neptune/core/lib/opensearch.py

# from copy import deepcopy
# from datetime import datetime, timedelta

from django.conf import settings
from opensearchpy import OpenSearch
from opensearchpy.exceptions import NotFoundError, RequestError

from core.lib.threshold import annotate_num_chans, annotate_num_users, annotate_online
from core.views.helpers import dedup_list

# from json import dumps
# pp = lambda x: print(dumps(x, indent=2))


def initialise_opensearch():
    """
    Inititialise the OpenSearch API endpoint.
    """
    auth = (settings.OPENSEARCH_USERNAME, settings.OPENSEARCH_PASSWORD)
    client = OpenSearch(
        # fmt: off
        hosts=[{"host": settings.OPENSEARCH_URL,
                "port": settings.OPENSEARCH_PORT}],
        http_compress=False,  # enables gzip compression for request bodies
        http_auth=auth,
        # client_cert = client_cert_path,
        # client_key = client_key_path,
        use_ssl=settings.OPENSEARCH_TLS,
        verify_certs=False,
        ssl_assert_hostname=False,
        ssl_show_warn=False,
        # a_certs=ca_certs_path,
    )
    return client


client = initialise_opensearch()


def annotate_results(results_parsed):
    """
    Accept a list of dict objects, search for the number of channels and users.
    Add them to the object.
    Mutate it in place. Does not return anything.
    """
    # Figure out items with net (not discord)
    nets = set()
    for x in results_parsed:
        if "net" in x:
            nets.add(x["net"])

    for net in nets:
        # Annotate the online attribute from Threshold
        nicks = list(
            set(
                [
                    x["nick"]
                    for x in results_parsed
                    if {"nick", "src", "net"}.issubset(x)
                    and x["src"] == "irc"
                    and x["net"] == net
                ]
            )
        )
        channels = list(
            set(
                [
                    x["channel"]
                    for x in results_parsed
                    if {"channel", "src", "net"}.issubset(x)
                    and x["src"] == "irc"
                    and x["net"] == net
                ]
            )
        )
        online_info = annotate_online(net, nicks)
        # Annotate the number of users in the channel
        num_users = annotate_num_users(net, channels)
        # Annotate the number channels the user is on
        num_chans = annotate_num_chans(net, nicks)
        for item in results_parsed:
            if "net" in item:
                if item["net"] == net:
                    if "nick" in item:
                        if item["nick"] in online_info:
                            item["online"] = online_info[item["nick"]]
                    if "channel" in item:
                        if item["channel"] in num_users:
                            item["num_users"] = num_users[item["channel"]]
                    if "nick" in item:
                        if item["nick"] in num_chans:
                            item["num_chans"] = num_chans[item["nick"]]


def filter_blacklisted(user, response):
    """
    Low level filter to take the raw OpenSearch response and remove
    objects from it we want to keep secret.
    Does not return, the object is mutated in place.
    """
    response["redacted"] = 0
    response["exemption"] = None
    if user.is_superuser:
        response["exemption"] = True
    # is_anonymous = isinstance(user, AnonymousUser)
    # For every hit from ES
    for index, item in enumerate(list(response["hits"]["hits"])):
        # For every blacklisted type
        for blacklisted_type in settings.OPENSEARCH_BLACKLISTED.keys():
            # Check this field we are matching exists
            if "_source" in item.keys():
                data_index = "_source"
            elif "fields" in item.keys():
                data_index = "fields"
            else:
                return False
            if blacklisted_type in item[data_index].keys():
                content = item[data_index][blacklisted_type]
                # For every item in the blacklisted array for the type
                for blacklisted_item in settings.OPENSEARCH_BLACKLISTED[
                    blacklisted_type
                ]:
                    if blacklisted_item == str(content):
                        # Remove the item
                        if item in response["hits"]["hits"]:
                            # Let the UI know something was redacted
                            if (
                                "exemption"
                                not in response["hits"]["hits"][index][data_index]
                            ):
                                response["redacted"] += 1
                            # Anonymous
                            if user.is_anonymous:
                                # Just set it to none so the index is not off
                                response["hits"]["hits"][index] = None
                            else:
                                if not user.has_perm("core.bypass_blacklist"):
                                    response["hits"]["hits"][index] = None
                                else:
                                    response["hits"]["hits"][index][data_index][
                                        "exemption"
                                    ] = True

