relational/test/driver.py

233 lines
7.1 KiB
Python

#!/usr/bin/env python
# -*- coding: utf-8 -*-
# coding=UTF-8
# Relational
# Copyright (C) 2010 Salvo "LtWorf" Tomaselli
#
# Relational is free software: you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation, either version 3 of the License, or
# (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program. If not, see <http://www.gnu.org/licenses/>.
#
# author Salvo "LtWorf" Tomaselli <tiposchi@tiscali.it>
from relational import relation, parser, optimizer
from relational_readline.termcolor import colored
import os
print relation
rels={}
examples_path='samples/'
tests_path='test/'
def readfile(fname):
'''Reads a file as string and returns its content'''
fd=open(fname)
expr=fd.read()
fd.close()
return expr
def load_relations():
'''Loads all the relations present in the directory indicated in the
examples_path variable and stores them in the rels dictionary'''
print "Loading relations"
for i in os.listdir(examples_path):
if i.endswith('.csv'): #It's a relation, loading it
#Naming the relation
relname=i[:-4]
print ("Loading relation %s with name %s..." % (i,relname)),
rels[relname]=relation.relation('%s%s' % (examples_path,i))
print 'done'
def execute_tests():
py_bad=0
py_good=0
py_tot=0
q_bad=0
q_good=0
q_tot=0
ex_bad=0
ex_good=0
ex_tot=0
for i in os.listdir(tests_path):
if i.endswith('.query'):
q_tot+=1
if run_test(i[:-6]):
q_good+=1
else:
q_bad+=1
elif i.endswith('.python'):
py_tot+=1
if run_py_test(i[:-7]):
py_good+=1
else:
py_bad+=1
elif i.endswith('.exec'):
ex_tot+=1
if run_exec_test(i[:-5]):
ex_good+=1
else:
ex_bad+=1
print "\n\033[36;1mResume of the results\033[0m"
print "\n\033[35;1mQuery tests\033[0m"
print "Total test count: %d" % q_tot
print "Passed tests: %d" % q_good
if q_bad>0:
print "\033[31;1mFailed tests count: %d\033[0m" % q_bad
print "\n\033[35;1mPython tests\033[0m"
print "Total test count: %d" % py_tot
print "Passed tests: %d" % py_good
if py_bad>0:
print "\033[31;1mFailed tests count: %d\033[0m" % py_bad
print "\n\033[35;1mExecute Python tests\033[0m"
print "Total test count: %d" % ex_tot
print "Passed tests: %d" % ex_good
if ex_bad>0:
print "\033[31;1mFailed tests count: %d\033[0m" % ex_bad
print "\n\033[36;1mTotal results\033[0m"
if q_bad+py_bad+ex_bad==0:
print "\033[32;1mNo failed tests\033[0m"
else:
print "\033[31;1mThere are %d failed tests\033[0m" % (py_bad+q_bad+ex_bad)
def run_exec_test(testname):
'''Runs a python test, which executes code directly rather than queries'''
print "Running python test: \033[35;1m%s\033[0m" % testname
glob=rels.copy()
exp_result={}
try:
expr=readfile('%s%s.exec' % (tests_path,testname))
exec(expr,glob) #Evaluating the expression
expr=readfile('%s%s.exec' % (tests_path,testname))
exec(expr,glob) #Evaluating the expression
expr=readfile('%s%s.result' % (tests_path,testname))
exp_result=eval(expr,rels) #Evaluating the expression
if isinstance(exp_result,dict):
fields_ok=True
for i in exp_result:
fields_ok = fields_ok and glob[i]==exp_result[i]
if fields_ok:
print colored('Test passed','green')
return True
except:
pass
print colored('ERROR','red')
print colored('=====================================','red')
print "Expected %s" % exp_result
#print "Got %s" % result
print colored('=====================================','red')
return False
def run_py_test(testname):
'''Runs a python test, which evaluates expressions directly rather than queries'''
print "Running expression python test: " + colored (testname,'magenta')
try:
expr=readfile('%s%s.python' % (tests_path,testname))
result=eval(expr,rels) #Evaluating the expression
expr=readfile('%s%s.result' % (tests_path,testname))
exp_result=eval(expr,rels) #Evaluating the expression
if result==exp_result:
print colored('Test passed','green')
return True
except:
pass
print colored('ERROR','red')
print colored('=====================================','red')
print "Expected %s" % exp_result
print "Got %s" % result
print colored('=====================================','red')
return False
def run_test(testname):
'''Runs a specific test executing the file
testname.query
and comparing the result with
testname.result
The query will be executed both unoptimized and
optimized'''
print "Running test: "+ colored(testname,'magenta')
query=None;expr=None;o_query=None;o_expr=None
result_rel=None
result=None
o_result=None
try:
result_rel=relation.relation('%s%s.result' % (tests_path,testname))
query=readfile('%s%s.query' % (tests_path,testname)).strip()
o_query=optimizer.optimize_all(query,rels)
expr=parser.parse(query)#Converting expression to python code
result=eval(expr,rels) #Evaluating the expression
o_expr=parser.parse(o_query)#Converting expression to python code
o_result=eval(o_expr,rels) #Evaluating the expression
if (o_result==result_rel) and (result==result_rel):
print colored('Test passed','green')
return True
except Exception as inst:
print inst
pass
print colored('ERROR','red')
print "Query: %s -> %s" % (query,expr)
print "Optimized query: %s -> %s" % (o_query,o_expr)
print colored('=====================================','red')
print "\033[33;1mExpected result\033[0m"
print result_rel
print "\033[33;1mResult\033[0m"
print result
print "\033[33;1mOptimized result\033[0m"
print o_result
print "\033[33;1moptimized result match\033[0m", result_rel==o_result
print "\033[33;1mresult match \033[0m", result==result_rel
print colored('=====================================','red')
return False
if __name__ == '__main__':
print "-> Starting testsuite for relational"
load_relations()
print "-> Starting tests"
execute_tests()