Merge pull request #195 from Mytherin/ctidscan

Add pg_use_ctid_scan setting, disable parallel ctid scans by default on older postgres versions

Author: Mytherin

src/postgres_extension.cpp

@@ -95,6 +95,8 @@ static void LoadInternal(DatabaseInstance &db) {
 
 	config.AddExtensionOption("pg_use_binary_copy", "Whether or not to use BINARY copy to read data",
 	                          LogicalType::BOOLEAN, Value::BOOLEAN(true));
+	config.AddExtensionOption("pg_use_ctid_scan", "Whether or not to parallelize scanning using table ctids",
+							  LogicalType::BOOLEAN, Value::BOOLEAN(true));
 	config.AddExtensionOption("pg_pages_per_task", "The amount of pages per task", LogicalType::UBIGINT,
 	                          Value::UBIGINT(PostgresBindData::DEFAULT_PAGES_PER_TASK));
 	config.AddExtensionOption("pg_connection_limit", "The maximum amount of concurrent Postgres connections",

src/postgres_scanner.cpp

@@ -106,6 +106,19 @@ void PostgresScanFunction::PrepareBind(PostgresVersion version, ClientContext &c
 			bind_data.pages_per_task = PostgresBindData::DEFAULT_PAGES_PER_TASK;
 		}
 	}
+	bool use_ctid_scan = true;
+	Value pg_use_ctid_scan;
+	if (context.TryGetCurrentSetting("pg_use_ctid_scan", pg_use_ctid_scan)) {
+		use_ctid_scan = BooleanValue::Get(pg_use_ctid_scan);
+	}
+	if (version.major_v < 14) {
+		// Disable parallel CTID scan on older Postgres versions since it is not efficient
+		// see https://github.com/duckdb/postgres_scanner/issues/186
+		use_ctid_scan = false;
+	}
+	if (!use_ctid_scan) {
+		approx_num_pages = 0;
+	}
 	bind_data.SetTablePages(approx_num_pages);
 	bind_data.version = version;
 }

test/sql/storage/attach_verify_big_table.test_slow

@@ -20,3 +20,11 @@ SELECT COUNT(*), SUM(generate_series) FROM s1.big_generated_table
 1000000	499999500000
 
 endloop
+
+statement ok
+SET pg_use_ctid_scan=false;
+
+query II
+SELECT COUNT(*), SUM(generate_series) FROM s1.big_generated_table
+----
+1000000	499999500000