    # Actually get rid of all the things we set to None
    response["hits"]["hits"] = [hit for hit in response["hits"]["hits"] if hit]


def construct_query(query, size, use_query_string=True, tokens=False):
    """
    Accept some query parameters and construct an OpenSearch query.
    """
    if not size:
        size = 5
    query_base = {
        "size": size,
        "query": {"bool": {"must": []}},
    }
    query_string = {
        "query_string": {
            "query": query,
            # "fields": fields,
            # "default_field": "msg",
            # "type": "best_fields",
            "fuzziness": "AUTO",
            "fuzzy_transpositions": True,
            "fuzzy_max_expansions": 50,
            "fuzzy_prefix_length": 0,
            # "minimum_should_match": 1,
            "default_operator": "or",
            "analyzer": "standard",
            "lenient": True,
            "boost": 1,
            "allow_leading_wildcard": True,
            # "enable_position_increments": False,
            "phrase_slop": 3,
            # "max_determinized_states": 10000,
            "quote_field_suffix": "",
            "quote_analyzer": "standard",
            "analyze_wildcard": False,
            "auto_generate_synonyms_phrase_query": True,
        }
    }
    query_tokens = {
        "simple_query_string": {
            # "tokens": query,
            "query": query,
            "fields": ["tokens"],
            "flags": "ALL",
            "fuzzy_transpositions": True,
            "fuzzy_max_expansions": 50,
            "fuzzy_prefix_length": 0,
            "default_operator": "and",
            "analyzer": "standard",
            "lenient": True,
            "boost": 1,
            "quote_field_suffix": "",
            "analyze_wildcard": False,
            "auto_generate_synonyms_phrase_query": False,
        }
    }
    if tokens:
        query_base["query"]["bool"]["must"].append(query_tokens)
        # query["query"]["bool"]["must"].append(query_string)
        # query["query"]["bool"]["must"][0]["query_string"]["fields"] = ["tokens"]
    elif use_query_string:
        query_base["query"]["bool"]["must"].append(query_string)
    return query_base


def run_main_query(client, user, query, custom_query=False, index=None, size=None):
    """
    Low level helper to run an ES query.
    Accept a user to pass it to the filter, so we can
    avoid filtering for superusers.
    Accept fields and size, for the fields we want to match and the
    number of results to return.
    """
    if not index:
        index = settings.OPENSEARCH_INDEX_MAIN
    if custom_query:
        search_query = query
    else:
        search_query = construct_query(query, size)
    try:
        response = client.search(body=search_query, index=index)
    except RequestError as err:
        print("OpenSearch error", err)
        return err
    except NotFoundError as err:
        print("OpenSearch error", err)
        return err
    filter_blacklisted(user, response)
    return response


def parse_results(results):
    results_parsed = []
    stringify = ["host", "channel"]
    if "hits" in results.keys():
        if "hits" in results["hits"]:
            for item in results["hits"]["hits"]:
                if "_source" in item.keys():
                    data_index = "_source"
                elif "fields" in item.keys():
                    data_index = "fields"
                else:
                    return False
                element = item[data_index]
                for field in stringify:
                    if field in element:
                        element[field] = str(element[field])
                # Why are fields in lists...
                if data_index == "fields":
                    element = {k: v[0] for k, v in element.items() if len(v)}
                element["id"] = item["_id"]

