nucliadb 6.7.2.post4862__py3-none-any.whl → 6.9.2.post5282__py3-none-any.whl

nucliadb/common/catalog/interface.py

@@ -0,0 +1,85 @@
+# Copyright (C) 2021 Bosutech XXI S.L.
+#
+# nucliadb is offered under the AGPL v3.0 and as commercial software.
+# For commercial licensing, contact us at info@nuclia.com.
+#
+# AGPL:
+# This program is free software: you can redistribute it and/or modify
+# it under the terms of the GNU Affero General Public License as
+# published by the Free Software Foundation, either version 3 of the
+# License, or (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+# GNU Affero General Public License for more details.
+#
+# You should have received a copy of the GNU Affero General Public License
+# along with this program. If not, see <http://www.gnu.org/licenses/>.
+#
+from __future__ import annotations
+
+import abc
+import datetime
+from dataclasses import dataclass
+from typing import Literal, Optional, Union
+
+from pydantic import BaseModel, Field
+
+from nucliadb.common.maindb.driver import Transaction
+from nucliadb_models import search as search_models
+from nucliadb_models.search import CatalogFacetsRequest, Resources
+
+
+class CatalogResourceData(BaseModel):
+    """
+    Data extracted from a resource to be indexed in the catalog
+    """
+
+    title: str = Field(description="Resource title")
+    created_at: datetime.datetime = Field(description="Resource creation date")
+    modified_at: datetime.datetime = Field(description="Resource last modification date")
+    labels: list[str] = Field(
+        description="Resource labels. This includes labels at the resource level and all classification labels of its fields"
+    )
+    slug: str = Field(description="Resource slug")
+
+
+@dataclass
+class CatalogExpression:
+    @dataclass
+    class Date:
+        field: Union[Literal["created_at"], Literal["modified_at"]]
+        since: Optional[datetime.datetime]
+        until: Optional[datetime.datetime]
+
+    bool_and: Optional[list["CatalogExpression"]] = None
+    bool_or: Optional[list["CatalogExpression"]] = None
+    bool_not: Optional["CatalogExpression"] = None
+    date: Optional[Date] = None
+    facet: Optional[str] = None
+    resource_id: Optional[str] = None
+
+
+class CatalogQuery(BaseModel):
+    kbid: str
+    query: Optional[search_models.CatalogQuery] = Field(description="Full-text search query")
+    filters: Optional[CatalogExpression] = Field(description="Filters to apply to the search")
+    sort: search_models.SortOptions = Field(description="Sorting option")
+    faceted: list[str] = Field(description="List of facets to compute during the search")
+    page_size: int = Field(description="Used for pagination. Maximum page size is 100")
+    page_number: int = Field(description="Used for pagination. First page is 0")
+
+
+class Catalog(abc.ABC, metaclass=abc.ABCMeta):
+    @abc.abstractmethod
+    async def update(self, txn: Transaction, kbid: str, rid: str, data: CatalogResourceData): ...
+
+    @abc.abstractmethod
+    async def delete(self, txn: Transaction, kbid: str, rid: str): ...
+
+    @abc.abstractmethod
+    async def search(self, query: CatalogQuery) -> Resources: ...
+
+    @abc.abstractmethod
+    async def facets(self, kbid: str, request: CatalogFacetsRequest) -> dict[str, int]: ...

nucliadb/{search/search/pgcatalog.py → common/catalog/pg.py}

@@ -26,111 +26,180 @@ from typing import Any, Literal, Union, cast
 from psycopg import AsyncCursor, sql
 from psycopg.rows import DictRow, dict_row
 
-from nucliadb.common.maindb.pg import PGDriver
+from nucliadb.common.catalog.interface import (
+    Catalog,
+    CatalogExpression,
+    CatalogQuery,
+    CatalogResourceData,
+)
+from nucliadb.common.exceptions import InvalidQueryError
+from nucliadb.common.maindb.driver import Transaction
+from nucliadb.common.maindb.pg import PGDriver, PGTransaction
 from nucliadb.common.maindb.utils import get_driver
-from nucliadb.search.search.query_parser.models import CatalogExpression, CatalogQuery
 from nucliadb_models import search as search_models
-from nucliadb_models.labels import translate_system_to_alias_label
-from nucliadb_models.search import CatalogFacetsRequest, ResourceResult, Resources, SortField, SortOrder
+from nucliadb_models.labels import translate_alias_to_system_label, translate_system_to_alias_label
+from nucliadb_models.search import (
+    CatalogFacetsRequest,
+    ResourceResult,
+    Resources,
+    SortField,
+    SortOrder,
+)
 from nucliadb_telemetry import metrics
 
