class OpenSearchDataStore(object):
"""Implements the datastore."""
# Number of events to queue up when bulk inserting events.
DEFAULT_FLUSH_INTERVAL = 1000
DEFAULT_SIZE = 100
DEFAULT_LIMIT = DEFAULT_SIZE # Max events to return
DEFAULT_FROM = 0
DEFAULT_STREAM_LIMIT = 5000 # Max events to return when streaming results
DEFAULT_FLUSH_RETRY_LIMIT = 3 # Max retries for flushing the queue.
DEFAULT_EVENT_IMPORT_TIMEOUT = "3m" # Timeout value for importing events.
def __init__(self, host="127.0.0.1", port=9200):
"""Create a OpenSearch client."""
super().__init__()
self._error_container = {}
self.user = current_app.config.get("OPENSEARCH_USER", "user")
self.password = current_app.config.get("OPENSEARCH_PASSWORD", "pass")
self.ssl = current_app.config.get("OPENSEARCH_SSL", False)
self.verify = current_app.config.get("OPENSEARCH_VERIFY_CERTS", True)
self.timeout = current_app.config.get("OPENSEARCH_TIMEOUT", 10)
parameters = {}
if self.ssl:
parameters["use_ssl"] = self.ssl
parameters["verify_certs"] = self.verify
if self.user and self.password:
parameters["http_auth"] = (self.user, self.password)
if self.timeout:
parameters["timeout"] = self.timeout
self.client = OpenSearch([{"host": host, "port": port}], **parameters)
self.import_counter = Counter()
self.import_events = []
self._request_timeout = current_app.config.get(
"TIMEOUT_FOR_EVENT_IMPORT", self.DEFAULT_EVENT_IMPORT_TIMEOUT)
@staticmethod
def _build_labels_query(sketch_id, labels):
"""Build OpenSearch query for Timesketch labels.
Args:
sketch_id: Integer of sketch primary key.
labels: List of label names.
Returns:
OpenSearch query as a dictionary.
"""
label_query = {"bool": {"must": []}}
for label in labels:
# Increase metrics counter per label
METRICS["search_filter_label"].labels(label=label).inc()
nested_query = {
"nested": {
"query": {
"bool": {
"must": [
{
"term": {
"timesketch_label.name.keyword": label
}
},
{
"term": {
"timesketch_label.sketch_id": sketch_id
}
},
]
}
},
"path": "timesketch_label",
}
}
label_query["bool"]["must"].append(nested_query)
return label_query
@staticmethod
def _build_events_query(events):
"""Build OpenSearch query for one or more document ids.
Args:
events: List of OpenSearch document IDs.
Returns:
OpenSearch query as a dictionary.
"""
events_list = [event["event_id"] for event in events]
query_dict = {"query": {"ids": {"values": events_list}}}
return query_dict
@staticmethod
def _build_query_dsl(query_dsl, timeline_ids):
"""Build OpenSearch Search DSL query by adding in timeline filtering.
Args:
query_dsl: A dict with the current query_dsl
timeline_ids: Either a list of timeline IDs (int) or None.
Returns:
OpenSearch query DSL as a dictionary.
"""
# Remove any aggregation coming from user supplied Query DSL.
# We have no way to display this data in a good way today.
if query_dsl.get("aggregations", None):
del query_dsl["aggregations"]
if not timeline_ids:
return query_dsl
if not isinstance(timeline_ids, (list, tuple)):
es_logger.error(
"Attempting to pass in timelines to a query DSL, but the "
"passed timelines are not a list.")
return query_dsl
if not all([isinstance(x, int) for x in timeline_ids]):
es_logger.error("All timeline IDs need to be an integer.")
return query_dsl
old_query = query_dsl.get("query")
if not old_query:
return query_dsl
query_dsl["query"] = {
"bool": {
"must": [],
"should": [
{
"bool": {
"must":
old_query,
"must_not": [{
"exists": {
"field": "__ts_timeline_id"
},
}],
}
},
{
"bool": {
"must": [
{
"terms": {
"__ts_timeline_id": timeline_ids
}
},
old_query,
],
"must_not": [],
"filter": [{
"exists": {
"field": "__ts_timeline_id"
}
}],
}
},
],
"must_not": [],
"filter": [],
}
}
return query_dsl
@staticmethod
def _convert_to_time_range(interval):
"""Convert an interval timestamp into start and end dates.
Args:
interval: Time frame representation
Returns:
Start timestamp in string format.
End timestamp in string format.