                # Split the timestamp into date and time
                if "ts" not in element:
                    if "time" in element:  # will fix data later
                        ts = element["time"]
                        del element["time"]
                        element["ts"] = ts
                if "ts" in element:
                    ts = element["ts"]
                    ts_spl = ts.split("T")
                    date = ts_spl[0]
                    time = ts_spl[1]
                    element["date"] = date
                    if "." in time:
                        time_spl = time.split(".")
                        if len(time_spl) == 2:
                            element["time"] = time.split(".")[0]
                        else:
                            element["time"] = time
                    else:
                        element["time"] = time
                results_parsed.append(element)
    return results_parsed


def query_results(
    request,
    query_params,
    size=None,
    annotate=True,
    custom_query=False,
    reverse=False,
    dedup=False,
    dedup_fields=None,
    lookup_hashes=True,
    tags=None,
):
    """
    API helper to alter the OpenSearch return format into something
    a bit better to parse.
    Accept a HTTP request object. Run the query, and annotate the
    results with the other data we have.
    """
    # is_anonymous = isinstance(request.user, AnonymousUser)
    query = None
    message = None
    message_class = None
    add_bool = []
    add_top = []
    add_top_negative = []
    sort = None
    query_created = False

    # Lookup the hash values but don't disclose them to the user
    # denied = []
    # if lookup_hashes:
    #     if settings.HASHING:
    #         query_params = deepcopy(query_params)
    #         denied_q = hash_lookup(request.user, query_params)
    #         denied.extend(denied_q)
    #         if tags:
    #             denied_t = hash_lookup(request.user, tags, query_params)
    #             denied.extend(denied_t)

    # message = "Permission denied: "
    # for x in denied:
    #     if isinstance(x, SearchDenied):
    #         message += f"Search({x.key}: {x.value}) "
    #     elif isinstance(x, LookupDenied):
    #         message += f"Lookup({x.key}: {x.value}) "
    # if denied:
    #     # message = [f"{i}" for i in message]
    #     # message = "\n".join(message)
    #     message_class = "danger"
    #     return {"message": message, "class": message_class}

    if request.user.is_anonymous:
        sizes = settings.OPENSEARCH_MAIN_SIZES_ANON
    else:
        sizes = settings.OPENSEARCH_MAIN_SIZES
    if not size:
        if "size" in query_params:
            size = query_params["size"]
            if size not in sizes:
                message = "Size is not permitted"
                message_class = "danger"
                return {"message": message, "class": message_class}
        else:
            size = 20
    source = None
    if "source" in query_params:
        source = query_params["source"]

        if source in settings.OPENSEARCH_SOURCES_RESTRICTED:
            if not request.user.has_perm("restricted_sources"):
                message = "Access denied"
                message_class = "danger"
                return {"message": message, "class": message_class}
        elif source not in settings.OPENSEARCH_MAIN_SOURCES:
            message = "Invalid source"
            message_class = "danger"
            return {"message": message, "class": message_class}

        if source == "all":
            source = None  # the next block will populate it

    if source:
        sources = [source]
    else:
        sources = settings.OPENSEARCH_MAIN_SOURCES
        if request.user.has_perm("restricted_sources"):
            for source_iter in settings.OPENSEARCH_SOURCES_RESTRICTED:
                sources.append(source_iter)
        else:
            sources = settings.OPENSEARCH_MAIN_SOURCES

    add_top_tmp = {"bool": {"should": []}}
    for source_iter in sources:
        add_top_tmp["bool"]["should"].append({"match_phrase": {"src": source_iter}})
    add_top.append(add_top_tmp)

    # date_query = False
    if set({"from_date", "to_date", "from_time", "to_time"}).issubset(
        query_params.keys()
    ):
        from_ts = f"{query_params['from_date']}T{query_params['from_time']}Z"
        to_ts = f"{query_params['to_date']}T{query_params['to_time']}Z"
        range_query = {
            "range": {
                "ts": {
                    "gt": from_ts,
                    "lt": to_ts,
                }
            }
        }
        add_top.append(range_query)