-from .filters import translate_label
+write_observer = metrics.Observer("pg_catalog_write", labels={"type": ""})
+search_observer = metrics.Observer("pg_catalog_search", labels={"op": ""})
 
-observer = metrics.Observer("pg_catalog_search", labels={"op": ""})
 logger = logging.getLogger(__name__)
 
 SPLIT_REGEX = re.compile(r"\W")
 
 
-def _filter_operands(operands: list[CatalogExpression]) -> tuple[list[str], list[CatalogExpression]]:
-    facets = []
-    nonfacets = []
-    for op in operands:
-        if op.facet:
-            facets.append(op.facet)
-        else:
-            nonfacets.append(op)
+def _pg_transaction(txn: Transaction) -> PGTransaction:
+    return cast(PGTransaction, txn)
 
-    return facets, nonfacets
 
-
-def _convert_filter(expr: CatalogExpression, filter_params: dict[str, Any]) -> sql.Composable:
-    if expr.bool_and:
-        return _convert_boolean_op(expr.bool_and, "and", filter_params)
-    elif expr.bool_or:
-        return _convert_boolean_op(expr.bool_or, "or", filter_params)
-    elif expr.bool_not:
-        return sql.SQL("(NOT {})").format(_convert_filter(expr.bool_not, filter_params))
-    elif expr.date:
-        return _convert_date_filter(expr.date, filter_params)
-    elif expr.facet:
-        param_name = f"param{len(filter_params)}"
-        filter_params[param_name] = [expr.facet]
-        if expr.facet == "/n/s/PROCESSED":
-            # Optimization for the most common case, we know PROCESSED is a full label and can use the smaller labels index
-            # This is needed because PROCESSED is present in most catalog entries and PG is unlikely to use any index
-            # for it, falling back to executing the extract_facets function which can be slow
-            return sql.SQL("labels @> {}").format(sql.Placeholder(param_name))
-        else:
-            return sql.SQL("extract_facets(labels) @> {}").format(sql.Placeholder(param_name))
-    elif expr.resource_id:
-        param_name = f"param{len(filter_params)}"
-        filter_params[param_name] = [expr.resource_id]
-        return sql.SQL("rid = {}").format(sql.Placeholder(param_name))
-    else:
-        return sql.SQL("")
+def _pg_driver() -> PGDriver:
+    return cast(PGDriver, get_driver())
 
