class Test(TestBase):
def setup(self):
datatypes = lambda *types: [validate_data_type(t) for t in types]
schema = Schema.from_lists(
['name', 'id', 'fid', 'isMale', 'scale', 'birth'],
datatypes('string', 'int64', 'float64', 'boolean', 'decimal',
'datetime'), ['ds'], datatypes('string'))
table = MockTable(name='pyodps_test_expr_table', schema=schema)
self.expr = CollectionExpr(_source_data=table,
_schema=Schema(columns=schema.columns))
table1 = MockTable(name='pyodps_test_expr_table1', schema=schema)
self.expr1 = CollectionExpr(_source_data=table1,
_schema=Schema(columns=schema.columns))
table2 = MockTable(name='pyodps_test_expr_table2', schema=schema)
self.expr2 = CollectionExpr(_source_data=table2,
_schema=Schema(columns=schema.columns))
schema2 = Schema.from_lists(['name', 'id', 'fid'],
datatypes('string', 'int64', 'float64'),
['part1', 'part2'],
datatypes('string', 'int64'))
table3 = MockTable(name='pyodps_test_expr_table2', schema=schema2)
self.expr3 = CollectionExpr(_source_data=table3,
_schema=Schema(columns=schema2.columns))
def testProjectPrune(self):
expr = self.expr.select('name', 'id')
new_expr = ColumnPruning(expr.to_dag()).prune()
self.assertIsInstance(new_expr, ProjectCollectionExpr)
self.assertIsNotNone(new_expr.input._source_data)
expected = 'SELECT t1.`name`, t1.`id` \n' \
'FROM mocked_project.`pyodps_test_expr_table` t1'
self.assertEqual(expected,
ODPSEngine(self.odps).compile(expr, prettify=False))
expr = self.expr[Scalar(3).rename('const'),
NullScalar('string').rename('string_const'),
self.expr.id]
expected = 'SELECT 3 AS `const`, CAST(NULL AS STRING) AS `string_const`, t1.`id` \n' \
'FROM mocked_project.`pyodps_test_expr_table` t1'
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr.select(
pt=BuiltinFunction('max_pt', args=(self.expr._source_data.name, )))
expected = "SELECT max_pt('pyodps_test_expr_table') AS `pt` \n" \
"FROM mocked_project.`pyodps_test_expr_table` t1"
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
def testApplyPrune(self):
@output(['name', 'id'], ['string', 'string'])
def h(row):
yield row[0], row[1]
expr = self.expr[self.expr.fid < 0].apply(h, axis=1)['id', ]
new_expr = ColumnPruning(expr.to_dag()).prune()
self.assertIsInstance(new_expr, ProjectCollectionExpr)
self.assertIsInstance(new_expr.input.input, FilterCollectionExpr)
self.assertIsNotNone(new_expr.input.input.input._source_data)
def testFilterPrune(self):
expr = self.expr.filter(self.expr.name == 'name1')
expr = expr['name', 'id']
new_expr = ColumnPruning(expr.to_dag()).prune()
self.assertIsInstance(new_expr.input, FilterCollectionExpr)
self.assertNotIsInstance(new_expr.input.input, ProjectCollectionExpr)
self.assertIsNotNone(new_expr.input.input._source_data)
expected = 'SELECT t1.`name`, t1.`id` \n' \
'FROM mocked_project.`pyodps_test_expr_table` t1 \n' \
'WHERE t1.`name` == \'name1\''
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr.filter(self.expr.name == 'name1')
new_expr = ColumnPruning(expr.to_dag()).prune()
self.assertIsInstance(new_expr, FilterCollectionExpr)
self.assertIsNotNone(new_expr.input._source_data)
expr = self.expr.filter(self.expr.id.isin(self.expr3.id))
expected = 'SELECT * \n' \
'FROM mocked_project.`pyodps_test_expr_table` t1 \n' \
'WHERE t1.`id` IN (SELECT t3.`id` FROM ( ' \
'SELECT t2.`id` FROM mocked_project.`pyodps_test_expr_table2` t2 ) t3)'
self.assertTrue(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
def testFilterPartitionPrune(self):
expr = self.expr.filter_partition('ds=today')[lambda x: x.fid < 0][
'name', lambda x: x.id + 1]
new_expr = ColumnPruning(expr.to_dag()).prune()
self.assertEqual(set(new_expr.input.input.schema.names),
set(['name', 'id', 'fid']))
expected = "SELECT t2.`name`, t2.`id` + 1 AS `id` \n" \
"FROM (\n" \
" SELECT t1.`name`, t1.`id`, t1.`fid` \n" \
" FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
" WHERE t1.`ds` == 'today' \n" \
") t2 \n" \
"WHERE t2.`fid` < 0"
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
def testSlicePrune(self):
expr = self.expr.filter(self.expr.fid < 0)[:4]['name',
lambda x: x.id + 1]
new_expr = ColumnPruning(expr.to_dag()).prune()
self.assertIsNotNone(new_expr.input.input.input._source_data)
expected = "SELECT t1.`name`, t1.`id` + 1 AS `id` \n" \
"FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
"WHERE t1.`fid` < 0 \n" \
"LIMIT 4"
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
def testGroupbyPrune(self):
expr = self.expr.groupby('name').agg(id=self.expr.id.max())
expr = expr[expr.id < 0]['name', ]
expected = "SELECT t1.`name` \n" \
"FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
"GROUP BY t1.`name` \n" \
"HAVING MAX(t1.`id`) < 0"
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr.groupby('name').agg(id=self.expr.id.max())
expr = expr[expr.id < 0]['id', ]
expected = "SELECT MAX(t1.`id`) AS `id` \n" \
"FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
"GROUP BY t1.`name` \n" \
"HAVING MAX(t1.`id`) < 0"
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
def testMutatePrune(self):
expr = self.expr[self.expr.exclude('birth'),
self.expr.fid.astype('int').rename('new_id')]
expr = expr[expr,
expr.groupby('name').
mutate(lambda x: x.new_id.cumsum().rename('new_id_sum'))]
expr = expr[expr.new_id, expr.new_id_sum]
expected = "SELECT t2.`new_id`, t2.`new_id_sum` \n" \
"FROM (\n" \
" SELECT CAST(t1.`fid` AS BIGINT) AS `new_id`, " \
"SUM(CAST(t1.`fid` AS BIGINT)) OVER (PARTITION BY t1.`name`) AS `new_id_sum` \n" \
" FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
") t2"
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
def testValueCountsPrune(self):
expr = self.expr.name.value_counts()['count', ]
new_expr = ColumnPruning(expr.to_dag()).prune()
self.assertIsInstance(new_expr.input.input, ProjectCollectionExpr)
self.assertEqual(set(new_expr.input.input.schema.names), set(['name']))
def testSortPrune(self):
expr = self.expr[self.expr.exclude('name'),
self.expr.name.rename('name2')].sort('name2')['id',
'fid']
expected = "SELECT t2.`id`, t2.`fid` \n" \
"FROM (\n" \
" SELECT t1.`id`, t1.`fid`, t1.`name` AS `name2` \n" \
" FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
" ORDER BY name2 \n" \
" LIMIT 10000\n" \
") t2"
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
def testDistinctPrune(self):
expr = self.expr.distinct(self.expr.id + 1, self.expr.name)['name', ]
expected = "SELECT t2.`name` \n" \
"FROM (\n" \
" SELECT DISTINCT t1.`id` + 1 AS `id`, t1.`name` \n" \
" FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
") t2"
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
def testSamplePrune(self):
expr = self.expr['name', 'id'].sample(parts=5)['id', ]
expected = "SELECT t1.`id` \n" \
"FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
"WHERE SAMPLE(5, 1)"
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
def testJoinPrune(self):
left = self.expr.select(self.expr, type='normal')
right = self.expr3[:4]
joined = left.left_join(right, on='id')
expr = joined.id_x.rename('id')
expected = "SELECT t2.`id` \n" \
"FROM (\n" \
" SELECT t1.`id` \n" \
" FROM mocked_project.`pyodps_test_expr_table` t1\n" \
") t2 \n" \
"LEFT OUTER JOIN \n" \
" (\n" \
" SELECT t3.`id` \n" \
" FROM mocked_project.`pyodps_test_expr_table2` t3 \n" \
" LIMIT 4\n" \
" ) t4\n" \
"ON t2.`id` == t4.`id`"
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
joined = self.expr.join(self.expr2, 'name')
expected = 'SELECT t1.`name`, t1.`id` AS `id_x`, t1.`fid` AS `fid_x`, ' \
't1.`isMale` AS `isMale_x`, t1.`scale` AS `scale_x`, ' \
't1.`birth` AS `birth_x`, t1.`ds` AS `ds_x`, t2.`id` AS `id_y`, ' \
't2.`fid` AS `fid_y`, t2.`isMale` AS `isMale_y`, t2.`scale` AS `scale_y`, ' \
't2.`birth` AS `birth_y`, t2.`ds` AS `ds_y` \n' \
'FROM mocked_project.`pyodps_test_expr_table` t1 \n' \
'INNER JOIN \n' \
' mocked_project.`pyodps_test_expr_table2` t2\n' \
'ON t1.`name` == t2.`name`'
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(joined, prettify=False)))
joined = self.expr.join(self.expr2,
on=[self.expr.name == self.expr2.name])
joined2 = joined.join(self.expr, on=[joined.id_x == self.expr.id])
expected = 'SELECT t1.`name` AS `name_x`, t1.`id` AS `id_x`, ' \
't1.`fid` AS `fid_x`, t1.`isMale` AS `isMale_x`, ' \
't1.`scale` AS `scale_x`, t1.`birth` AS `birth_x`, ' \
't1.`ds` AS `ds_x`, t2.`id` AS `id_y`, t2.`fid` AS `fid_y`, ' \
't2.`isMale` AS `isMale_y`, t2.`scale` AS `scale_y`, ' \
't2.`birth` AS `birth_y`, t2.`ds` AS `ds_y`, ' \
't3.`name` AS `name_y`, t3.`id`, t3.`fid`, t3.`isMale`, ' \
't3.`scale`, t3.`birth`, t3.`ds` \n' \
'FROM mocked_project.`pyodps_test_expr_table` t1 \n' \
'INNER JOIN \n' \
' mocked_project.`pyodps_test_expr_table2` t2\n' \
'ON t1.`name` == t2.`name` \n' \
'INNER JOIN \n' \
' mocked_project.`pyodps_test_expr_table` t3\n' \
'ON t1.`id` == t3.`id`'
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(joined2, prettify=False)))
joined = self.expr.join(self.expr2,
on=[self.expr.name == self.expr2.name],
mapjoin=True)
joined2 = joined.join(self.expr,
on=[joined.id_x == self.expr.id],
mapjoin=True)
expected = 'SELECT /*+mapjoin(t2,t3)*/ t1.`name` AS `name_x`, t1.`id` AS `id_x`, ' \
't1.`fid` AS `fid_x`, t1.`isMale` AS `isMale_x`, t1.`scale` AS `scale_x`, ' \
't1.`birth` AS `birth_x`, t1.`ds` AS `ds_x`, t2.`id` AS `id_y`, ' \
't2.`fid` AS `fid_y`, t2.`isMale` AS `isMale_y`, t2.`scale` AS `scale_y`, ' \
't2.`birth` AS `birth_y`, t2.`ds` AS `ds_y`, t3.`name` AS `name_y`, ' \
't3.`id`, t3.`fid`, t3.`isMale`, t3.`scale`, t3.`birth`, t3.`ds` \n' \
'FROM mocked_project.`pyodps_test_expr_table` t1 \n' \
'INNER JOIN \n' \
' mocked_project.`pyodps_test_expr_table2` t2\n' \
'ON t1.`name` == t2.`name` \n' \
'INNER JOIN \n' \
' mocked_project.`pyodps_test_expr_table` t3\n' \
'ON t1.`id` == t3.`id`'
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(joined2, prettify=False)))
def testUnionPrune(self):
left = self.expr.select('name', 'id')
right = self.expr3.select(
self.expr3.fid.astype('int').rename('id'), self.expr3.name)
expr = left.union(right)['id']
expected = "SELECT t3.`id` \n" \
"FROM (\n" \
" SELECT t1.`id` \n" \
" FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
" UNION ALL\n" \
" SELECT CAST(t2.`fid` AS BIGINT) AS `id` \n" \
" FROM mocked_project.`pyodps_test_expr_table2` t2\n" \
") t3"
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr.union(self.expr2)
expected = 'SELECT * \n' \
'FROM (\n' \
' SELECT * \n' \
' FROM mocked_project.`pyodps_test_expr_table` t1 \n' \
' UNION ALL\n' \
' SELECT * \n' \
' FROM mocked_project.`pyodps_test_expr_table2` t2\n' \
') t3'
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
class Test(TestBase):
def setup(self):
datatypes = lambda *types: [validate_data_type(t) for t in types]
schema = Schema.from_lists(
['name', 'id', 'fid', 'isMale', 'scale', 'birth'],
datatypes('string', 'int64', 'float64', 'boolean', 'decimal',
'datetime'), ['ds'], datatypes('string'))
table = MockTable(name='pyodps_test_expr_table', schema=schema)