    # if date_query:
    #     if settings.DELAY_RESULTS:
    #         if source not in settings.SAFE_SOURCES:
    #             if request.user.has_perm("core.bypass_delay"):
    #                 add_top.append(range_query)
    #             else:
    #                 delay_as_ts = datetime.now() - timedelta(
    #                     days=settings.DELAY_DURATION
    #                 )
    #                 lt_as_ts = datetime.strptime(
    #                     range_query["range"]["ts"]["lt"], "%Y-%m-%dT%H:%MZ"
    #                 )
    #                 if lt_as_ts > delay_as_ts:
    #                     range_query["range"]["ts"][
    #                         "lt"
    #                     ] = f"now-{settings.DELAY_DURATION}d"
    #                 add_top.append(range_query)
    #     else:
    #         add_top.append(range_query)
    # else:
    #     if settings.DELAY_RESULTS:
    #         if source not in settings.SAFE_SOURCES:
    #             if not request.user.has_perm("core.bypass_delay"):
    #                 range_query = {
    #                     "range": {
    #                         "ts": {
    #                             # "gt": ,
    #                             "lt": f"now-{settings.DELAY_DURATION}d",
    #                         }
    #                     }
    #                 }
    #                 add_top.append(range_query)

    if "sorting" in query_params:
        sorting = query_params["sorting"]
        if sorting not in ("asc", "desc", "none"):
            message = "Invalid sort"
            message_class = "danger"
            return {"message": message, "class": message_class}
        if sorting in ("asc", "desc"):
            sort = [
                {
                    "ts": {
                        "order": sorting,
                    }
                }
            ]

    if "check_sentiment" in query_params:
        if "sentiment_method" not in query_params:
            message = "No sentiment method"
            message_class = "danger"
            return {"message": message, "class": message_class}
        if "sentiment" in query_params:
            sentiment = query_params["sentiment"]
            try:
                sentiment = float(sentiment)
            except ValueError:
                message = "Sentiment is not a float"
                message_class = "danger"
                return {"message": message, "class": message_class}
        sentiment_method = query_params["sentiment_method"]
        range_query_compare = {"range": {"sentiment": {}}}
        range_query_precise = {
            "match": {
                "sentiment": None,
            }
        }
        if sentiment_method == "below":
            range_query_compare["range"]["sentiment"]["lt"] = sentiment
            add_top.append(range_query_compare)
        elif sentiment_method == "above":
            range_query_compare["range"]["sentiment"]["gt"] = sentiment
            add_top.append(range_query_compare)
        elif sentiment_method == "exact":
            range_query_precise["match"]["sentiment"] = sentiment
            add_top.append(range_query_precise)
        elif sentiment_method == "nonzero":
            range_query_precise["match"]["sentiment"] = 0
            add_top_negative.append(range_query_precise)

    # Only one of query or query_full can be active at once
    # We prefer query because it's simpler
    if "query" in query_params:
        query = query_params["query"]
        search_query = construct_query(query, size, tokens=True)
        query_created = True
    elif "query_full" in query_params:
        query_full = query_params["query_full"]
        # if request.user.has_perm("core.query_search"):
        search_query = construct_query(query_full, size)
        query_created = True
        # else:
        #     message = "You cannot search by query string"
        #     message_class = "danger"
        #     return {"message": message, "class": message_class}
    else:
        if custom_query:
            search_query = custom_query

    if tags:
        # Get a blank search query
        if not query_created:
            search_query = construct_query(None, size, use_query_string=False)
            query_created = True
        for tagname, tagvalue in tags.items():
            add_bool.append({tagname: tagvalue})

    required_any = ["query_full", "query", "tags"]
    if not any([field in query_params.keys() for field in required_any]):
        if not custom_query:
            message = "Empty query!"
            message_class = "warning"
            return {"message": message, "class": message_class}

    if add_bool:
        # if "bool" not in search_query["query"]:
        #     search_query["query"]["bool"] = {}
        # if "must" not in search_query["query"]["bool"]:
        #     search_query["query"]["bool"] = {"must": []}