 
-def _convert_boolean_op(
-    operands: list[CatalogExpression],
-    op: Union[Literal["and"], Literal["or"]],
-    filter_params: dict[str, Any],
-) -> sql.Composable:
-    array_op = sql.SQL("@>" if op == "and" else "&&")
-    operands_sql: list[sql.Composable] = []
-    facets, nonfacets = _filter_operands(operands)
-    if facets:
-        param_name = f"param{len(filter_params)}"
-        if facets == ["/n/s/PROCESSED"]:
-            # Optimization for the most common case, we know PROCESSED is a full label and can use the smaller labels index
-            # This is needed because PROCESSED is present in most catalog entries and PG is unlikely to use any index
-            # for it, falling back to executing the extract_facets function which can be slow
-            operands_sql.append(sql.SQL("labels @> {}").format(sql.Placeholder(param_name)))
-        else:
-            operands_sql.append(
-                sql.SQL("extract_facets(labels) {} {}").format(array_op, sql.Placeholder(param_name))
+class PGCatalog(Catalog):
+    @write_observer.wrap({"type": "update"})
+    async def update(self, txn: Transaction, kbid: str, rid: str, data: CatalogResourceData):
+        async with _pg_transaction(txn).connection.cursor() as cur:
+            await cur.execute(
+                """
+                INSERT INTO catalog
+                (kbid, rid, title, created_at, modified_at, labels, slug)
+                VALUES
+                (%(kbid)s, %(rid)s, %(title)s, %(created_at)s, %(modified_at)s, %(labels)s, %(slug)s)
+                ON CONFLICT (kbid, rid) DO UPDATE SET
+                title = excluded.title,
+                created_at = excluded.created_at,
+                modified_at = excluded.modified_at,
+                labels = excluded.labels,
+                slug = excluded.slug""",
+                {
+                    "kbid": kbid,
+                    "rid": rid,
+                    "title": data.title,
+                    "created_at": data.created_at,
+                    "modified_at": data.modified_at,
+                    "labels": data.labels,
+                    "slug": data.slug,
+                },
+            )
+            await cur.execute(
+                "DELETE FROM catalog_facets WHERE kbid = %(kbid)s AND rid = %(rid)s",
+                {
+                    "kbid": kbid,
+                    "rid": rid,
+                },
+            )
+            await cur.execute(
+                "INSERT INTO catalog_facets (kbid, rid, facet) SELECT %(kbid)s AS kbid, %(rid)s AS rid, unnest(%(facets)s::text[]) AS facet",
+                {
+                    "kbid": kbid,
+                    "rid": rid,
+                    "facets": list(extract_facets(data.labels)),
+                },
             )
-        filter_params[param_name] = facets
-    for nonfacet in nonfacets:
-        operands_sql.append(_convert_filter(nonfacet, filter_params))
-    return sql.SQL("({})").format(sql.SQL(f" {op.upper()} ").join(operands_sql))
 
+    @write_observer.wrap({"type": "delete"})
+    async def delete(self, txn: Transaction, kbid: str, rid: str):
+        async with _pg_transaction(txn).connection.cursor() as cur:
+            await cur.execute(
+                "DELETE FROM catalog where kbid = %(kbid)s AND rid = %(rid)s", {"kbid": kbid, "rid": rid}
+            )
 