"""
# return ('2018-12-05T00:00:00', '2018-12-05T23:59:59')
TS_FORMAT = "%Y-%m-%dT%H:%M:%S"
get_digits = lambda s: int("".join(filter(str.isdigit, s)))
get_alpha = lambda s: "".join(filter(str.isalpha, s))
ts_parts = interval.split(" ")
# The start date could be 1 or 2 first items
start = " ".join(ts_parts[0:len(ts_parts) - 2])
minus = get_digits(ts_parts[-2])
plus = get_digits(ts_parts[-1])
interval = get_alpha(ts_parts[-1])
start_ts = parser.parse(start)
rd = relativedelta.relativedelta
if interval == "s":
start_range = start_ts - rd(seconds=minus)
end_range = start_ts + rd(seconds=plus)
elif interval == "m":
start_range = start_ts - rd(minutes=minus)
end_range = start_ts + rd(minutes=plus)
elif interval == "h":
start_range = start_ts - rd(hours=minus)
end_range = start_ts + rd(hours=plus)
elif interval == "d":
start_range = start_ts - rd(days=minus)
end_range = start_ts + rd(days=plus)
else:
raise RuntimeError("Unable to parse the timestamp: " +
str(interval))
return start_range.strftime(TS_FORMAT), end_range.strftime(TS_FORMAT)
def build_query(
self,
sketch_id,
query_string,
query_filter,
query_dsl=None,
aggregations=None,
timeline_ids=None,
):
"""Build OpenSearch DSL query.
Args:
sketch_id: Integer of sketch primary key
query_string: Query string
query_filter: Dictionary containing filters to apply
query_dsl: Dictionary containing OpenSearch DSL query
aggregations: Dict of OpenSearch aggregations
timeline_ids: Optional list of IDs of Timeline objects that should
be queried as part of the search.
Returns:
OpenSearch DSL query as a dictionary
"""
if query_dsl:
if not isinstance(query_dsl, dict):
query_dsl = json.loads(query_dsl)
if not query_dsl:
query_dsl = {}
return self._build_query_dsl(query_dsl, timeline_ids)
if query_filter.get("events", None):
events = query_filter["events"]
return self._build_events_query(events)
query_dsl = {
"query": {
"bool": {
"must": [],
"must_not": [],
"filter": []
}
}
}
if query_string:
query_dsl["query"]["bool"]["must"].append({
"query_string": {
"query": query_string,
"default_operator": "AND"
}
})
# New UI filters
if query_filter.get("chips", None):
labels = []
must_filters = query_dsl["query"]["bool"]["must"]
must_not_filters = query_dsl["query"]["bool"]["must_not"]
datetime_ranges = {
"bool": {
"should": [],
"minimum_should_match": 1
}
}
for chip in query_filter["chips"]:
# Exclude chips that the user disabled
if not chip.get("active", True):
continue
# Increase metrics per chip type
METRICS["search_filter_type"].labels(type=chip["type"]).inc()
if chip["type"] == "label":
labels.append(chip["value"])
elif chip["type"] == "term":
term_filter = {
"match_phrase": {
"{}".format(chip["field"]): {
"query": "{}".format(chip["value"])
}
}
}
if chip["operator"] == "must":
must_filters.append(term_filter)
elif chip["operator"] == "must_not":
must_not_filters.append(term_filter)
elif chip["type"].startswith("datetime"):
range_filter = lambda start, end: {
"range": {
"datetime": {
"gte": start,
"lte": end
}
}
}
if chip["type"] == "datetime_range":
start, end = chip["value"].split(",")
elif chip["type"] == "datetime_interval":
start, end = self._convert_to_time_range(chip["value"])
else:
continue
datetime_ranges["bool"]["should"].append(
range_filter(start, end))
label_filter = self._build_labels_query(sketch_id, labels)
must_filters.append(label_filter)
must_filters.append(datetime_ranges)
# Pagination
if query_filter.get("from", None):
query_dsl["from"] = query_filter["from"]
# Number of events to return
if query_filter.get("size", None):
query_dsl["size"] = query_filter["size"]
# Make sure we are sorting.
if not query_dsl.get("sort", None):
query_dsl["sort"] = {"datetime": query_filter.get("order", "asc")}
# Add any pre defined aggregations
if aggregations:
# post_filter happens after aggregation so we need to move the
# filter to the query instead.
if query_dsl.get("post_filter", None):
query_dsl["query"]["bool"]["filter"] = query_dsl["post_filter"]
query_dsl.pop("post_filter", None)
