0]
except IndexError:
2021-04-06 20:39:10 +00:00
raise ConnectorException("Invalid book data")
return self.get_book_data(self.get_remote_id(uri))
def get_work_from_edition_data(self, data):
2021-04-30 22:48:52 +00:00
uri = data.get("wdt:P629", [None])[0]
if not uri:
2021-04-06 20:39:10 +00:00
raise ConnectorException("Invalid book data")
return self.get_book_data(self.get_remote_id(uri))
def get_authors_from_data(self, data):
2021-04-07 01:51:43 +00:00
authors = data.get("wdt:P50", [])
2021-04-07 00:46:06 +00:00
for author in authors:
yield self.get_or_create_author(self.get_remote_id(author))
def expand_book_data(self, book):
2021-04-07 01:34:55 +00:00
work = book
# go from the edition to the work, if necessary
if isinstance(book, models.Edition):
work = book.parent_work
try:
edition_options = self.load_edition_data(work.inventaire_id)
except ConnectorException:
# who knows, man
return
for edition_uri in edition_options.get("uris"):
remote_id = self.get_remote_id(edition_uri)
try:
data = self.get_book_data(remote_id)
except ConnectorException:
# who, indeed, knows
continue
self.create_edition_from_data(work, data)
def get_cover_url(self, cover_blob, *_):
2021-04-06 20:53:58 +00:00
"""format the relative cover url into an absolute one:
{"url": "/img/entities/e794783f01b9d4f897a1ea9820b96e00d346994f"}
"""
2021-04-29 00:18:14 +00:00
# covers may or may not be a list
2021-04-28 20:50:47 +00:00
if isinstance(cover_blob, list) and len(cover_blob) > 0:
cover_blob = cover_blob[0]
cover_id = cover_blob.get("url")
if not cover_id:
return None
2021-04-29 00:18:14 +00:00
# cover may or may not be an absolute url already
if re.match(r"^http", cover_id):
return cover_id
2021-09-18 18:32:00 +00:00
return f"{self.covers_url}{cover_id}"
2021-04-07 01:10:42 +00:00
def resolve_keys(self, keys):
"""cool, it's "wd:Q3156592" now what the heck does that mean"""
2021-04-07 01:10:42 +00:00
results = []
for uri in keys:
try:
data = self.get_book_data(self.get_remote_id(uri))
except ConnectorException:
continue
results.append(get_language_code(data.get("labels")))
return results
2021-04-29 17:30:38 +00:00
def get_description(self, links):
2021-04-29 17:54:36 +00:00
"""grab an extracted excerpt from wikipedia"""
2021-04-29 17:30:38 +00:00
link = links.get("enwiki")
if not link:
return ""
2021-09-18 18:32:00 +00:00
url = f"{self.base_url}/api/data?action=wp-extract&lang=en&title={link}"
2021-04-29 17:30:38 +00:00
try:
data = get_data(url)
except ConnectorException:
return ""
return data.get("extract")
2021-04-07 00:46:06 +00:00
def get_language_code(options, code="en"):
"""when there are a bunch of translation but we need a single field"""
result = options.get(code)
if result:
return result
values = list(options.values())
return values[0] if values else None