fix(cascade): ignore backquoted identifiers in the subquery check; assert the matched row

Review follow-up. The subquery check strips quoted text before looking for SELECT; it now strips MySQL's backquoted identifiers too, as it already did PostgreSQL's double-quoted ones, so a column named `select` no longer makes a seed look like a subquery (that case was safe, since the seed's keys select the same rows). The percent-sign test now checks which row each seed selects, not only how many. The seed helper's docstring notes that replacing a seed by its keys reads them all into one condition.

Author: dimitri-yatsenko

src/datajoint/diagram.py

@@ -160,7 +160,7 @@ def _has_subquery(condition):
     if not isinstance(condition, str):
         return False
     text = re.sub(r"(\\\")|(\\\')", "", condition)  # escaped quotes
-    text = re.sub(r"'[^']*'|\"[^\"]*\"", "", text)  # quoted text and identifiers
+    text = re.sub(r"'[^']*'|\"[^\"]*\"|`[^`]*`", "", text)  # quoted text and identifiers
     return re.search(r"\bselect\b", text, re.IGNORECASE) is not None
 
 
@@ -178,6 +178,10 @@ def _seed_restriction(table_expr):
       to a fresh copy of the table, which rejects a subquery that names other
       tables' columns. A delete would also remove the rows the subquery reads,
       when they are downstream, before it deletes the seed itself.
+
+    Replacing a seed by its keys reads every selected primary key into Python
+    and sends them back as one condition, so a very large seed makes a large
+    query; restrict by the keys in batches if that becomes a problem.
     """
     if table_expr._top is not None or any(_has_subquery(c) for c in table_expr.restriction):
         return AndList([table_expr.keys()]), set(table_expr.primary_key)

tests/integration/test_cascade_overdelete.py

@@ -319,7 +319,7 @@ class Use(dj.Manual):
         (Label & {"name": "50%"}, 1),
         (Label & "name = '5%%'", 3),
     ]:
-        assert len(seed) == 1
+        assert seed.to_arrays("label_id").tolist() == [label]
         assert dj.Diagram.cascade(seed).counts()[Use.full_table_name] == 1
 
     (Label & "name LIKE '50%'").delete(prompt=False)