query_dsl["aggregations"] = aggregations
# TODO: Simplify this when we don't have to support both timelines
# that have __ts_timeline_id set and those that don't.
# (query_string AND timeline_id NOT EXISTS) OR (
# query_string AND timeline_id in LIST)
if timeline_ids and isinstance(timeline_ids, (list, tuple)):
must_filters_pre = copy.copy(query_dsl["query"]["bool"]["must"])
must_not_filters_pre = copy.copy(
query_dsl["query"]["bool"]["must_not"])
must_filters_post = copy.copy(query_dsl["query"]["bool"]["must"])
must_not_filters_post = copy.copy(
query_dsl["query"]["bool"]["must_not"])
must_not_filters_pre.append({
"exists": {
"field": "__ts_timeline_id"
},
})
must_filters_post.append(
{"terms": {
"__ts_timeline_id": timeline_ids
}})
query_dsl["query"] = {
"bool": {
"must": [],
"should": [
{
"bool": {
"must": must_filters_pre,
"must_not": must_not_filters_pre,
}
},
{
"bool": {
"must":
must_filters_post,
"must_not":
must_not_filters_post,
"filter": [{
"exists": {
"field": "__ts_timeline_id"
}
}],
}
},
],
"must_not": [],
"filter": [],
}
}
return query_dsl
# pylint: disable=too-many-arguments
def search(
self,
sketch_id,
query_string,
query_filter,
query_dsl,
indices,
count=False,
aggregations=None,
return_fields=None,
enable_scroll=False,
timeline_ids=None,
):
"""Search OpenSearch. This will take a query string from the UI
together with a filter definition. Based on this it will execute the
search request on OpenSearch and get result back.
Args:
sketch_id: Integer of sketch primary key
query_string: Query string
query_filter: Dictionary containing filters to apply
query_dsl: Dictionary containing OpenSearch DSL query
indices: List of indices to query
count: Boolean indicating if we should only return result count
aggregations: Dict of OpenSearch aggregations
return_fields: List of fields to return
enable_scroll: If OpenSearch scroll API should be used
timeline_ids: Optional list of IDs of Timeline objects that should
be queried as part of the search.
Returns:
Set of event documents in JSON format
"""
scroll_timeout = None
if enable_scroll:
scroll_timeout = "1m" # Default to 1 minute scroll timeout
# Exit early if we have no indices to query
if not indices:
return {"hits": {"hits": [], "total": 0}, "took": 0}
# Make sure that the list of index names is uniq.
indices = list(set(indices))
# Check if we have specific events to fetch and get indices.
if query_filter.get("events", None):
indices = {
event["index"]
for event in query_filter["events"]
if event["index"] in indices
}
query_dsl = self.build_query(
sketch_id=sketch_id,
query_string=query_string,
query_filter=query_filter,
query_dsl=query_dsl,
aggregations=aggregations,
timeline_ids=timeline_ids,
)
# Default search type for OpenSearch is query_then_fetch.
search_type = "query_then_fetch"
# Only return how many documents matches the query.
if count:
if "sort" in query_dsl:
del query_dsl["sort"]
try:
count_result = self.client.count(body=query_dsl,
index=list(indices))
except NotFoundError:
es_logger.error(
"Unable to count due to an index not found: {0:s}".format(
",".join(indices)))
return 0
METRICS["search_requests"].labels(type="count").inc()
return count_result.get("count", 0)
if not return_fields:
# Suppress the lint error because opensearchpy adds parameters
# to the function with a decorator and this makes pylint sad.
# pylint: disable=unexpected-keyword-arg
return self.client.search(
body=query_dsl,
index=list(indices),
search_type=search_type,
scroll=scroll_timeout,
)
# The argument " _source_include" changed to "_source_includes" in
# ES version 7. This check add support for both version 6 and 7 clients.
# pylint: disable=unexpected-keyword-arg
try:
if self.version.startswith("6"):
_search_result = self.client.search(
body=query_dsl,
index=list(indices),
search_type=search_type,
_source_include=return_fields,
scroll=scroll_timeout,
)
else:
_search_result = self.client.search(
body=query_dsl,
index=list(indices),
search_type=search_type,
_source_includes=return_fields,
scroll=scroll_timeout,
)
except RequestError as e:
root_cause = e.info.get("error", {}).get("root_cause")
if root_cause:
error_items = []
for cause in root_cause:
error_items.append("[{0:s}] {1:s}".format(
cause.get("type", ""), cause.get("reason", "")))
cause = ", ".join(error_items)
else:
cause = str(e)
es_logger.error("Unable to run search query: {0:s}".format(cause),
exc_info=True)
raise ValueError(cause) from e
METRICS["search_requests"].labels(type="single").inc()
return _search_result
# pylint: disable=too-many-arguments
def search_stream(
self,
sketch_id=None,
query_string=None,
query_filter=None,
query_dsl=None,
indices=None,
return_fields=None,
enable_scroll=True,
timeline_ids=None,
):
"""Search OpenSearch. This will take a query string from the UI
together with a filter definition. Based on this it will execute the
search request on OpenSearch and get result back.
Args :
sketch_id: Integer of sketch primary key
query_string: Query string
query_filter: Dictionary containing filters to apply
query_dsl: Dictionary containing OpenSearch DSL query
indices: List of indices to query
return_fields: List of fields to return
enable_scroll: Boolean determining whether scrolling is enabled.
timeline_ids: Optional list of IDs of Timeline objects that should
be queried as part of the search.
