From 00b37e1462ad6aa4b85d3eec7d07a90a7b6c480c Mon Sep 17 00:00:00 2001 From: Mike Bayer Date: Thu, 31 Mar 2022 16:53:43 -0400 Subject: [PATCH] allow executemany values for ON CONFLICT DO NOTHING Scaled back a fix made for :ticket:`6581` where "executemany values" mode for psycopg2 were disabled for all "ON CONFLICT" styles of INSERT, to not apply to the "ON CONFLICT DO NOTHING" clause, which does not include any parameters and is safe for "executemany values" mode. "ON CONFLICT DO UPDATE" is still blocked from "executemany values" as there may be additional parameters in the DO UPDATE clause that cannot be batched (which is the original issue fixed by :ticket:`6581`). Fixes: #7880 Change-Id: Id3e23a0c6699333409a50148fa8923cb8e564bdc (cherry picked from commit cc8c5835a92b0035530b541c81b0c714b570b095) --- doc/build/changelog/unreleased_14/7880.rst | 11 ++++++++++ lib/sqlalchemy/dialects/postgresql/base.py | 19 ++++++++++++++++++ .../dialects/postgresql/psycopg2.py | 2 +- lib/sqlalchemy/sql/compiler.py | 9 +-------- test/dialect/postgresql/test_dialect.py | 20 +++++++++++++++++-- 5 files changed, 50 insertions(+), 11 deletions(-) create mode 100644 doc/build/changelog/unreleased_14/7880.rst diff --git a/doc/build/changelog/unreleased_14/7880.rst b/doc/build/changelog/unreleased_14/7880.rst new file mode 100644 index 0000000000..9abbac1cfa --- /dev/null +++ b/doc/build/changelog/unreleased_14/7880.rst @@ -0,0 +1,11 @@ +.. change:: + :tags: bug, postgresql + :tickets: 7880 + + Scaled back a fix made for :ticket:`6581` where "executemany values" mode + for psycopg2 were disabled for all "ON CONFLICT" styles of INSERT, to + not apply to the "ON CONFLICT DO NOTHING" clause, which does not include + any parameters and is safe for "executemany values" mode. "ON CONFLICT + DO UPDATE" is still blocked from "executemany values" as there may + be additional parameters in the DO UPDATE clause that cannot be batched + (which is the original issue fixed by :ticket:`6581`). diff --git a/lib/sqlalchemy/dialects/postgresql/base.py b/lib/sqlalchemy/dialects/postgresql/base.py index 37b2113e98..4f63002bc9 100644 --- a/lib/sqlalchemy/dialects/postgresql/base.py +++ b/lib/sqlalchemy/dialects/postgresql/base.py @@ -1429,6 +1429,7 @@ import re from uuid import UUID as _python_UUID from . import array as _array +from . import dml from . import hstore as _hstore from . import json as _json from . import ranges as _ranges @@ -2463,6 +2464,24 @@ class PGCompiler(compiler.SQLCompiler): return target_text + @util.memoized_property + def _is_safe_for_fast_insert_values_helper(self): + # don't allow fast executemany if _post_values_clause is + # present and is not an OnConflictDoNothing. what this means + # concretely is that the + # "fast insert executemany helper" won't be used, in other + # words we won't convert "executemany()" of many parameter + # sets into a single INSERT with many elements in VALUES. + # We can't apply that optimization safely if for example the + # statement includes a clause like "ON CONFLICT DO UPDATE" + + return self.insert_single_values_expr is not None and ( + self.statement._post_values_clause is None + or isinstance( + self.statement._post_values_clause, dml.OnConflictDoNothing + ) + ) + def visit_on_conflict_do_nothing(self, on_conflict, **kw): target_text = self._on_conflict_target(on_conflict, **kw) diff --git a/lib/sqlalchemy/dialects/postgresql/psycopg2.py b/lib/sqlalchemy/dialects/postgresql/psycopg2.py index 19d7b06ac9..adebc9b676 100644 --- a/lib/sqlalchemy/dialects/postgresql/psycopg2.py +++ b/lib/sqlalchemy/dialects/postgresql/psycopg2.py @@ -927,7 +927,7 @@ class PGDialect_psycopg2(PGDialect): self.executemany_mode & EXECUTEMANY_VALUES and context and context.isinsert - and context.compiled.insert_single_values_expr + and context.compiled._is_safe_for_fast_insert_values_helper ): executemany_values = ( "(%s)" % context.compiled.insert_single_values_expr diff --git a/lib/sqlalchemy/sql/compiler.py b/lib/sqlalchemy/sql/compiler.py index 671ca67492..7393629a40 100644 --- a/lib/sqlalchemy/sql/compiler.py +++ b/lib/sqlalchemy/sql/compiler.py @@ -3930,14 +3930,7 @@ class SQLCompiler(Compiled): [value for c, expr, value in crud_params] ) text += " VALUES (%s)" % insert_single_values_expr - if toplevel and insert_stmt._post_values_clause is None: - # don't assign insert_single_values_expr if _post_values_clause - # is present. what this means concretely is that the - # "fast insert executemany helper" won't be used, in other - # words we won't convert "executemany()" of many parameter - # sets into a single INSERT with many elements in VALUES. - # We can't apply that optimization safely if for example the - # statement includes a clause like "ON CONFLICT DO UPDATE" + if toplevel: self.insert_single_values_expr = insert_single_values_expr if insert_stmt._post_values_clause is not None: diff --git a/test/dialect/postgresql/test_dialect.py b/test/dialect/postgresql/test_dialect.py index c0eb4410cf..1d797a697d 100644 --- a/test/dialect/postgresql/test_dialect.py +++ b/test/dialect/postgresql/test_dialect.py @@ -30,6 +30,7 @@ from sqlalchemy import text from sqlalchemy import TypeDecorator from sqlalchemy import util from sqlalchemy.dialects.postgresql import base as postgresql +from sqlalchemy.dialects.postgresql import insert as pg_insert from sqlalchemy.dialects.postgresql import psycopg2 as psycopg2_dialect from sqlalchemy.dialects.postgresql.psycopg2 import EXECUTEMANY_BATCH from sqlalchemy.dialects.postgresql.psycopg2 import EXECUTEMANY_PLAIN @@ -357,7 +358,10 @@ class ExecuteManyMode(object): Column(ue("\u6e2c\u8a66"), Integer), ) - def test_insert(self, connection): + @testing.combinations( + "insert", "pg_insert", "pg_insert_on_conflict", argnames="insert_type" + ) + def test_insert(self, connection, insert_type): from psycopg2 import extras values_page_size = connection.dialect.executemany_values_page_size @@ -377,11 +381,23 @@ class ExecuteManyMode(object): else: assert False + if insert_type == "pg_insert_on_conflict": + stmt += " ON CONFLICT DO NOTHING" + with mock.patch.object( extras, meth.__name__, side_effect=meth ) as mock_exec: + if insert_type == "insert": + ins_stmt = self.tables.data.insert() + elif insert_type == "pg_insert": + ins_stmt = pg_insert(self.tables.data) + elif insert_type == "pg_insert_on_conflict": + ins_stmt = pg_insert(self.tables.data).on_conflict_do_nothing() + else: + assert False + connection.execute( - self.tables.data.insert(), + ins_stmt, [ {"x": "x1", "y": "y1"}, {"x": "x2", "y": "y2"}, -- 2.47.2