self.expr = CollectionExpr(_source_data=table, _schema=schema)
table1 = MockTable(name='pyodps_test_expr_table1', schema=schema)
self.expr1 = CollectionExpr(_source_data=table1, _schema=schema)
table2 = MockTable(name='pyodps_test_expr_table2', schema=schema)
self.expr2 = CollectionExpr(_source_data=table2, _schema=schema)
schema2 = Schema.from_lists(['name', 'id', 'fid'],
datatypes('string', 'int64', 'float64'),
['part1', 'part2'],
datatypes('string', 'int64'))
table3 = MockTable(name='pyodps_test_expr_table2', schema=schema2)
self.expr3 = CollectionExpr(_source_data=table3, _schema=schema2)
schema3 = Schema.from_lists(['id', 'name', 'relatives', 'hobbies'],
datatypes('int64', 'string',
'dict<string, string>',
'list<string>'))
table4 = MockTable(name='pyodps_test_expr_table', schema=schema)
self.expr4 = CollectionExpr(_source_data=table4, _schema=schema3)
self.maxDiff = None
def testFilterPushdownThroughProjection(self):
expr = self.expr[self.expr.id + 1, 'name'][lambda x: x.id < 10]
expected = 'SELECT t1.`id` + 1 AS `id`, t1.`name` \n' \
'FROM mocked_project.`pyodps_test_expr_table` t1 \n' \
'WHERE (t1.`id` + 1) < 10'
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr[
self.expr.id + 1, 'name',
self.expr.name.isnull().rename('is_null')][lambda x: x.is_null]
expected = 'SELECT t1.`id` + 1 AS `id`, t1.`name`, t1.`name` IS NULL AS `is_null` \n' \
'FROM mocked_project.`pyodps_test_expr_table` t1 \n' \
'WHERE t1.`name` IS NULL'
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr['name', self.expr.id ** 2]\
.filter(lambda x: x.name == 'name1').filter(lambda x: x.id < 3)
expected = "SELECT t1.`name`, CAST(POW(t1.`id`, 2) AS BIGINT) AS `id` \n" \
"FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
"WHERE (t1.`name` == 'name1') AND ((CAST(POW(t1.`id`, 2) AS BIGINT)) < 3)"
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr['name', self.expr.id + 1].filter(
lambda x: x.name == 'name1')[lambda x: 'tt' + x.name,
'id'].filter(lambda x: x.id < 3)
expected = "SELECT CONCAT('tt', t1.`name`) AS `name`, t1.`id` + 1 AS `id` \n" \
"FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
"WHERE (t1.`name` == 'name1') AND ((t1.`id` + 1) < 3)"
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr.filter(self.expr.name == 'name1').select(
'name', lambda x: (x.id + 1) * 2)[lambda x: 'tt' + x.name,
'id'].filter(lambda x: x.id < 3)
expected = "SELECT CONCAT('tt', t1.`name`) AS `name`, (t1.`id` + 1) * 2 AS `id` \n" \
"FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
"WHERE (t1.`name` == 'name1') AND (((t1.`id` + 1) * 2) < 3)"
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr.filter(
self.expr.id.between(2, 6),
self.expr.name.lower().contains('pyodps',
regex=False)).name.nunique()
expected = "SELECT COUNT(DISTINCT t2.`name`) AS `name_nunique` \n" \
"FROM (\n" \
" SELECT t1.`name`, t1.`id` \n" \
" FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
" WHERE ((t1.`id` >= 2) AND (t1.`id` <= 6)) AND INSTR(TOLOWER(t1.`name`), 'pyodps') > 0 \n" \
") t2"
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
def testFilterPushDownThroughJoin(self):
expr = self.expr.join(self.expr3, on='name')
expr = expr[(expr.id_x < 10) & (expr.fid_y > 3)]
expected = 'SELECT t2.`name`, t2.`id` AS `id_x`, t2.`fid` AS `fid_x`, ' \
't2.`isMale`, t2.`scale`, t2.`birth`, t2.`ds`, t4.`id` AS `id_y`, ' \
't4.`fid` AS `fid_y`, t4.`part1`, t4.`part2` \n' \
'FROM (\n' \
' SELECT * \n' \
' FROM mocked_project.`pyodps_test_expr_table` t1 \n' \
' WHERE t1.`id` < 10\n' \
') t2 \n' \
'INNER JOIN \n' \
' (\n' \
' SELECT * \n' \
' FROM mocked_project.`pyodps_test_expr_table2` t3 \n' \
' WHERE t3.`fid` > 3\n' \
' ) t4\n' \
'ON t2.`name` == t4.`name`'
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr.join(self.expr3, on='name')
expr = expr[(expr.id_x < 10) & (expr.fid_y > 3) & (expr.id_x > 3)]
expected = 'SELECT t2.`name`, t2.`id` AS `id_x`, t2.`fid` AS `fid_x`, ' \
't2.`isMale`, t2.`scale`, t2.`birth`, t2.`ds`, t4.`id` AS `id_y`, ' \
't4.`fid` AS `fid_y`, t4.`part1`, t4.`part2` \n' \
'FROM (\n' \
' SELECT * \n' \
' FROM mocked_project.`pyodps_test_expr_table` t1 \n' \
' WHERE (t1.`id` < 10) AND (t1.`id` > 3)\n' \
') t2 \n' \
'INNER JOIN \n' \
' (\n' \
' SELECT * \n' \
' FROM mocked_project.`pyodps_test_expr_table2` t3 \n' \
' WHERE t3.`fid` > 3\n' \
' ) t4\n' \
'ON t2.`name` == t4.`name`'
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr[self.expr.name, self.expr.id + 1]
expr2 = self.expr3['tt' + self.expr3.name, self.expr3.id.rename('id2')]
expr = expr.join(expr2, on='name')
expr = expr[((expr.id < 10) | (expr.id > 100)) & (expr.id2 > 3)]
expected = "SELECT t2.`name`, t2.`id`, t4.`id2` \n" \
"FROM (\n" \
" SELECT t1.`name`, t1.`id` + 1 AS `id` \n" \
" FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
" WHERE ((t1.`id` + 1) < 10) OR ((t1.`id` + 1) > 100)\n" \
") t2 \n" \
"INNER JOIN \n" \
" (\n" \
" SELECT CONCAT('tt', t3.`name`) AS `name`, t3.`id` AS `id2` \n" \
" FROM mocked_project.`pyodps_test_expr_table2` t3 \n" \
" WHERE t3.`id` > 3\n" \
" ) t4\n" \
"ON t2.`name` == t4.`name`"
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr.join(self.expr3, on='name')
expr = expr[(expr.id_x + expr.id_y < 10) & (expr.id_x > 3)]
expected = "SELECT * \n" \
"FROM (\n" \
" SELECT t2.`name`, t2.`id` AS `id_x`, t2.`fid` AS `fid_x`, t2.`isMale`, " \
"t2.`scale`, t2.`birth`, t2.`ds`, t3.`id` AS `id_y`, " \
"t3.`fid` AS `fid_y`, t3.`part1`, t3.`part2` \n" \
" FROM (\n" \
" SELECT * \n" \
" FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
" WHERE t1.`id` > 3\n" \
" ) t2 \n" \
" INNER JOIN \n" \
" mocked_project.`pyodps_test_expr_table2` t3\n" \
" ON t2.`name` == t3.`name` \n" \
") t4 \n" \
"WHERE (t4.`id_x` + t4.`id_y`) < 10"
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr.outer_join(self.expr3, on='name')
expr = expr[(expr.id_x + expr.id_y < 10) & (expr.id_x > 3)]
expected = "SELECT * \n" \
"FROM (\n" \
" SELECT t1.`name` AS `name_x`, t1.`id` AS `id_x`, t1.`fid` AS `fid_x`, " \
"t1.`isMale`, t1.`scale`, t1.`birth`, t1.`ds`, t2.`name` AS `name_y`, " \
"t2.`id` AS `id_y`, t2.`fid` AS `fid_y`, t2.`part1`, t2.`part2` \n" \
" FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
" FULL OUTER JOIN \n" \
" mocked_project.`pyodps_test_expr_table2` t2\n" \
" ON t1.`name` == t2.`name` \n" \
") t3 \n" \
"WHERE ((t3.`id_x` + t3.`id_y`) < 10) AND (t3.`id_x` > 3)"
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr.join(self.expr3,
on=[
'name', self.expr.id == self.expr3.id,
self.expr.id < 10,
self.expr3.name == 'name1', self.expr.id > 5
])
expected = 'SELECT t2.`name`, t2.`id`, t2.`fid` AS `fid_x`, t2.`isMale`, ' \
't2.`scale`, t2.`birth`, t2.`ds`, t4.`fid` AS `fid_y`, t4.`part1`, t4.`part2` \n' \
'FROM (\n' \
' SELECT * \n' \
' FROM mocked_project.`pyodps_test_expr_table` t1 \n' \
' WHERE (t1.`id` < 10) AND (t1.`id` > 5)\n' \
') t2 \n' \
'INNER JOIN \n' \
' (\n' \
' SELECT * \n' \
' FROM mocked_project.`pyodps_test_expr_table2` t3 \n' \
' WHERE t3.`name` == \'name1\'\n' \
' ) t4\n' \
'ON (t2.`name` == t4.`name`) AND (t2.`id` == t4.`id`)'
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr.left_join(self.expr3,
on=[
'name', self.expr.id == self.expr3.id,
self.expr.id < 10,
self.expr3.name == 'name1',
self.expr.id > 5
])
expected = 'SELECT t1.`name` AS `name_x`, t1.`id` AS `id_x`, t1.`fid` AS `fid_x`, t1.`isMale`, ' \
't1.`scale`, t1.`birth`, t1.`ds`, t2.`name` AS `name_y`, t2.`id` AS `id_y`, ' \
't2.`fid` AS `fid_y`, t2.`part1`, t2.`part2` \n' \
'FROM mocked_project.`pyodps_test_expr_table` t1 \n' \
'LEFT OUTER JOIN \n' \
' mocked_project.`pyodps_test_expr_table2` t2\n' \
'ON ((((t1.`name` == t2.`name`) AND (t1.`id` == t2.`id`)) ' \
"AND (t1.`id` < 10)) AND (t2.`name` == 'name1')) AND (t1.`id` > 5)"
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
def testFilterPushdownThroughUnion(self):
expr = self.expr['name', 'id'].union(self.expr2['id', 'name'])
expr = expr.filter(expr.id + 1 < 3)
expected = 'SELECT * \n' \
'FROM (\n' \
' SELECT t1.`name`, t1.`id` \n' \
' FROM mocked_project.`pyodps_test_expr_table` t1 \n' \
' WHERE (t1.`id` + 1) < 3 \n' \
' UNION ALL\n' \
' SELECT t2.`name`, t2.`id` \n' \
' FROM mocked_project.`pyodps_test_expr_table2` t2 \n' \
' WHERE (t2.`id` + 1) < 3\n' \
') t3'
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr1 = self.expr.filter(self.expr.id == 1)['name', 'id']
expr2 = self.expr.filter(self.expr.id == 0)['id', 'name']
expr = expr1.union(expr2)
expected = 'SELECT * \n' \
'FROM (\n' \
' SELECT t1.`name`, t1.`id` \n' \
' FROM mocked_project.`pyodps_test_expr_table` t1 \n' \
' WHERE t1.`id` == 1 \n' \
' UNION ALL\n' \
' SELECT t2.`name`, t2.`id` \n' \
' FROM mocked_project.`pyodps_test_expr_table` t2 \n' \
' WHERE t2.`id` == 0\n' \
') t3'
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
def testGroupbyProjection(self):
expr = self.expr['id', 'name', 'fid']
expr2 = expr.groupby('name').agg(count=expr.count(), id=expr.id.sum())
expr3 = expr2['count', 'id']
expected = "SELECT COUNT(1) AS `count`, SUM(t1.`id`) AS `id` \n" \
"FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
"GROUP BY t1.`name`"
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr3, prettify=False)))
expr = self.expr['id', 'name', 'fid'].filter(self.expr.id < 10)['name',
'id']
expr2 = expr.groupby('name').agg(count=expr.count(),
id=expr.id.sum(),
name2=expr.name.max())
expr3 = expr2[expr2.count + 1, 'id']
expected = "SELECT COUNT(1) + 1 AS `count`, SUM(t1.`id`) AS `id` \n" \
"FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
"WHERE t1.`id` < 10 \n" \
"GROUP BY t1.`name`"
self.assertEqual(expected,
ODPSEngine(self.odps).compile(expr3, prettify=False))
def testFilterPushdownThroughMultipleProjection(self):
schema = Schema.from_lists(list('abcde'), ['string'] * 5)
table = MockTable(name='pyodps_test_expr_table3', schema=schema)
tab = CollectionExpr(_source_data=table,
_schema=odps_schema_to_df_schema(schema))
labels2 = []
bins2 = []
for i in range(0, 30):
a = str(7 * i) + '-' + str(7 * (i + 1))
b = 7 * i
bins2.append(b)
labels2.append(a)
p1 = tab.select(
tab.a,
tab.c.astype('int').cut(bins2, labels=labels2,
include_over=True).rename('c_cut'),
tab.e.astype('int').rename('e'),
tab.c.astype('int').rename('c'))
p1['f'] = p1['e'] / p1['c']
t = []
l = []
for i in range(0, 20):
a = 1 * i
b = str(a)
t.append(a)
l.append(b)
p2 = p1.select(
p1.a, p1.c_cut,
p1.f.cut(bins=t, labels=l, include_over=True).rename('f_cut'))
expected = "SELECT t1.`a`, CASE WHEN (0 < CAST(t1.`c` AS BIGINT)) " \
"AND (CAST(t1.`c` AS BIGINT) <= 7) THEN '0-7' " \
"WHEN (7 < CAST(t1.`c` AS BIGINT)) " \
"AND (CAST(t1.`c` AS BIGINT) <= 14) " \
"THEN '7-14' WHEN (14 < CAST(t1.`c` AS BIGINT)) " \
"AND (CAST(t1.`c` AS BIGINT) <= 21) THEN '14-21' " \
"WHEN (21 < CAST(t1.`c` AS BIGINT)) " \
"AND (CAST(t1.`c` AS BIGINT) <= 28) " \
"THEN '21-28' WHEN (28 < CAST(t1.`c` AS BIGINT)) " \
"AND (CAST(t1.`c` AS BIGINT) <= 35) THEN '28-35' " \