Returns:
Generator of event documents in JSON format
"""
# Make sure that the list of index names is uniq.
indices = list(set(indices))
METRICS["search_requests"].labels(type="stream").inc()
if not query_filter.get("size"):
query_filter["size"] = self.DEFAULT_STREAM_LIMIT
if not query_filter.get("terminate_after"):
query_filter["terminate_after"] = self.DEFAULT_STREAM_LIMIT
result = self.search(
sketch_id=sketch_id,
query_string=query_string,
query_dsl=query_dsl,
query_filter=query_filter,
indices=indices,
return_fields=return_fields,
enable_scroll=enable_scroll,
timeline_ids=timeline_ids,
)
if enable_scroll:
scroll_id = result["_scroll_id"]
scroll_size = result["hits"]["total"]
else:
scroll_id = None
scroll_size = 0
# Elasticsearch version 7.x returns total hits as a dictionary.
# TODO: Refactor when version 6.x has been deprecated.
if isinstance(scroll_size, dict):
scroll_size = scroll_size.get("value", 0)
for event in result["hits"]["hits"]:
yield event
while scroll_size > 0:
# pylint: disable=unexpected-keyword-arg
result = self.client.scroll(scroll_id=scroll_id, scroll="5m")
scroll_id = result["_scroll_id"]
scroll_size = len(result["hits"]["hits"])
for event in result["hits"]["hits"]:
yield event
def get_filter_labels(self, sketch_id, indices):
"""Aggregate labels for a sketch.
Args:
sketch_id: The Sketch ID
indices: List of indices to aggregate on
Returns:
List with label names.
"""
# This is a workaround to return all labels by setting the max buckets
# to something big. If a sketch has more than this amount of labels
# the list will be incomplete but it should be uncommon to have >10k
# labels in a sketch.
max_labels = 10000
# pylint: disable=line-too-long
aggregation = {
"aggs": {
"nested": {
"nested": {
"path": "timesketch_label"
},
"aggs": {
"inner": {
"filter": {
"bool": {
"must": [{
"term": {
"timesketch_label.sketch_id":
sketch_id
}
}]
}
},
"aggs": {
"labels": {
"terms": {
"size": max_labels,
"field":
"timesketch_label.name.keyword",
}
}
},
}
},
}
}
}
# Make sure that the list of index names is uniq.
indices = list(set(indices))
labels = []
# pylint: disable=unexpected-keyword-arg
try:
result = self.client.search(index=indices,
body=aggregation,
size=0)
except NotFoundError:
es_logger.error("Unable to find the index/indices: {0:s}".format(
",".join(indices)))
return labels
buckets = (result.get("aggregations",
{}).get("nested",
{}).get("inner",
{}).get("labels",
{}).get("buckets", []))
for bucket in buckets:
new_bucket = {}
new_bucket["label"] = bucket.pop("key")
new_bucket["count"] = bucket.pop("doc_count")
labels.append(new_bucket)
return labels
# pylint: disable=inconsistent-return-statements
def get_event(self, searchindex_id, event_id):
"""Get one event from the datastore.
Args:
searchindex_id: String of OpenSearch index id
event_id: String of OpenSearch event id
Returns:
Event document in JSON format
"""
METRICS["search_get_event"].inc()
try:
# Suppress the lint error because opensearchpy adds parameters
# to the function with a decorator and this makes pylint sad.
# pylint: disable=unexpected-keyword-arg
if self.version.startswith("6"):
event = self.client.get(
index=searchindex_id,
id=event_id,
doc_type="_all",
_source_exclude=["timesketch_label"],
)
else:
event = self.client.get(
index=searchindex_id,
id=event_id,
doc_type="_all",
_source_excludes=["timesketch_label"],
)
return event
except NotFoundError:
abort(HTTP_STATUS_CODE_NOT_FOUND)
def count(self, indices):
"""Count number of documents.
Args:
indices: List of indices.
Returns:
Tuple containing number of documents and size on disk.
"""
if not indices:
return 0, 0
# Make sure that the list of index names is uniq.
indices = list(set(indices))
try:
es_stats = self.client.indices.stats(index=indices,
metric="docs, store")
except NotFoundError:
es_logger.error("Unable to count indices (index not found)")
return 0, 0
except RequestError:
es_logger.error("Unable to count indices (request error)",
exc_info=True)
return 0, 0
doc_count_total = (es_stats.get("_all",
{}).get("primaries",
{}).get("docs",
{}).get("count", 0))
doc_bytes_total = (es_stats.get("_all", {}).get("primaries", {}).get(
"store", {}).get("size_in_bytes", 0))
return doc_count_total, doc_bytes_total
def set_label(
self,
searchindex_id,
event_id,
event_type,
sketch_id,
user_id,
label,
toggle=False,
remove=False,
single_update=True,
):
"""Set label on event in the datastore.
Args:
searchindex_id: String of OpenSearch index id
event_id: String of OpenSearch event id
event_type: String of OpenSearch document type
sketch_id: Integer of sketch primary key
user_id: Integer of user primary key
label: String with the name of the label
remove: Optional boolean value if the label should be removed
toggle: Optional boolean value if the label should be toggled
single_update: Boolean if the label should be indexed immediately.
