Apply low-risk cleanup and docs consistency fixes

Author: philippemnoel

README.md

@@ -26,7 +26,7 @@ pip install -e .[test,dev]
 - `paradedb.sqlalchemy.search`: ParadeDB predicates (`match_all`, `fuzzy`, `parse`, `more_like_this`, etc.).
 - `paradedb.sqlalchemy.pdb`: function wrappers (`score`, `snippet`, `snippets`, `agg`).
 - `paradedb.sqlalchemy.facets`: aggregate/facet JSON builders and rows+facets helper.
-- `paradedb.sqlalchemy.select_with`: select decorators for score/snippet/snippet_positions columns.
+- `paradedb.sqlalchemy.select_with`: select decorators for score/snippet/snippets/snippet_positions columns.
 - `paradedb.sqlalchemy.alembic`: Alembic operations for BM25 index lifecycle.
 
 ## Quickstart
@@ -58,18 +58,20 @@ products_bm25_idx = Index(
     "products_bm25_idx",
     indexing.BM25Field(Product.id),
     indexing.BM25Field(
-        expr.json_text(Product.metadata, "color"),
+        expr.json_text(Product.metadata_, "color"),
         tokenizer=indexing.tokenize.literal(alias="metadata_color"),
     ),
     indexing.BM25Field(
-        expr.json_text(Product.metadata, "location"),
+        expr.json_text(Product.metadata_, "location"),
         tokenizer=indexing.tokenize.literal(alias="metadata_location"),
     ),
     postgresql_using="bm25",
     postgresql_with={"key_field": "id"},
 )
 ```
 
+Use `metadata_` as the ORM attribute name if your underlying column is named `metadata`.
+
 Tokenizer configs can use a Django/Rails-style structured shape:
 
 ```python
@@ -107,7 +109,7 @@ products_bm25_idx = Index(
 
 - Basic predicates: `match_all`, `match_any`, `term`, `phrase`, `fuzzy`, `regex`, `all`
 - Advanced predicates: `parse`, `phrase_prefix`, `regex_phrase`, `near`, `proximity`, `more_like_this`
-- Scoring/snippets: `pdb.score`, `pdb.snippet`, `pdb.snippets`, `pdb.snippet_positions`, `select_with.score`, `select_with.snippet`, `select_with.snippet_positions`
+- Scoring/snippets: `pdb.score`, `pdb.snippet`, `pdb.snippets`, `pdb.snippet_positions`, `select_with.score`, `select_with.snippet`, `select_with.snippets`, `select_with.snippet_positions`
 - Aggregations/facets: `facets.*` builders + `pdb.agg(...)`
 - Rows + facets: `facets.with_rows(...)`
 

paradedb/sqlalchemy/expr.py

@@ -3,16 +3,10 @@
 from sqlalchemy import func, literal
 from sqlalchemy.sql.elements import ClauseElement
 
-from .indexing import TokenizerSpec
-
 
 def json_text(json_expr: ClauseElement, key: str) -> ClauseElement:
     return json_expr.op("->>")(literal(key))
 
 
 def concat_ws(separator: str, *parts: ClauseElement) -> ClauseElement:
     return func.concat_ws(separator, *parts)
-
-
-def tokenizer_alias(tokenizer: TokenizerSpec | None) -> str | None:
-    return None if tokenizer is None else tokenizer.alias

tests/integration/test_paradedb_queries_integration.py

@@ -7,7 +7,7 @@
 from __future__ import annotations
 
 import pytest
-from sqlalchemy import select
+from sqlalchemy import func, select
 
 from conftest import MockItem, assert_uses_paradedb_scan
 from paradedb.sqlalchemy import pdb, search
@@ -583,7 +583,7 @@ def test_all_predicate_matches_everything(mock_session):
     """search.all() matches every indexed row."""
     total = mock_session.scalar(
         select(MockItem.id).where(search.all(MockItem.id)).with_only_columns(
-            __import__("sqlalchemy").func.count()
+            func.count()
         )
     )
     # mock_items typically has many rows (paradedb creates ~41)