Add script to convert minidumps from GPDB5 format to GPDB6+ (#548)

GPDB6 introduced additional statistitcs columns. This script adds
"dummy" values to those extra columns so that we can test older
minirepros on 6X. This should allow us to do more of our development on
6X instead of 5X.
Authored-by: NChris Hajas <chajas@pivotal.io>

CMakeLists.txt

@@ -6,7 +6,7 @@ project(gpopt LANGUAGES CXX C)
 set(CMAKE_CXX_STANDARD 98)
 set(GPORCA_VERSION_MAJOR 3)
 set(GPORCA_VERSION_MINOR 79)
-set(GPORCA_VERSION_PATCH 1)
+set(GPORCA_VERSION_PATCH 2)
 set(GPORCA_VERSION_STRING "${GPORCA_VERSION_MAJOR}.${GPORCA_VERSION_MINOR}.${GPORCA_VERSION_PATCH}")
 
 # Whenever an ABI-breaking change is made to GPORCA, this should be incremented.

scripts/convert_minirepro_5_to_6.py

+#!/usr/local/bin/python
+
+import argparse
+
+_help = """
+Converts a minirepro taken with GPDB5 or earlier for use with GPDB 6 and later
+"""
+
+
+# Description:
+#
+# Usage: ./convert_minirepro_5_to_6.py my_minirepro
+# creates my_minirepro_6X file
+#
+# GPDB6 added a new field to pg_statistic, making minirepros taken with GPDB5
+# or earlier incompatible with GPDB6+. This script adds "dummy" fields so the
+# minirepro can be loaded into a GPDB6+ system for testing/verification. ORCA
+# does not use this field currently, so the plan generated should be identical.
+# This script outputs a file suffixed with "_6X" in the same directory as the
+# original file.
+
+
+def process_minirepro(input_filepath, output_filepath):
+	with open(input_filepath, 'r') as infile, open(output_filepath, 'w+') as outfile:
+		line = infile.readline()
+		while line:
+			if 'allow_system_table_mods=on' in line:
+				raise Exception('Minirepro already in GPDB6 format')
+			# allow_system_table_mods GUC was changed to a boolean
+			if 'set allow_system_table_mods="DML";' in line:
+				line = 'set allow_system_table_mods=on;\n'
+			outfile.write(line)
+			if 'INSERT INTO pg_statistic' in line:
+				convert_insert_statement(infile, outfile)
+			line = infile.readline()
+
+
+# This expects and converts an insert statement from the old,
+# GPDB5 format (below) to the new GPDB6+ format:
+# INSERT INTO pg_statistic VALUES (
+# 	'mysch.mytbl'::regclass,
+# 	2::smallint,
+# 	0.0::real,
+# 	3::integer,
+# 	1.0::real,
+# 	1::smallint,
+# 	2::smallint,
+# 	0::smallint,
+# 	0::smallint,
+# 	23::oid,
+# 	42::oid,
+# 	0::oid,
+# 	0::oid,
+# 	E'{1}'::real[],
+# 	E'{1}'::real[],
+# 	NULL::real[],
+# 	NULL::real[],
+# 	E'{-1}'::int8[],
+# 	NULL::int8[],
+# 	NULL::int8[],
+# 	NULL::int8[]);
+
+# The following map describes the additional lines that must be added to the
+# existing insert statement. These lines are added at the line_number key,
+# and assume 0 indexing
+line_map = {2: "\tFalse::boolean,\n",
+			9: "\t0::smallint,\n",
+			13: "\t0::oid,\n",
+			17: "\tNULL::real[],\n",
+			21: "\tNULL::anyarray);\n",
+			}
+
+
+def convert_insert_statement(infile, outfile):
+	for line_number in range(0, 22):
+		line_to_insert = line_map.get(line_number)
+		if line_to_insert:
+			outfile.write(line_to_insert)
+		line = infile.readline()
+		# Remove the parenthesis and semicolon from the last line,
+		# since we're adding another line to the end
+		if line_number == 20:
+			line = line.replace(");", ",")
+		outfile.write(line)
+
+
+def parseargs():
+	parser = argparse.ArgumentParser(description=_help, version='1.0')
+
+	parser.add_argument("filepath", help="Path to minirepro file")
+
+	args = parser.parse_args()
+	return args
+
+
+def main():
+	args = parseargs()
+
+	input_filepath = args.filepath
+	output_filepath = input_filepath + "_6X"
+
+	process_minirepro(input_filepath, output_filepath)
+
+
+if __name__ == "__main__":
+	main()