"WHEN (35 < CAST(t1.`c` AS BIGINT)) " \
"AND (CAST(t1.`c` AS BIGINT) <= 42) THEN '35-42' " \
"WHEN (42 < CAST(t1.`c` AS BIGINT)) " \
"AND (CAST(t1.`c` AS BIGINT) <= 49) THEN '42-49' " \
"WHEN (49 < CAST(t1.`c` AS BIGINT)) " \
"AND (CAST(t1.`c` AS BIGINT) <= 56) " \
"THEN '49-56' WHEN (56 < CAST(t1.`c` AS BIGINT)) " \
"AND (CAST(t1.`c` AS BIGINT) <= 63) THEN '56-63' " \
"WHEN (63 < CAST(t1.`c` AS BIGINT)) " \
"AND (CAST(t1.`c` AS BIGINT) <= 70) THEN '63-70' " \
"WHEN (70 < CAST(t1.`c` AS BIGINT)) AND (CAST(t1.`c` AS BIGINT) <= 77) " \
"THEN '70-77' WHEN (77 < CAST(t1.`c` AS BIGINT)) " \
"AND (CAST(t1.`c` AS BIGINT) <= 84) " \
"THEN '77-84' WHEN (84 < CAST(t1.`c` AS BIGINT)) " \
"AND (CAST(t1.`c` AS BIGINT) <= 91) THEN '84-91' " \
"WHEN (91 < CAST(t1.`c` AS BIGINT)) AND (CAST(t1.`c` AS BIGINT) <= 98) " \
"THEN '91-98' WHEN (98 < CAST(t1.`c` AS BIGINT)) " \
"AND (CAST(t1.`c` AS BIGINT) <= 105) THEN '98-105' " \
"WHEN (105 < CAST(t1.`c` AS BIGINT)) AND (CAST(t1.`c` AS BIGINT) <= 112) " \
"THEN '105-112' WHEN (112 < CAST(t1.`c` AS BIGINT)) " \
"AND (CAST(t1.`c` AS BIGINT) <= 119) THEN '112-119' " \
"WHEN (119 < CAST(t1.`c` AS BIGINT)) AND (CAST(t1.`c` AS BIGINT) <= 126) " \
"THEN '119-126' WHEN (126 < CAST(t1.`c` AS BIGINT)) " \
"AND (CAST(t1.`c` AS BIGINT) <= 133) THEN '126-133' " \
"WHEN (133 < CAST(t1.`c` AS BIGINT)) AND (CAST(t1.`c` AS BIGINT) <= 140) " \
"THEN '133-140' WHEN (140 < CAST(t1.`c` AS BIGINT)) " \
"AND (CAST(t1.`c` AS BIGINT) <= 147) THEN '140-147' " \
"WHEN (147 < CAST(t1.`c` AS BIGINT)) AND (CAST(t1.`c` AS BIGINT) <= 154) " \
"THEN '147-154' WHEN (154 < CAST(t1.`c` AS BIGINT)) " \
"AND (CAST(t1.`c` AS BIGINT) <= 161) THEN '154-161' " \
"WHEN (161 < CAST(t1.`c` AS BIGINT)) AND (CAST(t1.`c` AS BIGINT) <= 168) " \
"THEN '161-168' WHEN (168 < CAST(t1.`c` AS BIGINT)) " \
"AND (CAST(t1.`c` AS BIGINT) <= 175) THEN '168-175' " \
"WHEN (175 < CAST(t1.`c` AS BIGINT)) AND (CAST(t1.`c` AS BIGINT) <= 182) " \
"THEN '175-182' WHEN (182 < CAST(t1.`c` AS BIGINT)) " \
"AND (CAST(t1.`c` AS BIGINT) <= 189) THEN '182-189' " \
"WHEN (189 < CAST(t1.`c` AS BIGINT)) AND (CAST(t1.`c` AS BIGINT) <= 196) " \
"THEN '189-196' WHEN (196 < CAST(t1.`c` AS BIGINT)) " \
"AND (CAST(t1.`c` AS BIGINT) <= 203) THEN '196-203' " \
"WHEN 203 < CAST(t1.`c` AS BIGINT) THEN '203-210' END AS `c_cut`, " \
"CASE WHEN (0 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 1) THEN '0' " \
"WHEN (1 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 2) " \
"THEN '1' WHEN (2 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 3) THEN '2' " \
"WHEN (3 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 4) " \
"THEN '3' WHEN (4 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 5) THEN '4' " \
"WHEN (5 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 6) THEN '5' " \
"WHEN (6 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 7) " \
"THEN '6' WHEN (7 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 8) THEN '7' " \
"WHEN (8 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 9) THEN '8' " \
"WHEN (9 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 10) " \
"THEN '9' WHEN (10 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 11) THEN '10' " \
"WHEN (11 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 12) " \
"THEN '11' WHEN (12 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 13) THEN '12' " \
"WHEN (13 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 14) THEN '13' " \
"WHEN (14 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 15) THEN '14' " \
"WHEN (15 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 16) THEN '15' " \
"WHEN (16 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 17) THEN '16' " \
"WHEN (17 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 18) " \
"THEN '17' WHEN (18 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 19) THEN '18' " \
"WHEN 19 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) THEN '19' END AS `f_cut` \n" \
"FROM mocked_project.`pyodps_test_expr_table3` t1 \n" \
"WHERE (CASE WHEN (0 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 1) THEN '0' " \
"WHEN (1 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 2) " \
"THEN '1' WHEN (2 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 3) THEN '2' " \
"WHEN (3 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 4) THEN '3' " \
"WHEN (4 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 5) THEN '4' " \
"WHEN (5 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 6) THEN '5' " \
"WHEN (6 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 7) THEN '6' " \
"WHEN (7 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 8) THEN '7' " \
"WHEN (8 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 9) THEN '8' " \
"WHEN (9 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 10) THEN '9' " \
"WHEN (10 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 11) THEN '10' " \
"WHEN (11 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 12) THEN '11' " \
"WHEN (12 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 13) THEN '12' " \
"WHEN (13 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 14) THEN '13' " \
"WHEN (14 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 15) THEN '14' " \
"WHEN (15 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 16) THEN '15' " \
"WHEN (16 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 17) THEN '16' " \
"WHEN (17 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 18) THEN '17' " \
"WHEN (18 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT))) " \
"AND ((CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) <= 19) THEN '18' " \
"WHEN 19 < (CAST(t1.`e` AS BIGINT) / CAST(t1.`c` AS BIGINT)) THEN '19' END) == '9'"
self.assertEqual(
str(expected),
str(
ODPSEngine(self.odps).compile(p2[p2.f_cut == '9'],
prettify=False)))
def testFilterPushdownThroughLateralView(self):
expr = self.expr4[self.expr4.id, self.expr4.name,
self.expr4.hobbies.explode('hobby')]
expr2 = expr[expr.hobby.notnull() & (expr.id < 4)]
expected = 'SELECT * \n' \
'FROM (\n' \
' SELECT t2.`id`, t2.`name`, t3.`hobby` \n' \
' FROM (\n' \
' SELECT * \n' \
' FROM mocked_project.`pyodps_test_expr_table` t1 \n' \
' WHERE t1.`id` < 4\n' \
' ) t2 \n' \
' LATERAL VIEW EXPLODE(t2.`hobbies`) t3 AS `hobby` \n' \
') t4 \n' \
'WHERE t4.`hobby` IS NOT NULL'
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr2, prettify=False)))
expected = 'SELECT * \n' \
'FROM (\n' \
' SELECT t1.`id`, t1.`name`, t2.`hobby` \n' \
' FROM mocked_project.`pyodps_test_expr_table` t1 \n' \
' LATERAL VIEW EXPLODE(t1.`hobbies`) t2 AS `hobby` \n' \
') t3 \n' \
'WHERE (t3.`hobby` IS NOT NULL) OR (t3.`id` < 4)'
expr3 = expr[expr.hobby.notnull() | (expr.id < 4)]
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr3, prettify=False)))
def testFilterPushdownJoinFilterProjection(self):
expr = self.expr.join(self.expr3, on='id')
expr2 = expr[expr.scale < 5]
expr3 = expr2['name_x', 'id', 'fid_y']
expr4 = expr3.query('name_x > "b" and fid_y < 3')
expected = "SELECT t2.`name` AS `name_x`, t2.`id`, t4.`fid` AS `fid_y` \n" \
"FROM (\n" \
" SELECT t1.`name`, t1.`id`, t1.`scale` \n" \
" FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
" WHERE (t1.`name` > 'b') AND (t1.`scale` < 5)\n" \
") t2 \n" \
"INNER JOIN \n" \
" (\n" \
" SELECT t3.`id`, t3.`fid` \n" \
" FROM mocked_project.`pyodps_test_expr_table2` t3 \n" \
" WHERE t3.`fid` < 3\n" \
" ) t4\n" \
"ON t2.`id` == t4.`id`"
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr4, prettify=False)))
def testFilterPushdownJoinUnion(self):
expr = self.expr.union(self.expr1)
expr2 = expr.join(self.expr3, on='name', suffixes=('', '_2'))
expr3 = expr2.filter(expr2.id < 3, expr2.fid > 4, expr2.isMale)
expected = "SELECT t3.`name`, t3.`id`, t3.`fid`, t3.`isMale`, t3.`scale`, t3.`birth`, " \
"t4.`id` AS `id_2`, t4.`fid` AS `fid_2`, t4.`part1`, t4.`part2` \n" \
"FROM (\n" \
" SELECT * \n" \
" FROM (\n" \
" SELECT * \n" \
" FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
" WHERE ((t1.`fid` > 4) AND (t1.`id` < 3)) AND t1.`isMale` \n" \
" UNION ALL\n" \
" SELECT * \n" \
" FROM mocked_project.`pyodps_test_expr_table1` t2 \n" \
" WHERE ((t2.`fid` > 4) AND (t2.`id` < 3)) AND t2.`isMale`\n" \
" ) t3\n" \
") t3 \n" \
"INNER JOIN \n" \
" mocked_project.`pyodps_test_expr_table2` t4\n" \
"ON t3.`name` == t4.`name`"
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr3, prettify=False)))
class Test(TestBase):
def setup(self):
datatypes = lambda *types: [validate_data_type(t) for t in types]
schema = Schema.from_lists(
['name', 'id', 'fid', 'isMale', 'scale', 'birth'],
datatypes('string', 'int64', 'float64', 'boolean', 'decimal',
'datetime'), ['ds'], datatypes('string'))
table = MockTable(name='pyodps_test_expr_table', schema=schema)
self.expr = CollectionExpr(_source_data=table, _schema=schema)
table1 = MockTable(name='pyodps_test_expr_table1', schema=schema)
self.expr1 = CollectionExpr(_source_data=table1, _schema=schema)
table2 = MockTable(name='pyodps_test_expr_table2', schema=schema)
self.expr2 = CollectionExpr(_source_data=table2, _schema=schema)
schema2 = Schema.from_lists(['name', 'id', 'fid'],
datatypes('string', 'int64', 'float64'),
['part1', 'part2'],
datatypes('string', 'int64'))
table3 = MockTable(name='pyodps_test_expr_table2', schema=schema2)
self.expr3 = CollectionExpr(_source_data=table3, _schema=schema2)
self.maxDiff = None
def testFilterPushdownThroughProjection(self):
expr = self.expr[self.expr.id + 1, 'name'][lambda x: x.id < 10]
expected = 'SELECT t1.`id` + 1 AS `id`, t1.`name` \n' \
'FROM mocked_project.`pyodps_test_expr_table` t1 \n' \
'WHERE (t1.`id` + 1) < 10'
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr['name', self.expr.id ** 2]\
.filter(lambda x: x.name == 'name1').filter(lambda x: x.id < 3)
expected = "SELECT t1.`name`, CAST(POW(t1.`id`, 2) AS BIGINT) AS `id` \n" \
"FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
"WHERE (t1.`name` == 'name1') AND ((CAST(POW(t1.`id`, 2) AS BIGINT)) < 3)"
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr['name', self.expr.id + 1].filter(
lambda x: x.name == 'name1')[lambda x: 'tt' + x.name,
'id'].filter(lambda x: x.id < 3)
expected = "SELECT CONCAT('tt', t1.`name`) AS `name`, t1.`id` + 1 AS `id` \n" \
"FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
"WHERE (t1.`name` == 'name1') AND ((t1.`id` + 1) < 3)"
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr.filter(self.expr.name == 'name1').select(
'name', lambda x: (x.id + 1) * 2)[lambda x: 'tt' + x.name,
'id'].filter(lambda x: x.id < 3)
expected = "SELECT CONCAT('tt', t1.`name`) AS `name`, (t1.`id` + 1) * 2 AS `id` \n" \
"FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
"WHERE (((t1.`id` + 1) * 2) < 3) AND (t1.`name` == 'name1')"
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr.filter(
self.expr.id.between(2, 6),
self.expr.name.lower().contains('pyodps',
regex=False)).name.nunique()