-def _convert_date_filter(date: CatalogExpression.Date, filter_params: dict[str, Any]) -> sql.Composable:
-    if date.since and date.until:
-        since_name = f"param{len(filter_params)}"
-        filter_params[since_name] = date.since
-        until_name = f"param{len(filter_params)}"
-        filter_params[until_name] = date.until
-        return sql.SQL("{field} BETWEEN {since} AND {until}").format(
-            field=sql.Identifier(date.field),
-            since=sql.Placeholder(since_name),
-            until=sql.Placeholder(until_name),
-        )
-    elif date.since:
-        since_name = f"param{len(filter_params)}"
-        filter_params[since_name] = date.since
-        return sql.SQL("{field} > {since}").format(
-            field=sql.Identifier(date.field), since=sql.Placeholder(since_name)
-        )
-    elif date.until:
-        until_name = f"param{len(filter_params)}"
-        filter_params[until_name] = date.until
-        return sql.SQL("{field} < {until}").format(
-            field=sql.Identifier(date.field), until=sql.Placeholder(until_name)
+    @search_observer.wrap({"op": "search"})
+    async def search(self, catalog_query: CatalogQuery) -> Resources:
+        # Prepare SQL query
+        query, query_params = _prepare_query_filters(catalog_query)
+
+        async with _pg_driver()._get_connection() as conn, conn.cursor(row_factory=dict_row) as cur:
+            facets = {}
+
+            # Faceted search
+            if catalog_query.faceted:
+                with search_observer({"op": "facets"}):
+                    tmp_facets: dict[str, dict[str, int]] = {
+                        translate_label(f): defaultdict(int) for f in catalog_query.faceted
+                    }
+
+                    if catalog_query.filters is None:
+                        await _faceted_search_unfiltered(cur, catalog_query, tmp_facets)
+                    else:
+                        await _faceted_search_filtered(
+                            cur, catalog_query, tmp_facets, query, query_params
+                        )
+
+                    facets = {translate_system_to_alias_label(k): v for k, v in tmp_facets.items()}
+
+            # Totals
+            with search_observer({"op": "totals"}):
+                await cur.execute(
+                    sql.SQL("SELECT COUNT(*) FROM ({}) fc").format(query),
+                    query_params,
+                )
+                total = (await cur.fetchone())["count"]  # type: ignore
+
+            # Query
+            with search_observer({"op": "query"}):
+                query, query_params = _prepare_query(catalog_query)
+                await cur.execute(query, query_params)
+                data = await cur.fetchall()
+
+        return Resources(
+            facets=facets,
+            results=[
+                ResourceResult(
+                    rid=str(r["rid"]).replace("-", ""),
+                    field="title",
+                    field_type="a",
+                    labels=[label for label in r["labels"] if label.startswith("/l/")],
+                    score=0,
+                )
+                for r in data
+            ],
+            query=catalog_query.query.query if catalog_query.query else "",
+            total=total,
+            page_number=catalog_query.page_number,
+            page_size=catalog_query.page_size,
+            next_page=(catalog_query.page_size * catalog_query.page_number + len(data) < total),
+            min_score=0,
         )
-    else:
-        raise ValueError(f"Invalid date operator")
+
+    @search_observer.wrap({"op": "catalog_facets"})
+    async def facets(self, kbid: str, request: CatalogFacetsRequest) -> dict[str, int]:
+        async with _pg_driver()._get_connection() as conn, conn.cursor() as cur:
+            prefix_filters: list[sql.Composable] = []
+            prefix_params: dict[str, Any] = {}
+            for cnt, prefix in enumerate(request.prefixes):
+                prefix_sql = sql.SQL("facet LIKE {}").format(sql.Placeholder(f"prefix{cnt}"))
+                prefix_params[f"prefix{cnt}"] = f"{prefix.prefix}%"
+                if prefix.depth is not None:
+                    prefix_parts = len(prefix.prefix.split("/"))
+                    depth_sql = sql.SQL("SPLIT_PART(facet, '/', {}) = ''").format(
+                        sql.Placeholder(f"depth{cnt}")
+                    )
+                    prefix_params[f"depth{cnt}"] = prefix_parts + prefix.depth + 1
+                    prefix_sql = sql.SQL("({} AND {})").format(prefix_sql, depth_sql)
+                prefix_filters.append(prefix_sql)
+
+            filter_sql: sql.Composable
+            if prefix_filters:
+                filter_sql = sql.SQL("AND {}").format(sql.SQL(" OR ").join(prefix_filters))
+            else:
+                filter_sql = sql.SQL("")
+
+            await cur.execute(
+                sql.SQL(
+                    "SELECT facet, COUNT(*) FROM catalog_facets WHERE kbid = %(kbid)s {} GROUP BY facet"
+                ).format(filter_sql),
+                {"kbid": kbid, **prefix_params},
+            )
+            return {k: v for k, v in await cur.fetchall()}
 
 
 def _prepare_query_filters(catalog_query: CatalogQuery) -> tuple[sql.Composable, dict[str, Any]]:
@@ -149,42 +218,16 @@ def _prepare_query_filters(catalog_query: CatalogQuery) -> tuple[sql.Composable,
     )
 
 
-def _prepare_query_search(query: search_models.CatalogQuery, params: dict[str, Any]) -> sql.Composable:
-    if query.match == search_models.CatalogQueryMatch.Exact:
-        params["query"] = query.query
-        return sql.SQL("{} = %(query)s").format(sql.Identifier(query.field.value))
-    elif query.match == search_models.CatalogQueryMatch.StartsWith:
-        params["query"] = query.query + "%"
-        if query.field == search_models.CatalogQueryField.Title:
-            # Insensitive search supported by pg_trgm for title
-            return sql.SQL("{} ILIKE %(query)s").format(sql.Identifier(query.field.value))
+def _filter_operands(operands: list[CatalogExpression]) -> tuple[list[str], list[CatalogExpression]]:
+    facets = []
+    nonfacets = []
+    for op in operands:
+        if op.facet:
+            facets.append(op.facet)
         else:
-            # Sensitive search for slug (btree does not support ILIKE and slugs are all lowercase anyway)
-            return sql.SQL("{} LIKE %(query)s").format(sql.Identifier(query.field.value))
-    # The rest of operators only supported by title
-    elif query.match == search_models.CatalogQueryMatch.Words:
-        # This is doing tokenization inside the SQL server (to keep the index updated). We could move it to
-        # the python code at update/query time if it ever becomes a problem but for now, a single regex
-        # executed per query is not a problem.
+            nonfacets.append(op)
 
