feat: add Typesense backend for search

Also run `make upgrade` to be able to successfully compile dependencies.

Private-ref: https://tasks.opencraft.com/browse/BB-9975

Author: samuelallan72

forum/search/typesense.py

@@ -0,0 +1,370 @@
+"""
+Typesense backend for searching comments and threads.
+"""
+
+from typing import Any, Optional
+
+from bs4 import BeautifulSoup
+from django.conf import settings
+from django.core.paginator import Paginator
+from typesense import Client
+from typesense.types.collection import CollectionCreateSchema
+from typesense.types.document import DocumentSchema, SearchParameters
+
+from forum.backends.mysql.models import Comment, CommentThread
+from forum.constants import FORUM_MAX_DEEP_SEARCH_COMMENT_COUNT
+from forum.search.base import (
+    BaseDocumentSearchBackend,
+    BaseIndexSearchBackend,
+    BaseSearchBackend,
+    BaseThreadSearchBackend,
+)
+
+_TYPESENSE_CLIENT: Client | None = None
+
+
+def get_typesense_client() -> Client:
+    """
+    Return a singleton Typesense client instance.
+    """
+    global _TYPESENSE_CLIENT
+    if _TYPESENSE_CLIENT is None:
+        _TYPESENSE_CLIENT = Client(
+            {
+                "api_key": settings.TYPESENSE_API_KEY,
+                "nodes": [settings.TYPESENSE_URL],
+            }
+        )
+    return _TYPESENSE_CLIENT
+
+
+class CommentsIndex:
+    """
+    Common data and operations relating to the comments index.
+    """
+
+    model = Comment
+
+    @staticmethod
+    def name() -> str:
+        """
+        Return the Typesense index name for the index.
+        """
+        return settings.TYPESENSE_COLLECTION_PREFIX + "comments"
+
+    @classmethod
+    def schema(cls) -> CollectionCreateSchema:
+        return {
+            "name": cls.name(),
+            "fields": [
+                {"name": "course_id", "type": "string"},
+                {"name": "comment_thread_id", "type": "string"},
+                {"name": "body", "type": "string"},
+            ],
+        }
+
+    @staticmethod
+    def build_document(doc_id: str | int, data: dict[str, Any]) -> DocumentSchema:
+        """
+        Build a Typesense document for this index.
+        """
+        # NOTE: Comments have no commentable_id or title, and the context is hardcoded to "course".
+        return {
+            "id": str(doc_id),
+            "course_id": str(data.get("course_id", "")),
+            "comment_thread_id": str(data.get("comment_thread_id", "")),
+            "body": (
+                BeautifulSoup(data["body"], features="html.parser").get_text()
+                if data.get("body")
+                else ""
+            ),
+        }
+
+    @staticmethod
+    def build_search_parameters(
+        *, search_text: str, course_id: str | None
+    ) -> SearchParameters:
+        """
+        Build Typesense search parameters for this index.
+        """
+        return {
+            "q": search_text,
+            "query_by": "body",
+            "filter_by": (
+                f"course_id:={quote_filter_value(course_id)}" if course_id else ""
+            ),
+            "per_page": FORUM_MAX_DEEP_SEARCH_COMMENT_COUNT,
+        }
+
+
+class CommentThreadsIndex:
+    """
+    Common data and operations relating to the comments index.
+    """
+
+    model = CommentThread
+
+    @staticmethod
+    def name() -> str:
+        """
+        Return the Typesense index name for the index.
+        """
+        return settings.TYPESENSE_COLLECTION_PREFIX + "comment_threads"
+
+    @classmethod
+    def schema(cls) -> CollectionCreateSchema:
+        return {
+            "name": cls.name(),
+            "fields": [
+                {"name": "course_id", "type": "string"},
+                {"name": "commentable_id", "type": "string"},
+                {"name": "context", "type": "string"},
+                {"name": "title", "type": "string"},
+                {"name": "body", "type": "string"},
+            ],
+        }
+
+    @staticmethod
+    def build_document(doc_id: str | int, data: dict[str, Any]) -> DocumentSchema:
+        """
+        Build a Typesense document for this index.
+        """
+        return {
+            "id": str(doc_id),
+            "course_id": str(data.get("course_id", "")),
+            "commentable_id": str(data.get("commentable_id", "")),
+            "context": str(data.get("context", "")),
+            "title": str(data.get("title", "")),
+            "body": (
+                BeautifulSoup(data["body"], features="html.parser").get_text()
+                if data.get("body")
+                else ""
+            ),