expected = "SELECT COUNT(DISTINCT t2.`name`) AS `name_nunique` \n" \
"FROM (\n" \
" SELECT t1.`name`, t1.`id` \n" \
" FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
" WHERE ((t1.`id` >= 2) AND (t1.`id` <= 6)) AND INSTR(TOLOWER(t1.`name`), 'pyodps') > 0 \n" \
") t2"
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
def testFilterPushDownThroughJoin(self):
expr = self.expr.join(self.expr3, on='name')
expr = expr[(expr.id_x < 10) & (expr.fid_y > 3)]
expected = 'SELECT t2.`name`, t2.`id` AS `id_x`, t2.`fid` AS `fid_x`, ' \
't2.`isMale`, t2.`scale`, t2.`birth`, t2.`ds`, t4.`id` AS `id_y`, ' \
't4.`fid` AS `fid_y`, t4.`part1`, t4.`part2` \n' \
'FROM (\n' \
' SELECT * \n' \
' FROM mocked_project.`pyodps_test_expr_table` t1 \n' \
' WHERE t1.`id` < 10\n' \
') t2 \n' \
'INNER JOIN \n' \
' (\n' \
' SELECT * \n' \
' FROM mocked_project.`pyodps_test_expr_table2` t3 \n' \
' WHERE t3.`fid` > 3\n' \
' ) t4\n' \
'ON t2.`name` == t4.`name`'
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr.join(self.expr3, on='name')
expr = expr[(expr.id_x < 10) & (expr.fid_y > 3) & (expr.id_x > 3)]
expected = 'SELECT t2.`name`, t2.`id` AS `id_x`, t2.`fid` AS `fid_x`, ' \
't2.`isMale`, t2.`scale`, t2.`birth`, t2.`ds`, t4.`id` AS `id_y`, ' \
't4.`fid` AS `fid_y`, t4.`part1`, t4.`part2` \n' \
'FROM (\n' \
' SELECT * \n' \
' FROM mocked_project.`pyodps_test_expr_table` t1 \n' \
' WHERE (t1.`id` < 10) AND (t1.`id` > 3)\n' \
') t2 \n' \
'INNER JOIN \n' \
' (\n' \
' SELECT * \n' \
' FROM mocked_project.`pyodps_test_expr_table2` t3 \n' \
' WHERE t3.`fid` > 3\n' \
' ) t4\n' \
'ON t2.`name` == t4.`name`'
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr[self.expr.name, self.expr.id + 1]
expr2 = self.expr3['tt' + self.expr3.name, self.expr3.id.rename('id2')]
expr = expr.join(expr2, on='name')
expr = expr[((expr.id < 10) | (expr.id > 100)) & (expr.id2 > 3)]
expected = "SELECT t2.`name`, t2.`id`, t4.`id2` \n" \
"FROM (\n" \
" SELECT t1.`name`, t1.`id` + 1 AS `id` \n" \
" FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
" WHERE ((t1.`id` + 1) < 10) OR ((t1.`id` + 1) > 100)\n" \
") t2 \n" \
"INNER JOIN \n" \
" (\n" \
" SELECT CONCAT('tt', t3.`name`) AS `name`, t3.`id` AS `id2` \n" \
" FROM mocked_project.`pyodps_test_expr_table2` t3 \n" \
" WHERE t3.`id` > 3\n" \
" ) t4\n" \
"ON t2.`name` == t4.`name`"
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr.join(self.expr3, on='name')
expr = expr[(expr.id_x + expr.id_y < 10) & (expr.id_x > 3)]
expected = "SELECT * \n" \
"FROM (\n" \
" SELECT t2.`name`, t2.`id` AS `id_x`, t2.`fid` AS `fid_x`, t2.`isMale`, " \
"t2.`scale`, t2.`birth`, t2.`ds`, t3.`id` AS `id_y`, " \
"t3.`fid` AS `fid_y`, t3.`part1`, t3.`part2` \n" \
" FROM (\n" \
" SELECT * \n" \
" FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
" WHERE t1.`id` > 3\n" \
" ) t2 \n" \
" INNER JOIN \n" \
" mocked_project.`pyodps_test_expr_table2` t3\n" \
" ON t2.`name` == t3.`name` \n" \
") t4 \n" \
"WHERE (t4.`id_x` + t4.`id_y`) < 10"
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr.join(self.expr3,
on=[
'name', self.expr.id == self.expr3.id,
self.expr.id < 10,
self.expr3.name == 'name1', self.expr.id > 5
])
expected = 'SELECT t2.`name`, t2.`id`, t2.`fid` AS `fid_x`, t2.`isMale`, ' \
't2.`scale`, t2.`birth`, t2.`ds`, t4.`fid` AS `fid_y`, t4.`part1`, t4.`part2` \n' \
'FROM (\n' \
' SELECT * \n' \
' FROM mocked_project.`pyodps_test_expr_table` t1 \n' \
' WHERE (t1.`id` < 10) AND (t1.`id` > 5)\n' \
') t2 \n' \
'INNER JOIN \n' \
' (\n' \
' SELECT * \n' \
' FROM mocked_project.`pyodps_test_expr_table2` t3 \n' \
' WHERE t3.`name` == \'name1\'\n' \
' ) t4\n' \
'ON (t2.`name` == t4.`name`) AND (t2.`id` == t4.`id`)'
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
def testFilterPushdownThroughUnion(self):
expr = self.expr['name', 'id'].union(self.expr2['id', 'name'])
expr = expr.filter(expr.id + 1 < 3)
expected = 'SELECT * \n' \
'FROM (\n' \
' SELECT t1.`name`, t1.`id` \n' \
' FROM mocked_project.`pyodps_test_expr_table` t1 \n' \
' WHERE (t1.`id` + 1) < 3 \n' \
' UNION ALL\n' \
' SELECT t2.`name`, t2.`id` \n' \
' FROM mocked_project.`pyodps_test_expr_table2` t2 \n' \
' WHERE (t2.`id` + 1) < 3\n' \
') t3'
self.assertEqual(
to_str(expected),
to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
class Test(TestBase):
def setup(self):
datatypes = lambda *types: [validate_data_type(t) for t in types]
schema = Schema.from_lists(['name', 'id', 'fid', 'isMale', 'scale', 'birth'],
datatypes('string', 'int64', 'float64', 'boolean', 'decimal', 'datetime'),
['ds'], datatypes('string'))
table = MockTable(name='pyodps_test_expr_table', schema=schema)
self.expr = CollectionExpr(_source_data=table, _schema=Schema(columns=schema.columns))
table1 = MockTable(name='pyodps_test_expr_table1', schema=schema)
self.expr1 = CollectionExpr(_source_data=table1, _schema=Schema(columns=schema.columns))
table2 = MockTable(name='pyodps_test_expr_table2', schema=schema)
self.expr2 = CollectionExpr(_source_data=table2, _schema=Schema(columns=schema.columns))
schema2 = Schema.from_lists(['name', 'id', 'fid'], datatypes('string', 'int64', 'float64'),
['part1', 'part2'], datatypes('string', 'int64'))
table3 = MockTable(name='pyodps_test_expr_table2', schema=schema2)
self.expr3 = CollectionExpr(_source_data=table3, _schema=Schema(columns=schema2.columns))
def testProjectPrune(self):
expr = self.expr.select('name', 'id')
new_expr = ColumnPruning(expr.to_dag()).prune()
self.assertIsInstance(new_expr, ProjectCollectionExpr)
self.assertIsNotNone(new_expr.input._source_data)
expected = 'SELECT t1.`name`, t1.`id` \n' \
'FROM mocked_project.`pyodps_test_expr_table` t1'
self.assertEqual(expected, ODPSEngine(self.odps).compile(expr, prettify=False))
expr = self.expr[Scalar(3).rename('const'),
NullScalar('string').rename('string_const'),
self.expr.id]
expected = 'SELECT 3 AS `const`, CAST(NULL AS STRING) AS `string_const`, t1.`id` \n' \
'FROM mocked_project.`pyodps_test_expr_table` t1'
self.assertEqual(to_str(expected), to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr.select(pt=BuiltinFunction('max_pt', args=(self.expr._source_data.name,)))
expected = "SELECT max_pt('pyodps_test_expr_table') AS `pt` \n" \
"FROM mocked_project.`pyodps_test_expr_table` t1"
self.assertEqual(to_str(expected), to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
def testApplyPrune(self):
@output(['name', 'id'], ['string', 'string'])
def h(row):
yield row[0], row[1]
expr = self.expr[self.expr.fid < 0].apply(h, axis=1)['id', ]
new_expr = ColumnPruning(expr.to_dag()).prune()
self.assertIsInstance(new_expr, ProjectCollectionExpr)
self.assertIsInstance(new_expr.input.input, FilterCollectionExpr)
self.assertIsNotNone(new_expr.input.input.input._source_data)
def testFilterPrune(self):
expr = self.expr.filter(self.expr.name == 'name1')
expr = expr['name', 'id']
new_expr = ColumnPruning(expr.to_dag()).prune()
self.assertIsInstance(new_expr.input, FilterCollectionExpr)
self.assertNotIsInstance(new_expr.input.input, ProjectCollectionExpr)
self.assertIsNotNone(new_expr.input.input._source_data)
expected = 'SELECT t1.`name`, t1.`id` \n' \
'FROM mocked_project.`pyodps_test_expr_table` t1 \n' \
'WHERE t1.`name` == \'name1\''
self.assertEqual(to_str(expected), to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr.filter(self.expr.name == 'name1')
new_expr = ColumnPruning(expr.to_dag()).prune()
self.assertIsInstance(new_expr, FilterCollectionExpr)
self.assertIsNotNone(new_expr.input._source_data)
expr = self.expr.filter(self.expr.id.isin(self.expr3.id))
expected = 'SELECT * \n' \
'FROM mocked_project.`pyodps_test_expr_table` t1 \n' \
'WHERE t1.`id` IN (SELECT t3.`id` FROM ( ' \
'SELECT t2.`id` FROM mocked_project.`pyodps_test_expr_table2` t2 ) t3)'
self.assertTrue(to_str(expected), to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
def testFilterPartitionPrune(self):
expr = self.expr.filter_partition('ds=today')[lambda x: x.fid < 0][
'name', lambda x: x.id + 1]
new_expr = ColumnPruning(expr.to_dag()).prune()
self.assertEqual(set(new_expr.input.input.schema.names), set(['name', 'id', 'fid']))
expected = "SELECT t2.`name`, t2.`id` + 1 AS `id` \n" \
"FROM (\n" \
" SELECT t1.`name`, t1.`id`, t1.`fid` \n" \
" FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
" WHERE t1.`ds` == 'today' \n" \
") t2 \n" \
"WHERE t2.`fid` < 0"
self.assertEqual(to_str(expected), to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
def testSlicePrune(self):
expr = self.expr.filter(self.expr.fid < 0)[:4]['name', lambda x: x.id + 1]
new_expr = ColumnPruning(expr.to_dag()).prune()
self.assertIsNotNone(new_expr.input.input.input._source_data)
expected = "SELECT t1.`name`, t1.`id` + 1 AS `id` \n" \
"FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
"WHERE t1.`fid` < 0 \n" \
"LIMIT 4"
self.assertEqual(to_str(expected), to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
def testGroupbyPrune(self):
expr = self.expr.groupby('name').agg(id=self.expr.id.max())
expr = expr[expr.id < 0]['name', ]
expected = "SELECT t1.`name` \n" \
"FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
"GROUP BY t1.`name` \n" \
"HAVING MAX(t1.`id`) < 0"
self.assertEqual(to_str(expected), to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr.groupby('name').agg(id=self.expr.id.max())
expr = expr[expr.id < 0]['id',]
expected = "SELECT MAX(t1.`id`) AS `id` \n" \
"FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
"GROUP BY t1.`name` \n" \
"HAVING MAX(t1.`id`) < 0"
self.assertEqual(to_str(expected), to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
def testMutatePrune(self):
expr = self.expr[self.expr.exclude('birth'), self.expr.fid.astype('int').rename('new_id')]
expr = expr[expr, expr.groupby('name').mutate(lambda x: x.new_id.cumsum().rename('new_id_sum'))]
expr = expr[expr.new_id, expr.new_id_sum]
expected = "SELECT t2.`new_id`, t2.`new_id_sum` \n" \
"FROM (\n" \
" SELECT CAST(t1.`fid` AS BIGINT) AS `new_id`, " \
"SUM(CAST(t1.`fid` AS BIGINT)) OVER (PARTITION BY t1.`name`) AS `new_id_sum` \n" \
" FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
") t2"
self.assertEqual(to_str(expected), to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
def testValueCountsPrune(self):
expr = self.expr.name.value_counts()['count', ]
new_expr = ColumnPruning(expr.to_dag()).prune()
self.assertIsInstance(new_expr.input.input, ProjectCollectionExpr)
self.assertEqual(set(new_expr.input.input.schema.names), set(['name']))
def testSortPrune(self):
expr = self.expr[self.expr.exclude('name'), self.expr.name.rename('name2')].sort('name2')['id', 'fid']
expected = "SELECT t2.`id`, t2.`fid` \n" \
"FROM (\n" \
" SELECT t1.`id`, t1.`fid`, t1.`name` AS `name2` \n" \
" FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
" ORDER BY name2 \n" \
" LIMIT 10000\n" \
") t2"
self.assertEqual(to_str(expected), to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
def testDistinctPrune(self):
expr = self.expr.distinct(self.expr.id + 1, self.expr.name)['name', ]