-        # Remove zero-length words from the split
-        params["query"] = [word.lower() for word in SPLIT_REGEX.split(query.query) if word]
-        return sql.SQL("regexp_split_to_array(lower(title), '\\W') @> %(query)s")
-    elif query.match == search_models.CatalogQueryMatch.Fuzzy:
-        params["query"] = query.query
-        # Note: the operator is %>, We use %%> for psycopg escaping
-        return sql.SQL("title %%> %(query)s")
-    elif query.match == search_models.CatalogQueryMatch.EndsWith:
-        params["query"] = "%" + query.query
-        return sql.SQL("title ILIKE %(query)s")
-    elif query.match == search_models.CatalogQueryMatch.Contains:
-        params["query"] = "%" + query.query + "%"
-        return sql.SQL("title ILIKE %(query)s")
-    else:  # pragma: nocover
-        # This is a trick so mypy generates an error if this branch can be reached,
-        # that is, if we are missing some ifs
-        _a: int = "a"
-        return sql.SQL("")
+    return facets, nonfacets
 
 
 def _prepare_query(catalog_query: CatalogQuery) -> tuple[sql.Composed, dict[str, Any]]:
@@ -219,67 +262,6 @@ def _prepare_query(catalog_query: CatalogQuery) -> tuple[sql.Composed, dict[str,
     return query, filter_params
 
 
-def _pg_driver() -> PGDriver:
-    return cast(PGDriver, get_driver())
-
-
-@observer.wrap({"op": "search"})
-async def pgcatalog_search(catalog_query: CatalogQuery) -> Resources:
-    # Prepare SQL query
-    query, query_params = _prepare_query_filters(catalog_query)
-
-    async with _pg_driver()._get_connection() as conn, conn.cursor(row_factory=dict_row) as cur:
-        facets = {}
-
-        # Faceted search
-        if catalog_query.faceted:
-            with observer({"op": "facets"}):
-                tmp_facets: dict[str, dict[str, int]] = {
-                    translate_label(f): defaultdict(int) for f in catalog_query.faceted
-                }
-
-                if catalog_query.filters is None:
-                    await _faceted_search_unfiltered(cur, catalog_query, tmp_facets)
-                else:
-                    await _faceted_search_filtered(cur, catalog_query, tmp_facets, query, query_params)
-
-                facets = {translate_system_to_alias_label(k): v for k, v in tmp_facets.items()}
-
-        # Totals
-        with observer({"op": "totals"}):
-            await cur.execute(
-                sql.SQL("SELECT COUNT(*) FROM ({}) fc").format(query),
-                query_params,
-            )
-            total = (await cur.fetchone())["count"]  # type: ignore
-
-        # Query
-        with observer({"op": "query"}):
-            query, query_params = _prepare_query(catalog_query)
-            await cur.execute(query, query_params)
-            data = await cur.fetchall()
-
-    return Resources(
-        facets=facets,
-        results=[
-            ResourceResult(
-                rid=str(r["rid"]).replace("-", ""),
-                field="title",
-                field_type="a",
-                labels=[label for label in r["labels"] if label.startswith("/l/")],
-                score=0,
-            )
-            for r in data
-        ],
-        query=catalog_query.query.query if catalog_query.query else "",
-        total=total,
-        page_number=catalog_query.page_number,
-        page_size=catalog_query.page_size,
-        next_page=(catalog_query.page_size * catalog_query.page_number + len(data) < total),
-        min_score=0,
-    )
-
-
 async def _faceted_search_unfiltered(
     cur: AsyncCursor[DictRow], catalog_query: CatalogQuery, tmp_facets: dict[str, dict[str, int]]
 ):
@@ -360,33 +342,137 @@ async def _faceted_search_filtered(
                 tmp_facets[grandparent][translate_system_to_alias_label(parent)] += count
 