        for item in add_bool:
            search_query["query"]["bool"]["must"].append({"match_phrase": item})
    if add_top:
        for item in add_top:
            search_query["query"]["bool"]["must"].append(item)
    if add_top_negative:
        for item in add_top_negative:
            if "must_not" in search_query["query"]["bool"]:
                search_query["query"]["bool"]["must_not"].append(item)
            else:
                search_query["query"]["bool"]["must_not"] = [item]
    if sort:
        search_query["sort"] = sort

    if "index" in query_params:
        index = query_params["index"]
        if index == "main":
            index = settings.OPENSEARCH_INDEX_MAIN
        else:
            if not request.user.has_perm(f"core.index_{index}"):
                message = "Not permitted to search by this index"
                message_class = "danger"
                return {
                    "message": message,
                    "class": message_class,
                }
            if index == "meta":
                index = settings.OPENSEARCH_INDEX_META
            elif index == "int":
                index = settings.OPENSEARCH_INDEX_INT
            else:
                message = "Index is not valid."
                message_class = "danger"
                return {
                    "message": message,
                    "class": message_class,
                }

    else:
        index = settings.OPENSEARCH_INDEX_MAIN

    results = run_main_query(
        client,
        request.user,  # passed through run_main_query to filter_blacklisted
        search_query,
        custom_query=True,
        index=index,
        size=size,
    )
    if not results:
        return False
    if isinstance(results, Exception):
        message = f"Error: {results.info['error']['root_cause'][0]['type']}"
        message_class = "danger"
        return {"message": message, "class": message_class}
    if len(results["hits"]["hits"]) == 0:
        message = "No results."
        message_class = "danger"
        return {"message": message, "class": message_class}

    results_parsed = parse_results(results)

    if annotate:
        annotate_results(results_parsed)
    if "dedup" in query_params:
        if query_params["dedup"] == "on":
            dedup = True
        else:
            dedup = False
    else:
        dedup = False

    if reverse:
        results_parsed = results_parsed[::-1]

    if dedup:
        if not dedup_fields:
            dedup_fields = ["msg", "nick", "ident", "host", "net", "channel"]
        results_parsed = dedup_list(results_parsed, dedup_fields)

    # if source not in settings.SAFE_SOURCES:
    #     if settings.ENCRYPTION:
    #         encrypt_list(request.user, results_parsed, settings.ENCRYPTION_KEY)

    #     if settings.HASHING:
    #         hash_list(request.user, results_parsed)

    #     if settings.OBFUSCATION:
    #         obfuscate_list(request.user, results_parsed)

    #     if settings.RANDOMISATION:
    #         randomise_list(request.user, results_parsed)

    # process_list(results)

    # IMPORTANT! - DO NOT PASS query_params to the user!
    context = {
        "object_list": results_parsed,
        "card": results["hits"]["total"]["value"],
        "took": results["took"],
    }
    if "redacted" in results:
        context["redacted"] = results["redacted"]
    if "exemption" in results:
        context["exemption"] = results["exemption"]
    if query:
        context["query"] = query
    # if settings.DELAY_RESULTS:
    #     if source not in settings.SAFE_SOURCES:
    #         if not request.user.has_perm("core.bypass_delay"):
    #             context["delay"] = settings.DELAY_DURATION
    # if settings.RANDOMISATION:
    #     if source not in settings.SAFE_SOURCES:
    #         if not request.user.has_perm("core.bypass_randomisation"):
    #             context["randomised"] = True
    return context


def query_single_result(request, query_params):
    context = query_results(request, query_params, size=100)

    if not context:
        return {"message": "Failed to run query", "message_class": "danger"}
    if "message" in context:
        return context
    dedup_set = {item["nick"] for item in context["object_list"]}
    if dedup_set:
        context["item"] = context["object_list"][0]

    return context