expected = "SELECT t2.`name` \n" \
"FROM (\n" \
" SELECT DISTINCT t1.`id` + 1 AS `id`, t1.`name` \n" \
" FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
") t2"
self.assertEqual(to_str(expected), to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
def testSamplePrune(self):
expr = self.expr['name', 'id'].sample(parts=5)['id', ]
expected = "SELECT t1.`id` \n" \
"FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
"WHERE SAMPLE(5, 1)"
self.assertEqual(to_str(expected), to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
def testJoinPrune(self):
left = self.expr.select(self.expr, type='normal')
right = self.expr3[:4]
joined = left.left_join(right, on='id')
expr = joined.id_x.rename('id')
expected = "SELECT t2.`id` \n" \
"FROM (\n" \
" SELECT t1.`id` \n" \
" FROM mocked_project.`pyodps_test_expr_table` t1\n" \
") t2 \n" \
"LEFT OUTER JOIN \n" \
" (\n" \
" SELECT t3.`id` \n" \
" FROM mocked_project.`pyodps_test_expr_table2` t3 \n" \
" LIMIT 4\n" \
" ) t4\n" \
"ON t2.`id` == t4.`id`"
self.assertEqual(to_str(expected), to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
joined = self.expr.join(self.expr2, 'name')
expected = 'SELECT t1.`name`, t1.`id` AS `id_x`, t1.`fid` AS `fid_x`, ' \
't1.`isMale` AS `isMale_x`, t1.`scale` AS `scale_x`, ' \
't1.`birth` AS `birth_x`, t1.`ds` AS `ds_x`, t2.`id` AS `id_y`, ' \
't2.`fid` AS `fid_y`, t2.`isMale` AS `isMale_y`, t2.`scale` AS `scale_y`, ' \
't2.`birth` AS `birth_y`, t2.`ds` AS `ds_y` \n' \
'FROM mocked_project.`pyodps_test_expr_table` t1 \n' \
'INNER JOIN \n' \
' mocked_project.`pyodps_test_expr_table2` t2\n' \
'ON t1.`name` == t2.`name`'
self.assertEqual(to_str(expected), to_str(ODPSEngine(self.odps).compile(joined, prettify=False)))
def testUnionPrune(self):
left = self.expr.select('name', 'id')
right = self.expr3.select(self.expr3.fid.astype('int').rename('id'), self.expr3.name)
expr = left.union(right)['id']
expected = "SELECT t3.`id` \n" \
"FROM (\n" \
" SELECT t1.`id` \n" \
" FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
" UNION ALL\n" \
" SELECT CAST(t2.`fid` AS BIGINT) AS `id` \n" \
" FROM mocked_project.`pyodps_test_expr_table2` t2\n" \
") t3"
self.assertEqual(to_str(expected), to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr.union(self.expr2)
expected = 'SELECT * \n' \
'FROM (\n' \
' SELECT * \n' \
' FROM mocked_project.`pyodps_test_expr_table` t1 \n' \
' UNION ALL\n' \
' SELECT * \n' \
' FROM mocked_project.`pyodps_test_expr_table2` t2\n' \
') t3'
self.assertEqual(to_str(expected), to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
class Test(TestBase):
def setup(self):
datatypes = lambda *types: [validate_data_type(t) for t in types]
schema = Schema.from_lists(
['name', 'id', 'fid', 'isMale', 'birth', 'scale'][:5],
datatypes('string', 'int64', 'float64', 'boolean', 'datetime',
'decimal')[:5])
self.schema = df_schema_to_odps_schema(schema)
table_name = tn('pyodps_test_%s' % str(uuid.uuid4()).replace('-', '_'))
self.odps.delete_table(table_name, if_exists=True)
self.table = self.odps.create_table(name=table_name,
schema=self.schema)
self.expr = CollectionExpr(_source_data=self.table, _schema=schema)
self.engine = SeahawksEngine(self.odps)
class FakeBar(object):
def update(self, *args, **kwargs):
pass
def inc(self, *args, **kwargs):
pass
def status(self, *args, **kwargs):
pass
self.faked_bar = FakeBar()
def teardown(self):
self.table.drop()
def _gen_data(self,
rows=None,
data=None,
nullable_field=None,
value_range=None):
if data is None:
data = []
for _ in range(rows):
record = []
for t in self.schema.types:
method = getattr(self, '_gen_random_%s' % t.name)
if t.name == 'bigint':
record.append(method(value_range=value_range))
else:
record.append(method())
data.append(record)
if nullable_field is not None:
j = self.schema._name_indexes[nullable_field]
for i, l in enumerate(data):
if i % 2 == 0:
data[i][j] = None
self.odps.write_table(self.table, 0, data)
return data
def testAsync(self):
data = self._gen_data(10, value_range=(-1000, 1000))
expr = self.expr.id.sum()
future = self.engine.execute(expr, async=True)
self.assertFalse(future.done())
res = future.result()
self.assertEqual(sum(it[1] for it in data), res)
def testCache(self):
data = self._gen_data(10, value_range=(-1000, 1000))
expr = self.expr[self.expr.id < 10].cache()
cnt = expr.count()
dag = self.engine.compile(expr)
self.assertEqual(len(dag.nodes()), 2)
res = self.engine.execute(cnt)
self.assertEqual(len([it for it in data if it[1] < 10]), res)
self.assertTrue(context.is_cached(expr))
table = context.get_cached(expr)
self.assertIsInstance(table, SeahawksTable)
def testBatch(self):
data = self._gen_data(10, value_range=(-1000, 1000))
expr = self.expr[self.expr.id < 10].cache()
expr1 = expr.id.sum()
expr2 = expr.id.mean()
dag = self.engine.compile([expr1, expr2])
self.assertEqual(len(dag.nodes()), 3)
self.assertEqual(sum(len(v) for v in dag._graph.values()), 2)
expect1 = sum(d[1] for d in data if d[1] < 10)
length = len([d[1] for d in data if d[1] < 10])
expect2 = (expect1 / float(length)) if length > 0 else 0.0
res = self.engine.execute([expr1, expr2], n_parallel=2)
self.assertEqual(res[0], expect1)
self.assertAlmostEqual(res[1], expect2)
self.assertTrue(context.is_cached(expr))
# test async and timeout
expr = self.expr[self.expr.id < 10]
expr1 = expr.id.sum()
expr2 = expr.id.mean()
fs = self.engine.execute([expr, expr1, expr2],
n_parallel=2,
async=True,
timeout=1)
self.assertEqual(len(fs), 3)
self.assertEqual(fs[1].result(), expect1)
self.assertAlmostEqual(fs[2].result(), expect2)
self.assertTrue(context.is_cached(expr))
def testBase(self):
data = self._gen_data(10, value_range=(-1000, 1000))
expr = self.expr[self.expr.id < 10]['name', lambda x: x.id]
result = self._get_result(self.engine.execute(expr).values)
self.assertEqual(len([it for it in data if it[1] < 10]), len(result))
if len(result) > 0:
self.assertEqual(2, len(result[0]))
expr = self.expr[Scalar(3).rename('const'), self.expr.id,
(self.expr.id + 1).rename('id2')]
res = self.engine.execute(expr)
result = self._get_result(res.values)
self.assertEqual([c.name for c in res.columns], ['const', 'id', 'id2'])
self.assertTrue(all(it[0] == 3 for it in result))
self.assertEqual(len(data), len(result))
self.assertEqual([it[1] + 1 for it in data], [it[2] for it in result])
expr = self.expr.sort('id')[:5]
res = self.engine.execute(expr)
result = self._get_result(res.values)
self.assertListAlmostEqual(
sorted(data, key=lambda it: it[1])[:5],
[r[:-1] + [r[-1].replace(tzinfo=None)] for r in result],
only_float=False,
delta=.001)
expr = self.expr[:1].filter(lambda x: x.name == data[1][0])
res = self.engine.execute(expr)
self.assertEqual(len(res), 0)
def testChinese(self):
data = [
['中文', 4, 5.3, None, None],
['\'中文2', 2, 3.5, None, None],
]
self._gen_data(data=data)
expr = self.expr.filter(self.expr.name == '中文')
res = self.engine.execute(expr)
self.assertEqual(len(res), 1)
expr = self.expr.filter(self.expr.name == '\'中文2')
res = self.engine.execute(expr)
self.assertEqual(len(res), 1)
expr = self.expr.filter(self.expr.name == u'中文')
res = self.engine.execute(expr)
self.assertEqual(len(res), 1)
def testElement(self):
data = self._gen_data(5, nullable_field='name')
fields = [
self.expr.name.isnull().rename('name1'),
self.expr.name.notnull().rename('name2'),
self.expr.name.fillna('test').rename('name3'),
self.expr.id.isin([1, 2, 3]).rename('id1'),
self.expr.id.isin(self.expr.fid.astype('int')).rename('id2'),
self.expr.id.notin([1, 2, 3]).rename('id3'),
self.expr.id.notin(self.expr.fid.astype('int')).rename('id4'),
self.expr.id.between(self.expr.fid, 3).rename('id5'),
self.expr.name.fillna('test').switch(
'test',
'test' + self.expr.name.fillna('test'),
'test2',
'test2' + self.expr.name.fillna('test'),
default=self.expr.name).rename('name4'),
self.expr.name.fillna('test').switch('test', 1, 'test2',
2).rename('name5'),
self.expr.id.cut([100, 200, 300],
labels=['xsmall', 'small', 'large', 'xlarge'],
include_under=True,
include_over=True).rename('id6'),
self.expr.id.between(self.expr.fid, 3,
inclusive=False).rename('id7'),
]
expr = self.expr[fields]
res = self.engine.execute(expr)
result = self._get_result(res)
self.assertEqual(len(data), len(result))
self.assertEqual(len([it for it in data if it[0] is None]),
len([it[0] for it in result if it[0]]))
self.assertEqual(len([it[0] for it in data if it[0] is not None]),
len([it[1] for it in result if it[1]]))
self.assertEqual([(it[0] if it[0] is not None else 'test')
for it in data], [it[2] for it in result])
self.assertEqual([(it[1] in (1, 2, 3)) for it in data],
[it[3] for it in result])
fids = [int(it[2]) for it in data]
self.assertEqual([(it[1] in fids) for it in data],
[it[4] for it in result])
self.assertEqual([(it[1] not in (1, 2, 3)) for it in data],
[it[5] for it in result])
self.assertEqual([(it[1] not in fids) for it in data],
[it[6] for it in result])
self.assertEqual([(it[2] <= it[1] <= 3) for it in data],
[it[7] for it in result])
self.assertEqual(
[to_str('testtest' if it[0] is None else it[0]) for it in data],
[to_str(it[8]) for it in result])
self.assertEqual([to_str(1 if it[0] is None else None) for it in data],
[to_str(it[9]) for it in result])
def get_val(val):
if val <= 100:
return 'xsmall'
elif 100 < val <= 200:
return 'small'
elif 200 < val <= 300:
return 'large'
else:
return 'xlarge'
self.assertEqual([to_str(get_val(it[1])) for it in data],
[to_str(it[10]) for it in result])
self.assertEqual([(it[2] < it[1] < 3) for it in data],
[it[11] for it in result])
def testArithmetic(self):
data = self._gen_data(5, value_range=(-1000, 1000))
fields = [
(self.expr.id + 1).rename('id1'),
(self.expr.fid - 1).rename('fid1'),
(self.expr.id / 2).rename('id2'),
(self.expr.id**2).rename('id3'),
abs(self.expr.id).rename('id4'),
(~self.expr.id).rename('id5'),
(-self.expr.fid).rename('fid2'),
(~self.expr.isMale).rename('isMale1'),
(-self.expr.isMale).rename('isMale2'),
(self.expr.id // 2).rename('id6'),
(self.expr.id % 2).rename('id7'),
]
expr = self.expr[fields]
res = self.engine.execute(expr)
result = self._get_result(res)
self.assertEqual(len(data), len(result))
self.assertEqual([it[1] + 1 for it in data], [it[0] for it in result])
self.assertListAlmostEqual([it[2] - 1 for it in data],
[it[1] for it in result],
delta=.001)
self.assertListAlmostEqual([float(it[1]) / 2 for it in data],
[it[2] for it in result],
delta=.001)
self.assertEqual([int(it[1]**2) for it in data],
[it[3] for it in result])
self.assertEqual([abs(it[1]) for it in data], [it[4] for it in result])
self.assertEqual([~it[1] for it in data], [it[5] for it in result])
self.assertListAlmostEqual([-it[2] for it in data],
[it[6] for it in result],
delta=.001)
self.assertEqual([not it[3] for it in data], [it[7] for it in result])
self.assertEqual([it[1] // 2 for it in data], [it[9] for it in result])
self.assertEqual([it[1] % 2 for it in data], [it[10] for it in result])
def testMath(self):