 
-@observer.wrap({"op": "catalog_facets"})
-async def pgcatalog_facets(kbid: str, request: CatalogFacetsRequest) -> dict[str, int]:
-    async with _pg_driver()._get_connection() as conn, conn.cursor() as cur:
-        prefix_filters: list[sql.Composable] = []
-        prefix_params: dict[str, Any] = {}
-        for cnt, prefix in enumerate(request.prefixes):
-            prefix_sql = sql.SQL("facet LIKE {}").format(sql.Placeholder(f"prefix{cnt}"))
-            prefix_params[f"prefix{cnt}"] = f"{prefix.prefix}%"
-            if prefix.depth is not None:
-                prefix_parts = len(prefix.prefix.split("/"))
-                depth_sql = sql.SQL("SPLIT_PART(facet, '/', {}) = ''").format(
-                    sql.Placeholder(f"depth{cnt}")
-                )
-                prefix_params[f"depth{cnt}"] = prefix_parts + prefix.depth + 1
-                prefix_sql = sql.SQL("({} AND {})").format(prefix_sql, depth_sql)
-            prefix_filters.append(prefix_sql)
+def _prepare_query_search(query: search_models.CatalogQuery, params: dict[str, Any]) -> sql.Composable:
+    if query.match == search_models.CatalogQueryMatch.Exact:
+        params["query"] = query.query
+        return sql.SQL("{} = %(query)s").format(sql.Identifier(query.field.value))
+    elif query.match == search_models.CatalogQueryMatch.StartsWith:
+        params["query"] = query.query + "%"
+        if query.field == search_models.CatalogQueryField.Title:
+            # Insensitive search supported by pg_trgm for title
+            return sql.SQL("{} ILIKE %(query)s").format(sql.Identifier(query.field.value))
+        else:
+            # Sensitive search for slug (btree does not support ILIKE and slugs are all lowercase anyway)
+            return sql.SQL("{} LIKE %(query)s").format(sql.Identifier(query.field.value))
+    # The rest of operators only supported by title
+    elif query.match == search_models.CatalogQueryMatch.Words:
+        # This is doing tokenization inside the SQL server (to keep the index updated). We could move it to
+        # the python code at update/query time if it ever becomes a problem but for now, a single regex
+        # executed per query is not a problem.
 
