def _parse_game(game):
for field in DEDUPE_FIELDS:
game.setdefault(field["field"], None)
if field["type"] == "Set":
game[field["field"]] = tuple(arg_to_iter(
game[field["field"]])) or None
game["names"] = tuple(
clear_list(
chain(arg_to_iter(game.get("name")),
arg_to_iter(game.get("alt_name")))))
for field in VALUE_ID_FIELDS:
game[field] = tuple(
clear_list(map(_parse_value_id, arg_to_iter(game.get(field)))))
return game
def parse(self, response):
"""
@url https://boardgamegeek.com/browse/boardgame/
@returns items 0 0
@returns requests 11
"""
next_page = response.xpath(
'//a[@title = "next page"]/@href').extract_first()
if next_page:
yield Request(
response.urljoin(next_page),
callback=self.parse,
priority=1,
meta={"max_retry_times": 10},
)
urls = response.xpath("//@href").extract()
bgg_ids = filter(None, map(extract_bgg_id, map(response.urljoin,
urls)))
yield from self._game_requests(*bgg_ids)
user_names = filter(None, map(extract_bgg_user_name, urls))
scraped_at = now()
for user_name in clear_list(user_names):
yield self.collection_request(
user_name
) if self.scrape_collections else self._user_item_or_request(
user_name, scraped_at=scraped_at)
def process_item(self, item, spider):
""" resolve resource image URLs to actual file locations """
for field in self.fields:
if item.get(field):
item[field] = clear_list(
map(self._parse_url, arg_to_iter(item[field])))
return item
def dates(self, request):
"""Find all available dates with rankings."""
query_set = self.get_queryset().order_by("ranking_type", "date")
ranking_types = clear_list(_extract_params(request, "ranking_type"))
if ranking_types:
query_set = query_set.filter(ranking_type__in=ranking_types)
return Response(query_set.values("ranking_type", "date").distinct())
def extract_ids(*urls: Optional[str]) -> Dict[str, List[Union[int, str]]]:
""" extract all possible IDs from all the URLs """
urls = tuple(map(urlparse, urls))
return {
"bgg_id": clear_list(map(extract_bgg_id, urls)),
"freebase_id": clear_list(map(extract_freebase_id, urls)),
"wikidata_id": clear_list(map(extract_wikidata_id, urls)),
"wikipedia_id": clear_list(map(extract_wikipedia_id, urls)),
"dbpedia_id": clear_list(map(extract_dbpedia_id, urls)),
"luding_id": clear_list(map(extract_luding_id, urls)),
"spielen_id": clear_list(map(extract_spielen_id, urls)),
"bga_id": clear_list(map(extract_bga_id, urls)),
}
def _process_df(data_frame,
columns=None,
required_columns=None,
target_column=None):
if data_frame is None or data_frame.empty:
LOGGER.error("DataFrame is empty")
return None
columns = clear_list(arg_to_iter(columns))
required_columns = clear_list(arg_to_iter(required_columns)) or columns
columns = clear_list(columns + required_columns)
if not columns:
LOGGER.error("No columns given")
return None
missing_columns = [
column for column in required_columns if column not in data_frame
]
if missing_columns:
LOGGER.error("DataFrame does not contain the expected columns %s",
missing_columns)
return None
for column in columns:
if column not in data_frame:
data_frame[column] = None
target_column = target_column or columns[0]
return (data_frame[columns][data_frame[target_column].notna()].sort_values(
target_column).rename(columns={
"bayes_rating": "score"
}).astype({
"rank": int,
"bgg_id": int
}))
def _extract_labels(self, response, value):
json_obj = parse_json(response.text) if hasattr(response, "text") else {}
labels = take_first(jmespath.search(f"entities.{value}.labels", json_obj)) or {}
labels = labels.values()
labels = sorted(
labels,
key=lambda label: self.lang_priorities.get(label.get("language"), math.inf),
)
labels = clear_list(label.get("value") for label in labels)
self.labels[value] = labels
self.logger.debug("resolved labels for %s: %s", value, labels)
return labels
def _game_requests(self,
*bgg_ids,
batch_size=10,
page=1,
priority=0,
**kwargs):
bgg_ids = clear_list(map(parse_int, bgg_ids))
if not bgg_ids:
return
bgg_ids = ((bgg_id for bgg_id in bgg_ids
if bgg_id not in self._ids_seen) if page == 1 else bgg_ids)
for batch in batchify(bgg_ids, batch_size):
batch = tuple(batch)
ids = ",".join(map(str, batch))
url = (self._api_url(
action="thing",
id=ids,
stats=1,
videos=1,
versions=int(self.scrape_ratings),
ratingcomments=int(self.scrape_ratings),
page=1,
) if page == 1 else self._api_url(action="thing",
id=ids,
versions=1,
ratingcomments=1,
page=page))
request = Request(url, callback=self.parse_game, priority=priority)
if len(batch) == 1:
request.meta["bgg_id"] = batch[0]
request.meta["page"] = page
request.meta.update(kwargs)
yield request
if page == 1:
self._ids_seen.update(batch)
def rankings(self, request, pk=None):
"""Find historical rankings of a game."""