Returns:
Dict with updated document body, or None if this is a single update.
"""
# OpenSearch painless script.
update_body = {
"script": {
"lang": "painless",
"source": UPDATE_LABEL_SCRIPT,
"params": {
"timesketch_label": {
"name": str(label),
"user_id": user_id,
"sketch_id": sketch_id,
},
remove: remove,
},
}
}
if toggle:
update_body["script"]["source"] = TOGGLE_LABEL_SCRIPT
if not single_update:
script = update_body["script"]
return dict(source=script["source"],
lang=script["lang"],
params=script["params"])
doc = self.client.get(index=searchindex_id,
id=event_id,
doc_type="_all")
try:
doc["_source"]["timesketch_label"]
except KeyError:
doc = {"doc": {"timesketch_label": []}}
self.client.update(index=searchindex_id,
doc_type=event_type,
id=event_id,
body=doc)
self.client.update(index=searchindex_id,
id=event_id,
doc_type=event_type,
body=update_body)
return None
def create_index(self,
index_name=uuid4().hex,
doc_type="generic_event",
mappings=None):
"""Create index with Timesketch settings.
Args:
index_name: Name of the index. Default is a generated UUID.
doc_type: Name of the document type. Default id generic_event.
mappings: Optional dict with the document mapping for OpenSearch.
Returns:
Index name in string format.
Document type in string format.
"""
if mappings:
_document_mapping = mappings
else:
_document_mapping = {
"properties": {
"timesketch_label": {
"type": "nested"
},
"datetime": {
"type": "date"
},
}
}
# TODO: Remove when we deprecate OpenSearch version 6.x
if self.version.startswith("6"):
_document_mapping = {doc_type: _document_mapping}
if not self.client.indices.exists(index_name):
try:
self.client.indices.create(
index=index_name, body={"mappings": _document_mapping})
except ConnectionError as e:
raise RuntimeError(
"Unable to connect to Timesketch backend.") from e
except RequestError:
index_exists = self.client.indices.exists(index_name)
es_logger.warning(
"Attempting to create an index that already exists "
"({0:s} - {1:s})".format(index_name, str(index_exists)))
return index_name, doc_type
def delete_index(self, index_name):
"""Delete OpenSearch index.
Args:
index_name: Name of the index to delete.
"""
if self.client.indices.exists(index_name):
try:
self.client.indices.delete(index=index_name)
except ConnectionError as e:
raise RuntimeError(
"Unable to connect to Timesketch backend: {}".format(
e)) from e
def import_event(
self,
index_name,
event_type,
event=None,
event_id=None,
flush_interval=DEFAULT_FLUSH_INTERVAL,
timeline_id=None,
):
"""Add event to OpenSearch.
Args:
index_name: Name of the index in OpenSearch
event_type: Type of event (e.g. plaso_event)
event: Event dictionary
event_id: Event OpenSearch ID
flush_interval: Number of events to queue up before indexing
timeline_id: Optional ID number of a Timeline object this event
belongs to. If supplied an additional field will be added to
the store indicating the timeline this belongs to.
"""
if event:
for k, v in event.items():
if not isinstance(k, six.text_type):
k = codecs.decode(k, "utf8")
# Make sure we have decoded strings in the event dict.
if isinstance(v, six.binary_type):
v = codecs.decode(v, "utf8")
event[k] = v
# Header needed by OpenSearch when bulk inserting.
header = {
"index": {
"_index": index_name,
}
}
update_header = {"update": {"_index": index_name, "_id": event_id}}
# TODO: Remove when we deprecate Elasticsearch version 6.x
if self.version.startswith("6"):
header["index"]["_type"] = event_type
update_header["update"]["_type"] = event_type
if event_id:
# Event has "lang" defined if there is a script used for import.
if event.get("lang"):
event = {"script": event}
else:
event = {"doc": event}
header = update_header
if timeline_id:
event["__ts_timeline_id"] = timeline_id
self.import_events.append(header)
self.import_events.append(event)
self.import_counter["events"] += 1
if self.import_counter["events"] % int(flush_interval) == 0:
_ = self.flush_queued_events()
self.import_events = []
else:
# Import the remaining events in the queue.
if self.import_events:
_ = self.flush_queued_events()
return self.import_counter["events"]
def flush_queued_events(self, retry_count=0):
"""Flush all queued events.
Returns:
dict: A dict object that contains the number of events
that were sent to OpenSearch as well as information
on whether there were any errors, and what the
details of these errors if any.
retry_count: optional int indicating whether this is a retry.
