Merge pull request #293 from Mytherin/noexperimentalpushdown

Enable filter pushdown by default - and add more extensive testing for filtering on all types
duckdb · Feb 1, 2025 · ecabb61 · ecabb61
2 parents 478ee8f + 2113d33
commit ecabb61
Show file tree

Hide file tree

Showing 3 changed files with 39 additions and 11 deletions.
diff --git a/src/postgres_extension.cpp b/src/postgres_extension.cpp
@@ -174,8 +174,8 @@ static void LoadInternal(DatabaseInstance &db) {
 	config.AddExtensionOption("pg_connection_cache", "Whether or not to use the connection cache", LogicalType::BOOLEAN,
 	                          Value::BOOLEAN(true), PostgresConnectionPool::PostgresSetConnectionCache);
 	config.AddExtensionOption("pg_experimental_filter_pushdown",
-	                          "Whether or not to use filter pushdown (currently experimental)", LogicalType::BOOLEAN,
-	                          Value::BOOLEAN(false));
+	                          "Whether or not to use filter pushdown", LogicalType::BOOLEAN,
+	                          Value::BOOLEAN(true));
 	config.AddExtensionOption("pg_null_byte_replacement",
 	                          "When writing NULL bytes to Postgres, replace them with the given character",
 	                          LogicalType::VARCHAR, Value(), SetPostgresNullByteReplacement);

diff --git a/src/postgres_filter_pushdown.cpp b/src/postgres_filter_pushdown.cpp
@@ -35,6 +35,28 @@ string PostgresFilterPushdown::TransformComparision(ExpressionType type) {
 	}
 }
 
+string TransformBlob(const string &val) {
+	char const HEX_DIGITS[] = "0123456789ABCDEF";
+
+	string result = "'\\x";
+	for(idx_t i = 0; i < val.size(); i++) {
+		uint8_t byte_val = static_cast<uint8_t>(val[i]);
+		result += HEX_DIGITS[(byte_val >> 4) & 0xf];
+		result += HEX_DIGITS[byte_val & 0xf];
+	}
+	result += "'::BYTEA";
+	return result;
+}
+
+string TransformLiteral(const Value &val) {
+	switch (val.type().id()) {
+	case LogicalTypeId::BLOB:
+		return TransformBlob(StringValue::Get(val));
+	default:
+		return KeywordHelper::WriteQuoted(val.ToString());
+	}
+}
+
 string PostgresFilterPushdown::TransformFilter(string &column_name, TableFilter &filter) {
 	switch (filter.filter_type) {
 	case TableFilterType::IS_NULL:
@@ -51,7 +73,7 @@ string PostgresFilterPushdown::TransformFilter(string &column_name, TableFilter
 	}
 	case TableFilterType::CONSTANT_COMPARISON: {
 		auto &constant_filter = filter.Cast<ConstantFilter>();
-		auto constant_string = KeywordHelper::WriteQuoted(constant_filter.constant.ToString());
+		auto constant_string = TransformLiteral(constant_filter.constant);
 		auto operator_string = TransformComparision(constant_filter.comparison_type);
 		return StringUtil::Format("%s %s %s", column_name, operator_string, constant_string);
 	}
@@ -72,7 +94,7 @@ string PostgresFilterPushdown::TransformFilter(string &column_name, TableFilter
 			if (!in_list.empty()) {
 				in_list += ", ";
 			}
-			in_list += KeywordHelper::WriteQuoted(val.ToString());
+			in_list += TransformLiteral(val);
 		}
 		return column_name + " IN (" + in_list + ")";
 	}

diff --git a/test/sql/storage/attach_types.test b/test/sql/storage/attach_types.test
@@ -56,13 +56,6 @@ FROM test_all_types();
 statement ok
 CREATE OR REPLACE TABLE s.all_types AS FROM all_types_tbl
 
-#mode output_result
-#
-#statement ok
-#DESCRIBE SELECT COLUMNS(*)::VARCHAR FROM all_types_tbl
-#
-#mode skip
-
 query IIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIIII
 SELECT COLUMNS(*)::VARCHAR FROM all_types_tbl
 ----
@@ -76,3 +69,16 @@ SELECT COLUMNS(*)::VARCHAR FROM s.all_types
 false	-128	-32768	-2147483648	-9223372036854775808	0	0	0	2000-01-01	00:00:00	2000-01-01 01:02:03	2000-01-01 01:02:03	2000-01-01 01:02:03	2000-01-01 01:02:03	00:00:00+15:00	2000-01-01 01:02:03	-999.9	-99999.9999	-999999999999.999999	-9999999999999999999999999999.9999999999	00000000-0000-0000-0000-000000000000	00:00:00	🦆🦆🦆🦆🦆🦆	thisisalongblob\x00withnullbytes	0010001001011100010101011010111	DUCK_DUCK_ENUM	enum_0	enum_0	[]	[]	[]	[]	[]	[🦆🦆🦆🦆🦆🦆, goose, NULL]
 true	127	32767	2147483647	9223372036854775807	255	65535	4294967295	2000-01-01	24:00:00	2000-01-01 01:02:03	2000-01-01 01:02:03	2000-01-01 01:02:03	2000-01-01 01:02:03	00:00:00+15:00	2000-01-01 01:02:03	999.9	99999.9999	999999999999.999999	9999999999999999999999999999.9999999999	ffffffff-ffff-ffff-ffff-ffffffffffff	83 years 3 months 999 days 00:16:39.999999	goo se	\x00\x00\x00a	10101	GOOSE	enum_299	enum_69999	[42, 999, NULL, NULL, -42]	[42.0, nan, inf, -inf, NULL, -42.0]	[1970-01-01, infinity, -infinity, NULL, 2022-05-12]	[1970-01-01 00:00:00, infinity, -infinity, NULL, 2022-05-12 16:23:45]	[1970-01-01 00:00:00+00, infinity, -infinity, NULL, 2022-05-12 23:23:45+00]	[🦆🦆🦆🦆🦆🦆, goose, NULL]
 NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL	NULL
+
+# filter pushdown
+foreach column_name bool tinyint smallint int bigint utinyint usmallint uint date time timestamp timestamp_s timestamp_ms timestamp_ns time_tz timestamp_tz dec_4_1 dec_9_4 dec_18_6 dec38_10 uuid interval varchar blob bit small_enum medium_enum large_enum int_array double_array date_array timestamp_array timestamptz_array varchar_array
+
+statement ok
+SET VARIABLE minimum_value=(SELECT MIN(${column_name}) min_val FROM s.all_types);
+
+query I
+SELECT ANY_VALUE(${column_name})=getvariable('minimum_value') FROM s.all_types WHERE ${column_name}=getvariable('minimum_value')
+----
+true
+
+endloop