# TODO: test sinh, cosh..., and acosh, asinh...
data = self._gen_data(5, value_range=(1, 90))
if hasattr(math, 'expm1'):
expm1 = math.expm1
else:
expm1 = lambda x: 2 * math.exp(x / 2.0) * math.sinh(x / 2.0)
methods_to_fields = [
(math.sin, self.expr.id.sin()),
(math.cos, self.expr.id.cos()),
(math.tan, self.expr.id.tan()),
(math.log, self.expr.id.log()),
(lambda v: math.log(v, 2), self.expr.id.log2()),
(math.log10, self.expr.id.log10()),
(math.log1p, self.expr.id.log1p()),
(math.exp, self.expr.id.exp()),
(expm1, self.expr.id.expm1()),
(math.atan, self.expr.id.arctan()),
(math.sqrt, self.expr.id.sqrt()),
(abs, self.expr.id.abs()),
(math.ceil, self.expr.id.ceil()),
(math.floor, self.expr.id.floor()),
(math.trunc, self.expr.id.trunc()),
]
fields = [
it[1].rename('id' + str(i))
for i, it in enumerate(methods_to_fields)
]
expr = self.expr[fields]
res = self.engine.execute(expr)
result = self._get_result(res)
for i, it in enumerate(methods_to_fields):
mt = it[0]
def method(v):
try:
return mt(v)
except ValueError:
return float('nan')
first = [method(it[1]) for it in data]
second = [it[i] for it in result]
self.assertEqual(len(first), len(second))
for it1, it2 in zip(first, second):
not_valid = lambda x: \
x is None or (isinstance(x, float) and (math.isnan(x) or math.isinf(x)))
if not_valid(it1) and not_valid(it2):
continue
if isinstance(it1, float) and it1 > 1.0e15:
scale = 0.1**(int(math.log10(it1)) - 15)
self.assertAlmostEqual(it1 * scale, it2 * scale, delta=8)
else:
self.assertAlmostEqual(it1, it2, delta=2)
def testString(self):
data = self._gen_data(5)
methods_to_fields = [
(lambda s: s.capitalize(), self.expr.name.capitalize()),
(lambda s: data[0][0] in s,
self.expr.name.contains(data[0][0], regex=False)),
(lambda s: s[0] + '|' + str(s[1]),
self.expr.name.cat(self.expr.id.astype('string'), sep='|')),
(lambda s: s.endswith(data[0][0]),
self.expr.name.endswith(data[0][0])),
(lambda s: s.startswith(data[0][0]),
self.expr.name.startswith(data[0][0])),
(lambda s: s.replace(data[0][0], 'test'),
self.expr.name.replace(data[0][0], 'test', regex=False)),
(lambda s: s[0], self.expr.name.get(0)),
(lambda s: len(s), self.expr.name.len()),
(lambda s: s.ljust(10), self.expr.name.ljust(10)),
(lambda s: s.ljust(20, '*'), self.expr.name.ljust(20,
fillchar='*')),
(lambda s: s.rjust(10), self.expr.name.rjust(10)),
(lambda s: s.rjust(20, '*'), self.expr.name.rjust(20,
fillchar='*')),
(lambda s: s * 4, self.expr.name.repeat(4)),
(lambda s: s[1:], self.expr.name.slice(1)),
(lambda s: s[1:6], self.expr.name.slice(1, 6)),
(lambda s: s.title(), self.expr.name.title()),
(lambda s: s.rjust(20, '0'), self.expr.name.zfill(20)),
]
fields = [
it[1].rename('id' + str(i))
for i, it in enumerate(methods_to_fields)
]
expr = self.expr[fields]
res = self.engine.execute(expr)
result = self._get_result(res)
for i, it in enumerate(methods_to_fields):
method = it[0]
if i != 2:
first = [method(it[0]) for it in data]
else:
# cat
first = [method(it) for it in data]
second = [it[i] for it in result]
self.assertEqual(first, second)
def testDatetime(self):
data = self._gen_data(5)
def date_value(sel):
if isinstance(sel, six.string_types):
fun = lambda v: getattr(v, sel)
else:
fun = sel
col_id = [
idx for idx, col in enumerate(self.schema.names)
if col == 'birth'
][0]
return [fun(row[col_id]) for row in data]
methods_to_fields = [
(partial(date_value, 'year'), self.expr.birth.year),
(partial(date_value, 'month'), self.expr.birth.month),
(partial(date_value, 'day'), self.expr.birth.day),
(partial(date_value, 'hour'), self.expr.birth.hour),
(partial(date_value, 'minute'), self.expr.birth.minute),
(partial(date_value, 'second'), self.expr.birth.second),
(partial(date_value, lambda d: d.isocalendar()[1]),
self.expr.birth.weekofyear),
(partial(date_value,
lambda d: d.weekday()), self.expr.birth.dayofweek),
(partial(date_value,
lambda d: d.weekday()), self.expr.birth.weekday),
(partial(date_value, lambda d: time.mktime(d.timetuple())),
self.expr.birth.unix_timestamp),
(partial(
date_value,
lambda d: datetime.combine(d.date(), datetime.min.time())),
self.expr.birth.date),
]
fields = [
it[1].rename('birth' + str(i))
for i, it in enumerate(methods_to_fields)
]
expr = self.expr[fields]
res = self.engine.execute(expr)
result = self._get_result(res)
for i, it in enumerate(methods_to_fields):
method = it[0]
first = method()
try:
import pandas as pd
def conv(v):
if isinstance(v, pd.Timestamp):
v = v.to_datetime()
if isinstance(v, datetime):
return v.replace(tzinfo=None)
return v
except ImportError:
conv = lambda v: v
second = [conv(it[i]) for it in result]
self.assertEqual(first, second)
def testSortDistinct(self):
data = [
['name1', 4, None, None, None],
['name2', 2, None, None, None],
['name1', 4, None, None, None],
['name1', 3, None, None, None],
]
self._gen_data(data=data)
expr = self.expr.sort(['name', -self.expr.id
]).distinct(['name', lambda x: x.id + 1])[:50]
res = self.engine.execute(expr)
result = self._get_result(res)
self.assertEqual(len(result), 3)
expected = [['name1', 5], ['name1', 4], ['name2', 3]]
self.assertEqual(sorted(expected), sorted(result))
def testPivotTable(self):
data = [['name1', 1, 1.0, True, None], ['name1', 1, 5.0, True, None],
['name1', 2, 2.0, True, None], ['name2', 1, 3.0, False, None],
['name2', 3, 4.0, False, None]]
self._gen_data(data=data)
expr = self.expr
expr1 = expr.pivot_table(rows='name', values='fid')
res = self.engine.execute(expr1)
result = self._get_result(res)
expected = [
['name1', 8.0 / 3],
['name2', 3.5],
]
self.assertListAlmostEqual(sorted(result),
sorted(expected),
only_float=False)
expr2 = expr.pivot_table(rows='name',
values='fid',
aggfunc=['mean', 'sum'])
res = self.engine.execute(expr2)
result = self._get_result(res)
expected = [
['name1', 8.0 / 3, 8.0],
['name2', 3.5, 7.0],
]
self.assertEqual(res.schema.names, ['name', 'fid_mean', 'fid_sum'])
self.assertListAlmostEqual(sorted(result),
sorted(expected),
only_float=False)
expr5 = expr.pivot_table(rows='id',
values='fid',
columns='name',
aggfunc=['mean', 'sum'])
expr6 = expr5['name1_fid_mean',
expr5.groupby(Scalar(1)).sort('name1_fid_mean').
name1_fid_mean.astype('float').cumsum()]
k = lambda x: list(0 if it is None else it for it in x)
expected = [[2, 2], [3, 5], [None, 5]]
res = self.engine.execute(expr6)
result = self._get_result(res)
self.assertEqual(sorted(result, key=k), sorted(expected, key=k))
expr3 = expr.pivot_table(rows='id',
values='fid',
columns='name',
fill_value=0).distinct()
res = self.engine.execute(expr3)
result = self._get_result(res)
expected = [
[2, 0, 2.0],
[3, 4.0, 0],
[1, 3.0, 3.0],
]
self.assertEqual(res.schema.names,
['id', 'name2_fid_mean', 'name1_fid_mean'])
self.assertEqual(result, expected)
expr7 = expr.pivot_table(rows='id',
values='fid',
columns='name',
aggfunc=['mean', 'sum']).cache()
self.assertEqual(len(self.engine.execute(expr7)), 3)
expr8 = self.expr.pivot_table(rows='id', values='fid', columns='name')
self.assertEqual(len(self.engine.execute(expr8)), 3)
self.assertNotIsInstance(expr8.schema, DynamicSchema)
expr9 = (expr8['name1_fid_mean'] -
expr8['name2_fid_mean']).rename('substract')
self.assertEqual(len(self.engine.execute(expr9)), 3)
expr10 = expr8.distinct()
self.assertEqual(len(self.engine.execute(expr10)), 3)
def testGroupbyAggregation(self):
data = [
['name1', 4, 5.3, None, None],
['name2', 2, 3.5, None, None],
['name1', 4, 4.2, None, None],
['name1', 3, 2.2, None, None],
['name1', 3, 4.1, None, None],
]
self._gen_data(data=data)
field = self.expr.groupby('name').sort(['id',
-self.expr.fid]).row_number()
expr = self.expr['name', 'id', 'fid', field]
res = self.engine.execute(expr)
result = self._get_result(res)
expected = [
['name1', 3, 4.1, 1],
['name1', 3, 2.2, 2],
['name1', 4, 5.3, 3],
['name1', 4, 4.2, 4],
['name2', 2, 3.5, 1],
]
result = sorted(result, key=lambda k: (k[0], k[1], -k[2]))
self.assertEqual(expected, result)
expr = self.expr.name.value_counts(dropna=True)[:25]
expected = [['name1', 4], ['name2', 1]]
res = self.engine.execute(expr)
result = self._get_result(res)
self.assertEqual(expected, result)
expr = self.expr.name.topk(25)
res = self.engine.execute(expr)
result = self._get_result(res)
self.assertEqual(expected, result)
expr = self.expr.groupby('name').count()
res = self.engine.execute(expr)
result = self._get_result(res)
self.assertEqual(sorted([it[1:] for it in expected]), sorted(result))
expected = [['name1', 2], ['name2', 1]]
expr = self.expr.groupby('name').id.nunique()
res = self.engine.execute(expr)
result = self._get_result(res)
self.assertEqual([it[1:] for it in expected], result)
expr = self.expr[self.expr['id'] > 2].name.value_counts()[:25]
expected = [['name1', 4]]
res = self.engine.execute(expr)
result = self._get_result(res)
self.assertEqual(expected, result)
expr = self.expr.groupby('name', Scalar(1).rename('constant')) \
.agg(id=self.expr.id.sum())
expected = [['name1', 1, 14], ['name2', 1, 2]]
res = self.engine.execute(expr)
result = self._get_result(res)
self.assertEqual(expected, result)
expr = self.expr[:1]
expr = expr.groupby('name').agg(expr.id.sum())
res = self.engine.execute(expr)
result = self._get_result(res)
expected = [['name1', 4]]
self.assertEqual(expected, result)
def testProjectionGroupbyFilter(self):
data = [
['name1', 4, 5.3, None, None],
['name2', 2, 3.5, None, None],
['name1', 4, 4.2, None, None],
['name1', 3, 2.2, None, None],
['name1', 3, 4.1, None, None],
]
self._gen_data(data=data)
df = self.expr.copy()
df['id'] = df.id + 1
df2 = df.groupby('name').agg(
id=df.id.sum())[lambda x: x.name == 'name2']
expected = [['name2', 3]]
res = self.engine.execute(df2)
result = self._get_result(res)
self.assertEqual(expected, result)
def testJoinGroupby(self):
data = [
['name1', 4, 5.3, None, None],
['name2', 2, 3.5, None, None],
['name1', 4, 4.2, None, None],
['name1', 3, 2.2, None, None],
['name1', 3, 4.1, None, None],
]
schema2 = Schema.from_lists(['name', 'id2', 'id3'],
[types.string, types.bigint, types.bigint])
table_name = tn('pyodps_test_engine_table2')
self.odps.delete_table(table_name, if_exists=True)
table2 = self.odps.create_table(name=table_name, schema=schema2)
expr2 = CollectionExpr(_source_data=table2,
_schema=odps_schema_to_df_schema(schema2))
self._gen_data(data=data)
data2 = [['name1', 4, -1], ['name2', 1, -2]]
self.odps.write_table(table2, 0, data2)
expr = self.expr.join(expr2, on='name')[self.expr]
expr = expr.groupby('id').agg(expr.fid.sum())
res = self.engine.execute(expr)
result = self._get_result(res)
id_idx = [
idx for idx, col in enumerate(self.expr.schema.names)
if col == 'id'
][0]
fid_idx = [
idx for idx, col in enumerate(self.expr.schema.names)
if col == 'fid'
][0]
expected = [[k, sum(
v[fid_idx] for v in row)] for k, row in itertools.groupby(
sorted(data, key=lambda r: r[id_idx]), lambda r: r[id_idx])]
for it in zip(sorted(expected, key=lambda it: it[0]),
sorted(result, key=lambda it: it[0])):
self.assertAlmostEqual(it[0][0], it[1][0])
self.assertAlmostEqual(it[0][1], it[1][1])
def testFilterGroupby(self):
data = [
['name1', 4, 5.3, None, None],
['name2', 2, 3.5, None, None],
['name1', 4, 4.2, None, None],
['name1', 3, 2.2, None, None],
['name1', 3, 4.1, None, None],
]
self._gen_data(data=data)
expr = self.expr.groupby(
['name']).agg(id=self.expr.id.max())[lambda x: x.id > 3]
res = self.engine.execute(expr)
result = self._get_result(res)
self.assertEqual(len(result), 1)
expected = [['name1', 4]]
self.assertEqual(expected, result)
def testWindowFunction(self):
data = [
['name1', 4, 5.3, None, None],
['name2', 2, 3.5, None, None],
['name1', 4, 4.2, None, None],
['name1', 3, 2.2, None, None],
['name1', 3, 6.1, None, None],
]
self._gen_data(data=data)
expr = self.expr.groupby('name').id.cumsum()
res = self.engine.execute(expr)
result = self._get_result(res)
expected = [[14]] * 4 + [[2]]
self.assertEqual(sorted(expected), sorted(result))
expr = self.expr.groupby('name').sort('fid').id.cummax()
res = self.engine.execute(expr)
result = self._get_result(res)
expected = [[3], [4], [4], [4], [2]]
self.assertEqual(sorted(expected), sorted(result))
expr = self.expr[
self.expr.groupby('name', 'id').sort('fid').id.cummean(), ]
res = self.engine.execute(expr)
result = self._get_result(res)
expected = [[3], [3], [4], [4], [2]]
self.assertEqual(sorted(expected), sorted(result))
expr = self.expr.groupby('name').mutate(
id2=lambda x: x.id.cumcount(),
fid2=lambda x: x.fid.cummin(sort='id'))
res = self.engine.execute(expr['name', 'id2', 'fid2'])
result = self._get_result(res)
expected = [
['name1', 4, 2.2],
['name1', 4, 2.2],
['name1', 4, 2.2],
['name1', 4, 2.2],
['name2', 1, 3.5],
]
self.assertEqual(sorted(expected), sorted(result))
expr = self.expr[
self.expr.id,
self.expr.groupby('name').rank('id'),
self.expr.groupby('name').dense_rank('fid', ascending=False),
self.expr.groupby('name').