"""
if not self.import_events:
return {}
return_dict = {
"number_of_events": len(self.import_events) / 2,
"total_events": self.import_counter["events"],
}
try:
# pylint: disable=unexpected-keyword-arg
results = self.client.bulk(body=self.import_events,
timeout=self._request_timeout)
except (ConnectionTimeout, socket.timeout):
if retry_count >= self.DEFAULT_FLUSH_RETRY_LIMIT:
es_logger.error("Unable to add events, reached recount max.",
exc_info=True)
return {}
es_logger.error("Unable to add events (retry {0:d}/{1:d})".format(
retry_count, self.DEFAULT_FLUSH_RETRY_LIMIT))
return self.flush_queued_events(retry_count + 1)
errors_in_upload = results.get("errors", False)
return_dict["errors_in_upload"] = errors_in_upload
if errors_in_upload:
items = results.get("items", [])
return_dict["errors"] = []
es_logger.error("Errors while attempting to upload events.")
for item in items:
index = item.get("index", {})
index_name = index.get("_index", "N/A")
_ = self._error_container.setdefault(index_name, {
"errors": [],
"types": Counter(),
"details": Counter()
})
error_counter = self._error_container[index_name]["types"]
error_detail_counter = self._error_container[index_name][
"details"]
error_list = self._error_container[index_name]["errors"]
error = index.get("error", {})
status_code = index.get("status", 0)
doc_id = index.get("_id", "(unable to get doc id)")
caused_by = error.get("caused_by", {})
caused_reason = caused_by.get("reason",
"Unkown Detailed Reason")
error_counter[error.get("type")] += 1
detail_msg = "{0:s}/{1:s}".format(
caused_by.get("type", "Unknown Detailed Type"),
" ".join(caused_reason.split()[:5]),
)
error_detail_counter[detail_msg] += 1
error_msg = "<{0:s}> {1:s} [{2:s}/{3:s}]".format(
error.get("type", "Unknown Type"),
error.get("reason", "No reason given"),
caused_by.get("type", "Unknown Type"),
caused_reason,
)
error_list.append(error_msg)
try:
es_logger.error(
"Unable to upload document: {0:s} to index {1:s} - "
"[{2:d}] {3:s}".format(doc_id, index_name, status_code,
error_msg))
# We need to catch all exceptions here, since this is a crucial
# call that we do not want to break operation.
except Exception: # pylint: disable=broad-except
es_logger.error(
"Unable to upload document, and unable to log the "
"error itself.",
exc_info=True,
)
return_dict["error_container"] = self._error_container
self.import_events = []
return return_dict
@property
def version(self):
"""Get OpenSearch version.
Returns:
Version number as a string.
"""
version_info = self.client.info().get("version")
return version_info.get("number")
class SearchApiClient:
def __init__(self, host=settings.OPENSEARCH_HOST):
protocol = settings.OPENSEARCH_PROTOCOL
protocol_config = {}
if protocol == "https":
protocol_config = {
"scheme": "https",
"port": 443,
"use_ssl": True,
"verify_certs": settings.OPENSEARCH_VERIFY_CERTS,
}
if settings.IS_AWS:
http_auth = ("supersurf", settings.OPENSEARCH_PASSWORD)
else:
http_auth = (None, None)
self.client = OpenSearch([host],
http_auth=http_auth,
connection_class=RequestsHttpConnection,
**protocol_config)
self.index_nl = settings.OPENSEARCH_NL_INDEX
self.index_en = settings.OPENSEARCH_EN_INDEX
self.index_unk = settings.OPENSEARCH_UNK_INDEX
self.languages = {"nl": self.index_nl, "en": self.index_en}
@staticmethod
def parse_search_result(search_result):
"""
Parses the search result into the correct format that the frontend uses
:param search_result: result from search
:return result: list of results ready for frontend
"""
hits = search_result.pop("hits")
aggregations = search_result.get("aggregations", {})
result = dict()
result['recordcount'] = hits['total']['value']
# Transform aggregations into drilldowns
drilldowns = {}
for aggregation_name, aggregation in aggregations.items():
buckets = aggregation["filtered"][
"buckets"] if "filtered" in aggregation else aggregation[
"buckets"]
for bucket in buckets:
drilldowns[f"{aggregation_name}-{bucket['key']}"] = bucket[
"doc_count"]
result['drilldowns'] = drilldowns
# Parse spelling suggestions
did_you_mean = {}
if 'suggest' in search_result:
spelling_suggestion = search_result['suggest'][
'did-you-mean-suggestion'][0]
spelling_option = spelling_suggestion['options'][0] if len(
spelling_suggestion['options']) else None
if spelling_option is not None and spelling_option["score"] >= 0.01:
did_you_mean = {
'original': spelling_suggestion['text'],
'suggestion': spelling_option['text']
}
result['did_you_mean'] = did_you_mean
# Transform hits into records
result['records'] = [
SearchApiClient.parse_search_hit(hit) for hit in hits['hits']
]
return result
@staticmethod
def parse_search_hit(hit, transform=True):
"""
Parses the search hit into the format that is also used by the edurep endpoint.