+        }
+
+    @staticmethod
+    def build_search_parameters(
+        *,
+        search_text: str,
+        course_id: str | None,
+        context: str,
+        commentable_ids: list[str] | None,
+    ) -> SearchParameters:
+        """
+        Build Typesense search parameters for this index.
+        """
+        # Context is always a single word, so we can use the faster `:` operator, without sacrificing accuracy.
+        filters = [f"context:{quote_filter_value(context)}"]
+        if commentable_ids:
+            safe_ids = ", ".join(quote_filter_value(value) for value in commentable_ids)
+            filters.append(f"commentable_ids:[{safe_ids}]")
+        if course_id:
+            filters.append(f"course_id:={quote_filter_value(course_id)}")
+
+        return {
+            "q": search_text,
+            "query_by": "title,body",
+            "filter_by": " && ".join(filters),
+            "per_page": FORUM_MAX_DEEP_SEARCH_COMMENT_COUNT,
+        }
+
+
+INDICES: dict[str, type[CommentsIndex] | type[CommentThreadsIndex]] = {
+    "comments": CommentsIndex,
+    "comment_threads": CommentThreadsIndex,
+}
+
+
+class TypesenseDocumentBackend(BaseDocumentSearchBackend):
+    """
+    Document backend implementation for Typesense.
+    """
+
+    def index_document(
+        self, index_name: str, doc_id: str | int, document: dict[str, Any]
+    ) -> None:
+        """
+        Index a document in Typesense.
+        """
+        client = get_typesense_client()
+        index = INDICES[index_name]
+        typesense_document = index.build_document(doc_id, document)
+        client.collections[index.name()].documents.upsert(typesense_document)
+
+    def update_document(
+        self, index_name: str, doc_id: str | int, update_data: dict[str, Any]
+    ) -> None:
+        """
+        Same operation as index_document, because upsert is used.
+        """
+        return self.index_document(index_name, doc_id, update_data)
+
+    def delete_document(self, index_name: str, doc_id: str | int) -> None:
+        """
+        Delete a document from Typesense.
+        """
+        client = get_typesense_client()
+        index = INDICES[index_name]
+        client.collections[index.name()].documents[str(doc_id)].delete(
+            delete_parameters={"ignore_not_found": True},
+        )
+
+
+class TypesenseIndexBackend(BaseIndexSearchBackend):
+    """
+    Manage indexes for the Typesense backend.
+
+    Typesense calls these "collections". https://typesense.org/docs/29.0/api/collections.html
+    """
+
+    def initialize_indices(self, force_new_index: bool = False) -> None:
+        """
+        Initialize the indices in Typesense.
+
+        If force_new_index is True, the indexes will be dropped before being recreated.
+        """
+        client = get_typesense_client()
+        for index in INDICES.values():
+            if force_new_index:
+                client.collections[index.name()].delete()
+            # XXX: this fails if the index already exists
+            client.collections.create(index.schema())
+
+    def rebuild_indices(
+        self, batch_size: int = 500, extra_catchup_minutes: int = 5
+    ) -> None:
+        """
+        Reindex everything in Typesense
+
+        The Typesense collections are dropped and recreated,
+        and data is reindexed from the MySQL database.
+
+        Only MySQL-backed instances are supported.
+        Note that the `extra_catchup_minutes` argument is ignored.
+        """
+        client = get_typesense_client()
+        self.initialize_indices(force_new_index=True)
+        for index in INDICES.values():
+            paginator = Paginator(index.model.objects.all(), per_page=batch_size)
+            for page_number in paginator.page_range:
+                page = paginator.get_page(page_number)
+                documents = [
+                    index.build_document(obj.pk, obj.doc_to_hash())
+                    for obj in page.object_list
+                ]
+                if documents:
+                    client.collections[index.name()].documents.import_(
+                        documents, {"action": "upsert"}
+                    )
+
+    def validate_indices(self) -> None:
+        """
+        Check if the indices exist and are valid.
+
+        Raise an exception if any do not exist or if any are not valid.
+        """
+        client = get_typesense_client()
+        for index in INDICES.values():
+            collection = client.collections[index.name()].retrieve()
+            # TODO: collection returns more information than the initial create schema,