filters = {
"game": pk,
"ranking_type__in": clear_list(_extract_params(request, "ranking_type")),
"date__gte": parse_date(
request.query_params.get("date__gte"), tzinfo=timezone.utc
),
"date__lte": parse_date(
request.query_params.get("date__lte"), tzinfo=timezone.utc
),
}
filters = {k: v for k, v in filters.items() if v}
queryset = Ranking.objects.filter(**filters)
serializer = RankingSerializer(
queryset, many=True, context=self.get_serializer_context()
)
return Response(serializer.data)
def _add_value(self, result, field, item):
labels = clear_list(flatten(r[1] for r in arg_to_iter(result))) or None
self.logger.debug("resolved labels for %s: %s", item.get(field),
labels)
item[field] = labels
return item
def merge_files(
in_paths,
out_path,
keys="id",
key_types=None,
latest=None,
latest_types=None,
latest_min=None,
latest_required=False,
fieldnames=None,
fieldnames_exclude=None,
sort_keys=False,
sort_latest=False,
sort_fields=None,
sort_descending=False,
concat_output=False,
log_level=None,
):
""" merge files into one """
spark = _spark_session(log_level=log_level)
if spark is None:
raise RuntimeError(
"Please make sure Spark is installed and configured correctly!")
in_paths = list(map(str, arg_to_iter(in_paths)))
LOGGER.info(
"Merging items from %s into <%s> with Spark session %r",
f"[{len(in_paths) } paths]" if len(in_paths) > 10 else in_paths,
out_path,
spark,
)
fieldnames = clear_list(arg_to_iter(fieldnames))
fieldnames_exclude = frozenset(arg_to_iter(fieldnames_exclude))
if fieldnames and fieldnames_exclude:
LOGGER.warning(
"Both <fieldnames> and <fieldnames_exclude> were specified, please choose one"
)
sort_fields = tuple(arg_to_iter(sort_fields))
if sum(map(bool, (sort_keys, sort_latest, sort_fields))) > 1:
LOGGER.warning(
"Only use at most one of <sort_keys>, <sort_latest>, and <sort_fields>"
)
keys = tuple(arg_to_iter(keys))
key_types = tuple(arg_to_iter(key_types))
key_types += (None, ) * (len(keys) - len(key_types))
assert len(keys) == len(key_types)
LOGGER.info("Using keys %s with types %s", keys, key_types)
latest = tuple(arg_to_iter(latest))
latest_types = tuple(arg_to_iter(latest_types))
latest_types += (None, ) * (len(latest) - len(latest_types))
assert len(latest) == len(latest_types)
LOGGER.info("Using latest %s with types %s", latest, latest_types)
data = spark.read.json(path=in_paths,
mode="DROPMALFORMED",
dropFieldIfAllNull=True)
key_column_names = [f"_key_{i}" for i in range(len(keys))]
key_columns = [
_column_type(data[column], column_type).alias(name)
for column, column_type, name in zip(keys, key_types, key_column_names)
]
key_columns_str = (column.cast("string") for column in key_columns)
latest_column_names = [f"_latest_{i}" for i in range(len(latest))]
latest_columns = [
_column_type(data[column], column_type).alias(name) for column,
column_type, name in zip(latest, latest_types, latest_column_names)
]
latest_columns_str = (column.cast("string") for column in latest_columns)
drop_subset = keys + tuple(key_column_names)
if latest_required:
drop_subset += latest + tuple(latest_column_names)
LOGGER.info("Dropping rows without values in columns %s", drop_subset)
data = data.select(
"*",
*key_columns,
array(*key_columns_str).alias("_key"),
*latest_columns,
array(*latest_columns_str).alias("_latest"),
).dropna(how="any", subset=drop_subset)
if latest_min is not None:
LOGGER.info("Filter out items before %s", latest_min)
data = data.filter(latest_columns[0] >= latest_min)
rdd = (data.rdd.keyBy(lambda row: tuple(arg_to_iter(row["_key"]))).
reduceByKey(_compare).values())
data = rdd.toDF(schema=data.schema)
if sort_keys:
LOGGER.info(
"Sorting %s by columns %s",
"descending" if sort_descending else "ascending",
keys,
)
data = data.sort(*key_column_names, ascending=not sort_descending)
elif sort_latest:
LOGGER.info(
"Sorting %s by columns %s",
"descending" if sort_descending else "ascending",
latest,
)
data = data.sort(*latest_column_names, ascending=not sort_descending)
elif sort_fields:
LOGGER.info(
"Sorting %s by columns %s",
"descending" if sort_descending else "ascending",
sort_fields,
)
data = data.sort(*sort_fields, ascending=not sort_descending)
data = data.drop("_key", *key_column_names, "_latest",
*latest_column_names)
columns = frozenset(data.columns) - fieldnames_exclude
if fieldnames:
fieldnames = [column for column in fieldnames if column in columns]
LOGGER.info("Only use columns: %s", fieldnames)
else:
fieldnames = sorted(columns)
LOGGER.info("Use sorted column names: %s", fieldnames)
data = data.select(*fieldnames)
data = _remove_empty(data)
if concat_output:
with tempfile.TemporaryDirectory() as temp_path:
path = Path(temp_path) / "out"
LOGGER.info("Saving temporary output to <%s>", path)
data.write.json(path=str(path))
LOGGER.info("Concatenate temporary files to <%s>", out_path)
files = path.glob("part-*")
concat_files(dst=out_path, srcs=sorted(files), ensure_newline=True)
else:
LOGGER.info("Saving output to <%s>", out_path)
data.write.json(path=str(out_path))
LOGGER.info("Done merging.")