-        filter_sql: sql.Composable
-        if prefix_filters:
-            filter_sql = sql.SQL("AND {}").format(sql.SQL(" OR ").join(prefix_filters))
+        # Remove zero-length words from the split
+        params["query"] = [word.lower() for word in SPLIT_REGEX.split(query.query) if word]
+        return sql.SQL("regexp_split_to_array(lower(title), '\\W') @> %(query)s")
+    elif query.match == search_models.CatalogQueryMatch.Fuzzy:
+        params["query"] = query.query
+        # Note: the operator is %>, We use %%> for psycopg escaping
+        return sql.SQL("title %%> %(query)s")
+    elif query.match == search_models.CatalogQueryMatch.EndsWith:
+        params["query"] = "%" + query.query
+        return sql.SQL("title ILIKE %(query)s")
+    elif query.match == search_models.CatalogQueryMatch.Contains:
+        params["query"] = "%" + query.query + "%"
+        return sql.SQL("title ILIKE %(query)s")
+    else:  # pragma: no cover
+        # This is a trick so mypy generates an error if this branch can be reached,
+        # that is, if we are missing some ifs
+        _a: int = "a"
+        return sql.SQL("")
+
+
+def _convert_filter(expr: CatalogExpression, filter_params: dict[str, Any]) -> sql.Composable:
+    if expr.bool_and:
+        return _convert_boolean_op(expr.bool_and, "and", filter_params)
+    elif expr.bool_or:
+        return _convert_boolean_op(expr.bool_or, "or", filter_params)
+    elif expr.bool_not:
+        return sql.SQL("(NOT {})").format(_convert_filter(expr.bool_not, filter_params))
+    elif expr.date:
+        return _convert_date_filter(expr.date, filter_params)
+    elif expr.facet:
+        param_name = f"param{len(filter_params)}"
+        filter_params[param_name] = [expr.facet]
+        if expr.facet == "/n/s/PROCESSED":
+            # Optimization for the most common case, we know PROCESSED is a full label and can use the smaller labels index
+            # This is needed because PROCESSED is present in most catalog entries and PG is unlikely to use any index
+            # for it, falling back to executing the extract_facets function which can be slow
+            return sql.SQL("labels @> {}").format(sql.Placeholder(param_name))
         else:
-            filter_sql = sql.SQL("")
+            return sql.SQL("extract_facets(labels) @> {}").format(sql.Placeholder(param_name))
+    elif expr.resource_id:
+        param_name = f"param{len(filter_params)}"
+        filter_params[param_name] = [expr.resource_id]
+        return sql.SQL("rid = {}").format(sql.Placeholder(param_name))
+    else:
+        return sql.SQL("")
+
+
+def _convert_boolean_op(
+    operands: list[CatalogExpression],
+    op: Union[Literal["and"], Literal["or"]],
+    filter_params: dict[str, Any],
+) -> sql.Composable:
+    array_op = sql.SQL("@>" if op == "and" else "&&")
+    operands_sql: list[sql.Composable] = []
+    facets, nonfacets = _filter_operands(operands)
+    if facets:
+        param_name = f"param{len(filter_params)}"
+        if facets == ["/n/s/PROCESSED"]:
+            # Optimization for the most common case, we know PROCESSED is a full label and can use the smaller labels index
+            # This is needed because PROCESSED is present in most catalog entries and PG is unlikely to use any index
+            # for it, falling back to executing the extract_facets function which can be slow
+            operands_sql.append(sql.SQL("labels @> {}").format(sql.Placeholder(param_name)))
+        else:
+            operands_sql.append(
+                sql.SQL("extract_facets(labels) {} {}").format(array_op, sql.Placeholder(param_name))
+            )
+        filter_params[param_name] = facets
+    for nonfacet in nonfacets:
+        operands_sql.append(_convert_filter(nonfacet, filter_params))
+    return sql.SQL("({})").format(sql.SQL(f" {op.upper()} ").join(operands_sql))
+
 
-        await cur.execute(
-            sql.SQL(
-                "SELECT facet, COUNT(*) FROM catalog_facets WHERE kbid = %(kbid)s {} GROUP BY facet"
-            ).format(filter_sql),
-            {"kbid": kbid, **prefix_params},
+def _convert_date_filter(date: CatalogExpression.Date, filter_params: dict[str, Any]) -> sql.Composable:
+    if date.since and date.until:
+        since_name = f"param{len(filter_params)}"
+        filter_params[since_name] = date.since
+        until_name = f"param{len(filter_params)}"
+        filter_params[until_name] = date.until
+        return sql.SQL("{field} BETWEEN {since} AND {until}").format(
+            field=sql.Identifier(date.field),
+            since=sql.Placeholder(since_name),
+            until=sql.Placeholder(until_name),
         )
-        return {k: v for k, v in await cur.fetchall()}
+    elif date.since:
+        since_name = f"param{len(filter_params)}"
+        filter_params[since_name] = date.since
+        return sql.SQL("{field} > {since}").format(
+            field=sql.Identifier(date.field), since=sql.Placeholder(since_name)
+        )
+    elif date.until:
+        until_name = f"param{len(filter_params)}"
+        filter_params[until_name] = date.until
+        return sql.SQL("{field} < {until}").format(
+            field=sql.Identifier(date.field), until=sql.Placeholder(until_name)
+        )
+    else:
+        raise ValueError(f"Invalid date operator")
+
+
+def translate_label(literal: str) -> str:
+    if len(literal) == 0:
+        raise InvalidQueryError("filters", "Invalid empty label")
+    if literal[0] != "/":
+        raise InvalidQueryError("filters", f"Invalid label. It must start with a `/`: {literal}")
+    return translate_alias_to_system_label(literal)
+
+
+def extract_facets(labels: list[str]) -> set[str]:
+    facets = set()
+    for label in labels:
+        parts = label.split("/")
+        facet = ""
+        for part in parts[1:]:
+            facet += f"/{part}"
+            facets.add(facet)
+    return facets