It's mostly just mapping the variables we need into the places that we expect them to be.
:param hit: result from search
:return record: parsed record
"""
data = hit["_source"]
serializer = SearchResultSerializer()
# Basic mapping between field and data (excluding any method fields with a source of "*")
field_mapping = {
field.source: field_name if transform else field.source
for field_name, field in serializer.fields.items()
if field.source != "*"
}
record = {
field_mapping[field]: value
for field, value in data.items() if field in field_mapping
}
# Reformatting some fields if a relations field is desired
if "relations" in field_mapping:
publishers = [{
"name": publisher
} for publisher in data.get("publishers", [])]
record["relations"] = {
"authors":
data.get("authors", []),
"parties":
data.get("parties", []) or publishers,
"projects":
data.get("projects", []),
"keywords": [{
"label": keyword
} for keyword in data.get("keywords", [])],
"themes": [{
"label": theme
} for theme in data.get("research_themes", [])],
"parents":
data.get("is_part_of", []),
"children":
data.get("has_parts", [])
}
# Calling methods on serializers to set data for method fields
for field_name, field in serializer.fields.items():
if field.source != "*":
continue
record[field_name] = getattr(serializer, field.method_name)(data)
# Add highlight to the record
if hit.get("highlight", 0):
record["highlight"] = hit["highlight"]
return record
def autocomplete(self, query):
"""
Use the suggest query to get typing hints during searching.
:param query: the input from the user so far
:return: a list of options matching the input query, sorted by length
"""
# build the query for search engine
query_dictionary = {
'suggest': {
"autocomplete": {
'text': query,
"completion": {
"field": "suggest_completion",
"size": 100
}
}
}
}
result = self.client.search(
index=[self.index_nl, self.index_en, self.index_unk],
body=query_dictionary)
# extract the options from the search result, remove duplicates,
# remove non-matching prefixes (engine will suggest things that don't match _exactly_)
# and sort by length
autocomplete = result['suggest']['autocomplete']
options = autocomplete[0]['options']
flat_options = list(
set([
item for option in options
for item in option['_source']['suggest_completion']
]))
options_with_prefix = [
option for option in flat_options
if option.lower().startswith(query.lower())
]
options_with_prefix.sort(key=lambda option: len(option))
return options_with_prefix
def drilldowns(self, drilldown_names, search_text=None, filters=None):
"""
This function is named drilldowns is because it's also named drilldowns in the original edurep search code.
It passes on information to search, and returns the search without the records.
This allows calculation of 'item counts' (i.e. how many results there are in through a certain filter)
"""
search_results = self.search(search_text=search_text,
filters=filters,
drilldown_names=drilldown_names)
search_results["records"] = []
return search_results
def search(self,
search_text,
drilldown_names=None,
filters=None,
ordering=None,
page=1,
page_size=5):
"""
Build and send a query to search engine and parse it before returning.
:param search_text: A list of strings to search for.
:param drilldown_names: A list of the 'drilldowns' (filters) that are to be counted by engine.
:param filters: The filters that are applied for this search.
:param ordering: Sort the results by this ordering (or use default search ordering otherwise)
:param page: The page index of the results
:param page_size: How many items are loaded per page.
:return:
"""
start_record = page_size * (page - 1)
body = {
'query': {
"bool": defaultdict(list)
},
'from': start_record,
'size': page_size,
'post_filter': {
"bool": defaultdict(list)
},
'highlight': {
'number_of_fragments': 1,
'fragment_size': 120,
'fields': {
'description': {},
'text': {}
}
}
}
if search_text:
query_string = {
"simple_query_string": {
"fields": SEARCH_FIELDS,
"query": search_text,
"default_operator": "and"
}
}
body["query"]["bool"]["must"] += [query_string]
body["query"]["bool"]["should"] = {
"distance_feature": {
"field": "publisher_date",
"pivot": "90d",
"origin": "now",
"boost": 1.15
}
}
body["suggest"] = {
'did-you-mean-suggestion': {
'text': search_text,
'phrase': {
'field':
'suggest_phrase',
'size':
1,
'gram_size':
3,
'direct_generator': [{
'field': 'suggest_phrase',
'suggest_mode': 'always'
}],
},
}
}
indices = self.parse_index_language(self, filters)
if drilldown_names:
body["aggs"] = self.parse_aggregations(drilldown_names, filters)
filters = self.parse_filters(filters)
if filters:
body["post_filter"]["bool"]["must"] += filters
if ordering:
body["sort"] = [self.parse_ordering(ordering), "_score"]
# make query and parse
result = self.client.search(index=indices, body=body)
return self.parse_search_result(result)
def get_materials_by_id(self,
external_ids,
page=1,
page_size=10,
**kwargs):
"""
Retrieve specific materials from search engine through their external id.
:param external_ids: the id's of the materials to retrieve
:param page: The page index of the results
:param page_size: How many items are loaded per page.
