Merge pull request #83 from Mytherin/issue75

Fix #75: no longer throw when interpreting TINYINT(1) columns with non-0 or 1 values as booleans
duckdb · Sep 5, 2024 · 248b976 · 248b976
2 parents 64cb6ae + 9f2400a
commit 248b976
Show file tree

Hide file tree

Showing 7 changed files with 38 additions and 13 deletions.
diff --git a/duckdb b/duckdb
diff --git a/src/mysql_scanner.cpp b/src/mysql_scanner.cpp
@@ -92,13 +92,24 @@ void CastBoolFromMySQL(ClientContext &context, Vector &input, Vector &result, id
 		}
 		auto str_data = input_data[r].GetData();
 		auto str_size = input_data[r].GetSize();
-		if (str_size != 1) {
-			throw BinderException("Failed to cast MySQL boolean - expected 1 byte "
-			                      "element but got element of size %s",
-			                      str_size);
+		if (str_size == 0) {
+			throw BinderException(
+			    "Failed to cast MySQL boolean - expected 1 byte element but got element of size %d\n* SET "
+			    "mysql_tinyint1_as_boolean=false to disable loading TINYINT(1) columns as booleans\n* SET "
+			    "mysql_bit1_as_boolean=false to disable loading BIT(1) columns as booleans",
+			    str_size);
+		}
+		// booleans are EITHER binary "1" or "0" (BIT(1))
+		// OR a number
+		// in both cases we can figure out what value it is from the first character:
+		// \0 -> zero byte, false
+		// - -> negative number, false
+		// 0 -> zero number, false
+		if (*str_data == '\0' || *str_data == '0' || *str_data == '-') {
+			result_data[r] = false;
+		} else {
+			result_data[r] = true;
 		}
-		auto bool_char = *str_data;
-		result_data[r] = bool_char == '\1' || bool_char == '1';
 	}
 }
 

diff --git a/src/storage/mysql_table_entry.cpp b/src/storage/mysql_table_entry.cpp
@@ -81,7 +81,8 @@ unique_ptr<BaseStatistics> MySQLTableEntry::GetStatistics(ClientContext &context
 	return nullptr;
 }
 
-void MySQLTableEntry::BindUpdateConstraints(Binder &binder, LogicalGet &, LogicalProjection &, LogicalUpdate &, ClientContext &) {
+void MySQLTableEntry::BindUpdateConstraints(Binder &binder, LogicalGet &, LogicalProjection &, LogicalUpdate &,
+                                            ClientContext &) {
 }
 
 TableFunction MySQLTableEntry::GetScanFunction(ClientContext &context, unique_ptr<FunctionData> &bind_data) {

diff --git a/test/sql/attach_fake_booleans.test b/test/sql/attach_fake_booleans.test
@@ -18,6 +18,8 @@ SELECT * FROM s1.fake_booleans
 true	true
 false	false
 NULL	NULL
+false	true
+true	false
 
 query II
 SELECT typeof(tinyint_bool), typeof(bit_bool) FROM s1.fake_booleans LIMIT 1
@@ -40,6 +42,15 @@ SELECT typeof(tinyint_bool), typeof(bit_bool) FROM s1.fake_booleans LIMIT 1
 ----
 TINYINT	BLOB
 
+query II
+SELECT * FROM s1.fake_booleans
+----
+1	\x01
+0	\x00
+NULL	NULL
+-128	\x01
+127	\x00
+
 statement ok
 SET mysql_tinyint1_as_boolean=true
 

diff --git a/test/sql/attach_types.test b/test/sql/attach_types.test
@@ -33,7 +33,7 @@ NULL	NULL	NULL	NULL
 # test all types
 statement ok
 CREATE TABLE all_types_tbl AS SELECT *
-EXCLUDE (float, double, ubigint, hugeint, uhugeint, int_array, double_array, date_array, timestamp_array, timestamptz_array, varchar_array, nested_int_array, struct, struct_of_arrays, array_of_structs, map, "union",fixed_int_array,fixed_varchar_array,fixed_nested_varchar_array,list_of_fixed_int_array,fixed_array_of_int_list,fixed_nested_int_array,struct_of_fixed_array,fixed_struct_array
+EXCLUDE (float, double, ubigint, hugeint, uhugeint, int_array, double_array, date_array, timestamp_array, timestamptz_array, varchar_array, nested_int_array, struct, struct_of_arrays, array_of_structs, map, "union",fixed_int_array,fixed_varchar_array,fixed_nested_varchar_array,list_of_fixed_int_array,fixed_array_of_int_list,fixed_nested_int_array,struct_of_fixed_array,fixed_struct_array, varint
 )
 REPLACE(
 	CASE WHEN int IS NOT NULL THEN '2000-01-01' ELSE NULL END AS date,

diff --git a/test/sql/mysql_query.test b/test/sql/mysql_query.test
@@ -90,14 +90,16 @@ DECIMAL(2,1)	DECIMAL(5,1)	DECIMAL(10,2)	DECIMAL(20,3)	DOUBLE
 query II
 SELECT * FROM mysql_query('simple', 'SELECT * FROM fake_booleans')
 ----
-true	true
-false	false
+1	true
+0	false
 NULL	NULL
+-128	true
+127	false
 
 query II
 SELECT * FROM mysql_query('simple', 'SELECT * FROM bits')
 ----
-false	\x00\x00\x00\x00\x00\x01UU
+true	\x00\x00\x00\x00\x00\x01UU
 NULL	NULL
 
 query III

diff --git a/test/test_data.sql b/test/test_data.sql
@@ -82,7 +82,7 @@ INSERT INTO decimals VALUES (
 
 -- MySQL doesn't support a "proper" boolean so people often use TINYINT(1) or BIT(1) as boolean
 CREATE TABLE fake_booleans(tinyint_bool TINYINT(1), bit_bool BIT(1));
-INSERT INTO fake_booleans VALUES (true, true), (false, false), (NULL, NULL);
+INSERT INTO fake_booleans VALUES (true, true), (false, false), (NULL, NULL), (-128, b'1'), (127, b'0');
 
 CREATE TABLE bits(b BIT(6), bl BIT(64));
 INSERT INTO bits VALUES (b'000101', b'010101010101010101');