row_number(sort=['id', 'fid'], ascending=[True, False]),
self.expr.groupby('name').percent_rank('id'), ]
res = self.engine.execute(expr)
result = self._get_result(res)
expected = [[4, 3, 2, 3, float(2) / 3], [2, 1, 1, 1, 0.0],
[4, 3, 3, 4, float(2) / 3], [3, 1, 4, 2,
float(0) / 3],
[3, 1, 1, 1, float(0) / 3]]
[
self.assertListAlmostEqual(l, r)
for l, r in zip(sorted(expected), sorted(result))
]
expr = self.expr[
self.expr.id,
self.expr.groupby('name').id.
lag(offset=3, default=0, sort=['id', 'fid']).rename('id2'),
self.expr.groupby('name').id.lead(offset=1,
default=-1,
sort=['id', 'fid'],
ascending=[False, False]
).rename('id3'), ]
res = self.engine.execute(expr)
result = self._get_result(res)
expected = [[4, 3, 4], [2, 0, -1], [4, 0, 3], [3, 0, -1], [3, 0, 3]]
self.assertEqual(sorted(expected), sorted(result))
def testWindowRewrite(self):
data = [
['name1', 4, 5.3, None, None],
['name2', 2, 3.5, None, None],
['name1', 4, 4.2, None, None],
['name1', 3, 2.2, None, None],
['name1', 3, 4.1, None, None],
]
self._gen_data(data=data)
expr = self.expr[self.expr.id - self.expr.id.mean() < 10][[
lambda x: x.id - x.id.max()
]][[lambda x: x.id - x.id.min()]][lambda x: x.id - x.id.std() > 0]
res = self.engine.execute(expr)
result = self._get_result(res)
id_idx = [
idx for idx, col in enumerate(self.expr.schema.names)
if col == 'id'
][0]
expected = [r[id_idx] for r in data]
maxv = max(expected)
expected = [v - maxv for v in expected]
minv = min(expected)
expected = [v - minv for v in expected]
meanv = sum(expected) * 1.0 / len(expected)
meanv2 = sum([v**2 for v in expected]) * 1.0 / len(expected)
std = math.sqrt(meanv2 - meanv**2)
expected = [v for v in expected if v > std]
self.assertEqual(expected, [it[0] for it in result])
def testReduction(self):
data = self._gen_data(rows=5, value_range=(-100, 100))
def stats(col, func):
col_idx = [
idx for idx, cn in enumerate(self.expr.schema.names)
if cn == col
][0]
return func([r[col_idx] for r in data])
def var(vct, ddof=0):
meanv = mean(vct)
meanv2 = mean([v**2 for v in vct])
return (meanv2 - meanv**2) * len(vct) / (len(vct) - ddof)
def moment(vct, order, central=False, absolute=False):
abs_fun = abs if absolute else lambda x: x
if central:
m = mean(vct)
return mean([abs_fun(v - m)**order for v in vct])
else:
return mean([abs_fun(v)**order for v in vct])
def skew(vct):
n = len(vct)
return moment(vct, 3, central=True) / (std(
vct, 1)**3) * (n**2) / (n - 1) / (n - 2)
def kurtosis(vct):
n = len(vct)
m4 = moment(vct, 4, central=True)
m2 = var(vct, 0)
return 1.0 / (n - 2) / (n - 3) * ((n * n - 1.0) * m4 / m2**2 - 3 *
(n - 1)**2)
mean = lambda v: sum(v) * 1.0 / len(v)
std = lambda v, ddof=0: math.sqrt(var(v, ddof))
nunique = lambda v: len(set(v))
cat = lambda v: len([it for it in v if it is not None])
methods_to_fields = [
(partial(stats, 'id', mean), self.expr.id.mean()),
(partial(len, data), self.expr.count()),
(partial(stats, 'id', var), self.expr.id.var(ddof=0)),
(partial(stats, 'id',
lambda x: var(x, 1)), self.expr.id.var(ddof=1)),
(partial(stats, 'id', std), self.expr.id.std(ddof=0)),
(partial(stats, 'id', lambda x: moment(x, 3, central=True)),
self.expr.id.moment(3, central=True)),
(partial(stats, 'id', skew), self.expr.id.skew()),
(partial(stats, 'id', kurtosis), self.expr.id.kurtosis()),
(partial(stats, 'id', sum), self.expr.id.sum()),
(partial(stats, 'id', min), self.expr.id.min()),
(partial(stats, 'id', max), self.expr.id.max()),
(partial(stats, 'isMale', min), self.expr.isMale.min()),
(partial(stats, 'isMale', sum), self.expr.isMale.sum()),
(partial(stats, 'isMale', any), self.expr.isMale.any()),
(partial(stats, 'isMale', all), self.expr.isMale.all()),
(partial(stats, 'name', nunique), self.expr.name.nunique()),
(partial(stats, 'name', cat), self.expr.name.cat(sep='|')),
(partial(stats, 'id', lambda x: len(x)), self.expr.id.count()),
]
fields = [
it[1].rename('f' + str(i))
for i, it in enumerate(methods_to_fields)
]
expr = self.expr[fields]
res = self.engine.execute(expr)
result = self._get_result(res)
for i, it in enumerate(methods_to_fields):
method = it[0]
first = method()
second = [it[i] for it in result][0]
if i == len(methods_to_fields) - 2: # cat
second = len(second.split('|'))
if isinstance(first, float):
self.assertAlmostEqual(first, second)
else:
if first != second:
pass
self.assertEqual(first, second)
def testJoin(self):
data = [
['name1', 4, 5.3, None, None],
['name2', 2, 3.5, None, None],
['name1', 4, 4.2, None, None],
['name1', 3, 2.2, None, None],
['name1', 3, 4.1, None, None],
]
schema2 = Schema.from_lists(['name', 'id2', 'id3'],
[types.string, types.bigint, types.bigint])
table_name = tn('pyodps_test_engine_table2')
self.odps.delete_table(table_name, if_exists=True)
table2 = self.odps.create_table(name=table_name, schema=schema2)
expr2 = CollectionExpr(_source_data=table2,
_schema=odps_schema_to_df_schema(schema2))
self._gen_data(data=data)
data2 = [['name1', 4, -1], ['name2', 1, -2]]
self.odps.write_table(table2, 0, data2)
try:
expr = self.expr.join(expr2)['name', 'id2']
res = self.engine.execute(expr)
result = self._get_result(res)
self.assertEqual(len(result), 5)
expected = [[to_str('name1'), 4], [to_str('name2'), 1]]
self.assertTrue(all(it in expected for it in result))
expr = self.expr.join(expr2, on=['name',
('id', 'id2')])[self.expr.name,
expr2.id2]
res = self.engine.execute(expr)
result = self._get_result(res)
self.assertEqual(len(result), 2)
expected = [to_str('name1'), 4]
self.assertTrue(all(it == expected for it in result))
expr = self.expr.left_join(expr2,
on=['name',
('id', 'id2')])[self.expr.name,
expr2.id2]
res = self.engine.execute(expr)
result = self._get_result(res)
expected = [['name1', 4], ['name2', None], ['name1', 4],
['name1', None], ['name1', None]]
self.assertEqual(len(result), 5)
self.assertTrue(all(it in expected for it in result))
expr = self.expr.right_join(expr2,
on=['name',
('id', 'id2')])[self.expr.name,
expr2.id2]
res = self.engine.execute(expr)
result = self._get_result(res)
expected = [
['name1', 4],
['name1', 4],
[None, 1],
]
self.assertEqual(len(result), 3)
self.assertTrue(all(it in expected for it in result))
expr = self.expr.outer_join(expr2,
on=['name',
('id', 'id2')])[self.expr.name,
expr2.id2]
res = self.engine.execute(expr)
result = self._get_result(res)
expected = [
['name1', 4],
['name1', 4],
['name2', None],
['name1', None],
['name1', None],
[None, 1],
]
self.assertEqual(len(result), 6)
self.assertTrue(all(it in expected for it in result))
grouped = self.expr.groupby('name').agg(
new_id=self.expr.id.sum()).cache()
self.engine.execute(self.expr.join(grouped, on='name'))
expr = self.expr.join(expr2, on=[
'name', ('id', 'id2')
])[lambda x: x.groupby(Scalar(1)).sort('name').row_number(), ]
self.engine.execute(expr)
finally:
table2.drop()
def testUnion(self):
data = [
['name1', 4, 5.3, None, None],
['name2', 2, 3.5, None, None],
['name1', 4, 4.2, None, None],
['name1', 3, 2.2, None, None],
['name1', 3, 4.1, None, None],
]
schema2 = Schema.from_lists(['name', 'id2', 'id3'],
[types.string, types.bigint, types.bigint])
table_name = tn('pyodps_test_engine_table2')
self.odps.delete_table(table_name, if_exists=True)
table2 = self.odps.create_table(name=table_name, schema=schema2)
expr2 = CollectionExpr(_source_data=table2,
_schema=odps_schema_to_df_schema(schema2))
self._gen_data(data=data)
data2 = [['name3', 5, -1], ['name4', 6, -2]]
self.odps.write_table(table2, 0, data2)
try:
expr = self.expr['name', 'id'].distinct().union(
expr2[expr2.id2.rename('id'), 'name'])
res = self.engine.execute(expr)
result = self._get_result(res)
expected = [['name1', 4], ['name1', 3], ['name2', 2], ['name3', 5],
['name4', 6]]
result = sorted(result)
expected = sorted(expected)
self.assertEqual(len(result), len(expected))
for e, r in zip(result, expected):
self.assertEqual([to_str(t) for t in e],
[to_str(t) for t in r])
finally:
table2.drop()
def testScaleValue(self):
data = [
['name1', 4, 5.3],
['name2', 2, 3.5],
['name1', 4, 4.2],
['name1', 3, 2.2],
['name1', 3, 4.1],
]
schema = Schema.from_lists(['name', 'id', 'fid'],
[types.string, types.bigint, types.double])
table_name = tn('pyodps_test_engine_scale_table')
self.odps.delete_table(table_name, if_exists=True)
table = self.odps.create_table(name=table_name, schema=schema)
self.odps.write_table(table_name, 0, data)
expr_input = CollectionExpr(_source_data=table,
_schema=odps_schema_to_df_schema(schema))
expr = expr_input.min_max_scale(columns=['fid'])
res = self.engine.execute(expr)
result = self._get_result(res)
expected = [['name1', 4, 1.0], ['name2', 2, 0.41935483870967744],
['name1', 4, 0.6451612903225807], ['name1', 3, 0.0],
['name1', 3, 0.6129032258064515]]
result = sorted(result)
expected = sorted(expected)
for first, second in zip(result, expected):
self.assertEqual(len(first), len(second))
for it1, it2 in zip(first, second):
self.assertAlmostEqual(it1, it2)
expr = expr_input.std_scale(columns=['fid'])
res = self.engine.execute(expr)
result = self._get_result(res)
expected = [['name1', 4, 1.4213602653434203],
['name2', 2, -0.3553400663358544],
['name1', 4, 0.3355989515394193],
['name1', 3, -1.6385125281042194],
['name1', 3, 0.23689337755723686]]
result = sorted(result)
expected = sorted(expected)
for first, second in zip(result, expected):
self.assertEqual(len(first), len(second))
for it1, it2 in zip(first, second):
self.assertAlmostEqual(it1, it2)
def testPersist(self):
data = [
['name1', 4, 5.3, None, None],
['name2', 2, 3.5, None, None],
['name1', 4, 4.2, None, None],
['name1', 3, 2.2, None, None],
['name1', 3, 4.1, None, None],