:return: a list of search results (like a regular search).
"""
start_record = page_size * (page - 1)
normalized_external_ids = []
for external_id in external_ids:
if not external_id.startswith("surf"):
normalized_external_ids.append(external_id)
else:
external_id_parts = external_id.split(":")
normalized_external_ids.append(external_id_parts[-1])
result = self.client.search(
index=[self.index_nl, self.index_en, self.index_unk],
body={
"query": {
"bool": {
"must": [{
"terms": {
"external_id": normalized_external_ids
}
}]
}
},
'from': start_record,
'size': page_size,
},
)
results = self.parse_search_result(result)
materials = {
material["external_id"]: material
for material in results["records"]
}
records = []
for external_id in normalized_external_ids:
if external_id not in materials:
continue
records.append(materials[external_id])
results["recordcount"] = len(records)
results["records"] = records
return results
def stats(self):
stats = self.client.count(
index=",".join([self.index_nl, self.index_en, self.index_unk]))
return stats.get("count", 0)
def more_like_this(self, external_id, language):
index = self.languages.get(language, self.index_unk)
body = {
"query": {
"more_like_this": {
"fields": ["title", "description"],
"like": [{
"_index": index,
"_id": external_id
}],
"min_term_freq": 1,
"max_query_terms": 12
}
}
}
search_result = self.client.search(index=index, body=body)
hits = search_result.pop("hits")
result = dict()
result["records_total"] = hits["total"]["value"]
result["results"] = [
SearchApiClient.parse_search_hit(hit, transform=False)
for hit in hits["hits"]
]
return result
def author_suggestions(self, author_name):
body = {
"query": {
"bool": {
"must": {
"multi_match": {
"fields": [
field for field in SEARCH_FIELDS
if "authors" not in field
],
"query":
author_name,
},
},
"must_not": {
"match": {
"authors.name.folded": author_name
}
}
}
}
}
search_result = self.client.search(
index=[self.index_nl, self.index_en, self.index_unk], body=body)
hits = search_result.pop("hits")
result = dict()
result["records_total"] = hits["total"]["value"]
result["results"] = [
SearchApiClient.parse_search_hit(hit, transform=False)
for hit in hits["hits"]
]
return result
@staticmethod
def parse_filters(filters):
"""
Parse filters from the frontend format into the search engine format.
Not every filter is handled by search engine in the same way so it's a lot of manual parsing.
:param filters: the list of filters to be parsed
:return: the filters in the format for a search query.
"""
if not filters:
return {}
filter_items = []
for filter_item in filters:
# skip filter_items that are empty
# and the language filter item (it's handled by telling search engine in what index to search).
if not filter_item['items'] or 'language.keyword' in filter_item[
'external_id']:
continue
search_type = filter_item['external_id']
# date range query
if search_type == "publisher_date":
lower_bound, upper_bound = filter_item["items"]
if lower_bound is not None or upper_bound is not None:
filter_items.append({
"range": {
"publisher_date": {
"gte": lower_bound,
"lte": upper_bound
}
}
})
# all other filter types are handled by just using terms with the 'translated' filter items
else:
filter_items.append(
{"terms": {
search_type: filter_item["items"]
}})
return filter_items
def parse_aggregations(self, aggregation_names, filters):
"""
Parse the aggregations so search engine can count the items properly.
:param aggregation_names: the names of the aggregations to
:param filters: the filters for the query
:return:
"""
aggregation_items = {}
for aggregation_name in aggregation_names:
other_filters = []
if filters:
other_filters = list(
filter(lambda x: x['external_id'] != aggregation_name,
filters))
other_filters = self.parse_filters(other_filters)
search_type = aggregation_name
if len(other_filters) > 0:
# Filter the aggregation by the filters applied to other categories
aggregation_items[aggregation_name] = {
"filter": {
"bool": {
"must": other_filters
}
},
"aggs": {
"filtered": {
"terms": {
"field": search_type,
"size": 2000,
}
}
},
}
else:
aggregation_items[aggregation_name] = {
"terms": {
"field": search_type,
"size": 2000,
}
}
return aggregation_items
@staticmethod
def parse_ordering(ordering):
"""
Parse the frontend ordering format ('asc', 'desc' or None) into the type that search engine expects.
"""
order = "asc"
if ordering.startswith("-"):
order = "desc"
ordering = ordering[1:]
search_type = ordering
return {search_type: {"order": order}}
@staticmethod
def parse_index_language(self, filters):
"""
Select the index to search on based on language.
"""
# if no language is selected, search on both.
indices = [self.index_nl, self.index_en, self.index_unk]
if not filters:
return indices
language_item = [
filter_item for filter_item in filters
if filter_item['external_id'] == 'language.keyword'
]
if not language_item:
return indices
language_indices = [
f"latest-{language}" for language in language_item[0]['items']
]
return language_indices if len(language_indices) else indices