+            # so we need a better comparison here
+            if collection != index.schema():
+                print(f"Expected schema: {index.schema()}")
+                print(f"Found schema: {collection}")
+                raise AssertionError(f"Collection {index.name()} exists, but schema does not match expected.")
+
+    def refresh_indices(self) -> None:
+        """
+        Noop on Typesense, as all write API operations are synchronous.
+
+        See https://typesense.org/docs/guide/migrating-from-algolia.html#synchronous-write-apis for more information.
+        """
+        return None
+
+    def delete_unused_indices(self) -> int:
+        """
+        Noop on Typesense.
+        """
+        return 0
+
+
+def quote_filter_value(value: str) -> str:
+    """
+    Sanitize and safely quote a value for use in a Typesense filter.
+
+    https://typesense.org/docs/guide/tips-for-filtering.html#escaping-special-characters
+    """
+    # TODO: It may be possible to escape backticks, rather than removing them.
+    #       It also may be possible that there are issues with the backtick method.
+    #       See https://github.com/typesense/typesense/issues/196
+    return "`" + value.replace("`", "") + "`"
+
+
+class TypesenseThreadSearchBackend(BaseThreadSearchBackend):
+    """
+    Thread search backend implementation for Typesense.
+    """
+
+    def get_thread_ids(
+        self,
+        context: str,
+        # This argument is unsupported. Anyway, its only role was to boost some results,
+        # which did not have much effect because they are shuffled anyway downstream.
+        group_ids: list[int],
+        search_text: str,
+        # This parameter is unsupported, but as far as we know it's not used anywhere.
+        sort_criteria: Optional[list[dict[str, str]]] = None,
+        commentable_ids: Optional[list[str]] = None,
+        course_id: Optional[str] = None,
+    ) -> list[str]:
+        """
+        Retrieve thread IDs based on search criteria.
+        """
+        client = get_typesense_client()
+        thread_ids: set[str] = set()
+
+        # All comments have "course" as their context, and none of them have a commentable_id.
+        if context == "course" and not commentable_ids:
+            comment_results = client.collections[CommentsIndex.name()].documents.search(
+                CommentsIndex.build_search_parameters(
+                    search_text=search_text, course_id=course_id
+                )
+            )
+            for hit in comment_results.get("hits", []):
+                thread_ids.add(hit["document"]["comment_thread_id"])
+
+        thread_results = client.collections[
+            CommentThreadsIndex.name()
+        ].documents.search(
+            CommentThreadsIndex.build_search_parameters(
+                search_text=search_text,
+                course_id=course_id,
+                context=context,
+                commentable_ids=commentable_ids,
+            )
+        )
+        for hit in thread_results.get("hits", []):
+            thread_ids.add(hit["document"]["id"])
+
+        return list(thread_ids)
+
+    def get_suggested_text(self, search_text: str) -> Optional[str]:
+        """
+        Retrieve text suggestions for a given search query.
+
+        :param search_text: Text to search for suggestions
+        :return: Suggested text or None
+        """
+        # TODO: https://typesense.org/docs/guide/query-suggestions.html
+        # TODO: if this is implemented, do we need to also implement get_thread_ids_with_corrected_text?
+        return None
+
+
+class TypesenseBackend(BaseSearchBackend):
+    """
+    Typesense-powered search backend.
+    """
+
+    DOCUMENT_SEARCH_CLASS = TypesenseDocumentBackend
+    INDEX_SEARCH_CLASS = TypesenseIndexBackend
+    THREAD_SEARCH_CLASS = TypesenseThreadSearchBackend

forum/settings/common.py

@@ -10,7 +10,13 @@ def plugin_settings(settings: Any) -> None:
     Common settings for forum app
     """
     # Search backend
-    if getattr(settings, "MEILISEARCH_ENABLED", False):
+    if getattr(settings, "TYPESENSE_ENABLED", False):
+        settings.FORUM_SEARCH_BACKEND = getattr(
+            settings,
+            "FORUM_SEARCH_BACKEND",
+            "forum.search.typesense.TypesenseBackend",
+        )
+    elif getattr(settings, "MEILISEARCH_ENABLED", False):
         settings.FORUM_SEARCH_BACKEND = getattr(
             settings,
             "FORUM_SEARCH_BACKEND",

requirements/base.in

@@ -11,4 +11,5 @@ requests
 pymongo
 elasticsearch
 edx-search # meilisearch backend
+typesense
 mysqlclient