]
self._gen_data(data=data)
table_name = tn('pyodps_test_engine_persist_seahawks_table')
try:
df = self.engine.persist(self.expr, table_name)
res = self.engine.execute(df)
result = self._get_result(res)
self.assertEqual(len(result), 5)
self.assertEqual(data, result)
finally:
self.odps.delete_table(table_name, if_exists=True)
try:
schema = Schema.from_lists(self.schema.names, self.schema.types,
['ds'], ['string'])
self.odps.create_table(table_name, schema)
df = self.engine.persist(self.expr,
table_name,
partition='ds=today',
create_partition=True)
res = self.engine.execute(df)
result = self._get_result(res)
self.assertEqual(len(result), 5)
self.assertEqual(data, [r[:-1] for r in result])
finally:
self.odps.delete_table(table_name, if_exists=True)
try:
self.engine.persist(self.expr, table_name, partitions=['name'])
t = self.odps.get_table(table_name)
self.assertEqual(2, len(list(t.partitions)))
with t.open_reader(partition='name=name1', reopen=True) as r:
self.assertEqual(4, r.count)
with t.open_reader(partition='name=name2', reopen=True) as r:
self.assertEqual(1, r.count)
finally:
self.odps.delete_table(table_name, if_exists=True)
def testMakeKV(self):
from odps import types as odps_types
data = [
['name1', 1.0, 3.0, None, 10.0, None, None],
['name1', None, 3.0, 5.1, None, None, None],
['name1', 7.1, None, None, None, 8.2, None],
['name2', None, 1.2, 1.5, None, None, None],
['name2', None, 1.0, None, None, None, 1.1],
]
kv_cols = ['k1', 'k2', 'k3', 'k5', 'k7', 'k9']
schema = Schema.from_lists(['name'] + kv_cols, [odps_types.string] +
[odps_types.double] * 6)
table_name = tn('pyodps_test_engine_make_kv')
self.odps.delete_table(table_name, if_exists=True)
table = self.odps.create_table(name=table_name, schema=schema)
expr = CollectionExpr(_source_data=table,
_schema=odps_schema_to_df_schema(schema))
try:
self.odps.write_table(table, 0, data)
expr1 = expr.to_kv(columns=kv_cols, kv_delim='=')
res = self.engine.execute(expr1)
result = self._get_result(res)
expected = [
['name1', 'k1=1,k2=3,k5=10'],
['name1', 'k2=3,k3=5.1'],
['name1', 'k1=7.1,k7=8.2'],
['name2', 'k2=1.2,k3=1.5'],
['name2', 'k2=1,k9=1.1'],
]
self.assertListEqual(result, expected)
finally:
table.drop()
def testFilterOrder(self):
table_name = tn('pyodps_test_division_error')
self.odps.delete_table(table_name, if_exists=True)
table = self.odps.create_table(table_name,
'divided bigint, divisor bigint',
lifecycle=1)
try:
self.odps.write_table(table_name,
[[2, 0], [1, 1], [1, 2], [5, 1], [5, 0]])
df = CollectionExpr(_source_data=table,
_schema=odps_schema_to_df_schema(table.schema))
fdf = df[df.divisor > 0]
ddf = fdf[(fdf.divided / fdf.divisor).rename('result'), ]
expr = ddf[ddf.result > 1]
res = self.engine.execute(expr)
result = self._get_result(res)
self.assertEqual(result, [[
5,
]])
finally:
table.drop()
def testAXFException(self):
import sqlalchemy
data = [
['name1', 4, 5.3, None, None],
['name2', 2, 3.5, None, None],
['name1', 4, 4.2, None, None],
['name1', 3, 2.2, None, None],
['name1', 3, 4.1, None, None],
]
self._gen_data(data=data)
table_name = tn('pyodps_test_engine_axf_seahawks_table')
try:
schema = Schema.from_lists(self.schema.names, self.schema.types,
['ds'], ['string'])
self.odps.create_table(table_name, schema)
df = self.engine.persist(self.expr,
table_name,
partition='ds=today',
create_partition=True)
with self.assertRaises(sqlalchemy.exc.DatabaseError):
self.engine.execute(df.input)
finally:
self.odps.delete_table(table_name, if_exists=True)
class Test(TestBase):
def setup(self):
datatypes = lambda *types: [validate_data_type(t) for t in types]
schema = Schema.from_lists(['name', 'id', 'fid', 'isMale', 'scale', 'birth'],
datatypes('string', 'int64', 'float64', 'boolean', 'decimal', 'datetime'),
['ds'], datatypes('string'))
table = MockTable(name='pyodps_test_expr_table', schema=schema)
self.expr = CollectionExpr(_source_data=table, _schema=schema)
table1 = MockTable(name='pyodps_test_expr_table1', schema=schema)
self.expr1 = CollectionExpr(_source_data=table1, _schema=schema)
table2 = MockTable(name='pyodps_test_expr_table2', schema=schema)
self.expr2 = CollectionExpr(_source_data=table2, _schema=schema)
schema2 = Schema.from_lists(['name', 'id', 'fid'], datatypes('string', 'int64', 'float64'),
['part1', 'part2'], datatypes('string', 'int64'))
table3 = MockTable(name='pyodps_test_expr_table2', schema=schema2)
self.expr3 = CollectionExpr(_source_data=table3, _schema=schema2)
self.maxDiff = None
def testFilterPushdownThroughProjection(self):
expr = self.expr[self.expr.id + 1, 'name'][lambda x: x.id < 10]
expected = 'SELECT t1.`id` + 1 AS `id`, t1.`name` \n' \
'FROM mocked_project.`pyodps_test_expr_table` t1 \n' \
'WHERE (t1.`id` + 1) < 10'
self.assertEqual(to_str(expected), to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr['name', self.expr.id ** 2]\
.filter(lambda x: x.name == 'name1').filter(lambda x: x.id < 3)
expected = "SELECT t1.`name`, CAST(POW(t1.`id`, 2) AS BIGINT) AS `id` \n" \
"FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
"WHERE (t1.`name` == 'name1') AND ((CAST(POW(t1.`id`, 2) AS BIGINT)) < 3)"
self.assertEqual(to_str(expected), to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr['name', self.expr.id + 1].filter(lambda x: x.name == 'name1')[
lambda x: 'tt' + x.name, 'id'
].filter(lambda x: x.id < 3)
expected = "SELECT CONCAT('tt', t1.`name`) AS `name`, t1.`id` + 1 AS `id` \n" \
"FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
"WHERE (t1.`name` == 'name1') AND ((t1.`id` + 1) < 3)"
self.assertEqual(to_str(expected), to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr.filter(self.expr.name == 'name1').select('name', lambda x: (x.id + 1) * 2)[
lambda x: 'tt' + x.name, 'id'
].filter(lambda x: x.id < 3)
expected = "SELECT CONCAT('tt', t1.`name`) AS `name`, (t1.`id` + 1) * 2 AS `id` \n" \
"FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
"WHERE (((t1.`id` + 1) * 2) < 3) AND (t1.`name` == 'name1')"
self.assertEqual(to_str(expected), to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr.filter(self.expr.id.between(2, 6),
self.expr.name.lower().contains('pyodps', regex=False)).name.nunique()
expected = "SELECT COUNT(DISTINCT t2.`name`) AS `name_nunique` \n" \
"FROM (\n" \
" SELECT t1.`id`, t1.`name` \n" \
" FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
" WHERE ((t1.`id` >= 2) AND (t1.`id` <= 6)) AND INSTR(TOLOWER(t1.`name`), 'pyodps') > 0 \n" \
") t2"
self.assertEqual(to_str(expected), to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
def testFilterPushDownThroughJoin(self):
expr = self.expr.join(self.expr3, on='name')
expr = expr[(expr.id_x < 10) & (expr.fid_y > 3)]
expected = 'SELECT t2.`name`, t2.`id` AS `id_x`, t2.`fid` AS `fid_x`, ' \
't2.`isMale`, t2.`scale`, t2.`birth`, t2.`ds`, t4.`id` AS `id_y`, ' \
't4.`fid` AS `fid_y`, t4.`part1`, t4.`part2` \n' \
'FROM (\n' \
' SELECT * \n' \
' FROM mocked_project.`pyodps_test_expr_table` t1 \n' \
' WHERE t1.`id` < 10\n' \
') t2 \n' \
'INNER JOIN \n' \
' (\n' \
' SELECT * \n' \
' FROM mocked_project.`pyodps_test_expr_table2` t3 \n' \
' WHERE t3.`fid` > 3\n' \
' ) t4\n' \
'ON t2.`name` == t4.`name`'
self.assertEqual(to_str(expected), to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr.join(self.expr3, on='name')
expr = expr[(expr.id_x < 10) & (expr.fid_y > 3) & (expr.id_x > 3)]
expected = 'SELECT t2.`name`, t2.`id` AS `id_x`, t2.`fid` AS `fid_x`, ' \
't2.`isMale`, t2.`scale`, t2.`birth`, t2.`ds`, t4.`id` AS `id_y`, ' \
't4.`fid` AS `fid_y`, t4.`part1`, t4.`part2` \n' \
'FROM (\n' \
' SELECT * \n' \
' FROM mocked_project.`pyodps_test_expr_table` t1 \n' \
' WHERE (t1.`id` < 10) AND (t1.`id` > 3)\n' \
') t2 \n' \
'INNER JOIN \n' \
' (\n' \
' SELECT * \n' \
' FROM mocked_project.`pyodps_test_expr_table2` t3 \n' \
' WHERE t3.`fid` > 3\n' \
' ) t4\n' \
'ON t2.`name` == t4.`name`'
self.assertEqual(to_str(expected), to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr[self.expr.name, self.expr.id + 1]
expr2 = self.expr3['tt' + self.expr3.name, self.expr3.id.rename('id2')]
expr = expr.join(expr2, on='name')
expr = expr[((expr.id < 10) | (expr.id > 100)) & (expr.id2 > 3)]
expected = "SELECT t2.`name`, t2.`id`, t4.`id2` \n" \
"FROM (\n" \
" SELECT t1.`name`, t1.`id` + 1 AS `id` \n" \
" FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
" WHERE ((t1.`id` + 1) < 10) OR ((t1.`id` + 1) > 100)\n" \
") t2 \n" \
"INNER JOIN \n" \
" (\n" \
" SELECT CONCAT('tt', t3.`name`) AS `name`, t3.`id` AS `id2` \n" \
" FROM mocked_project.`pyodps_test_expr_table2` t3 \n" \
" WHERE t3.`id` > 3\n" \
" ) t4\n" \
"ON t2.`name` == t4.`name`"
self.assertEqual(to_str(expected), to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr.join(self.expr3, on='name')
expr = expr[(expr.id_x + expr.id_y < 10) & (expr.id_x > 3)]
expected = "SELECT * \n" \
"FROM (\n" \
" SELECT t2.`name`, t2.`id` AS `id_x`, t2.`fid` AS `fid_x`, t2.`isMale`, " \
"t2.`scale`, t2.`birth`, t2.`ds`, t3.`id` AS `id_y`, " \
"t3.`fid` AS `fid_y`, t3.`part1`, t3.`part2` \n" \
" FROM (\n" \
" SELECT * \n" \
" FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
" WHERE t1.`id` > 3\n" \
" ) t2 \n" \
" INNER JOIN \n" \
" mocked_project.`pyodps_test_expr_table2` t3\n" \
" ON t2.`name` == t3.`name` \n" \
") t4 \n" \
"WHERE (t4.`id_x` + t4.`id_y`) < 10"
self.assertEqual(to_str(expected), to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr.outer_join(self.expr3, on='name')
expr = expr[(expr.id_x + expr.id_y < 10) & (expr.id_x > 3)]
expected = "SELECT * \n" \
"FROM (\n" \
" SELECT t1.`name` AS `name_x`, t1.`id` AS `id_x`, t1.`fid` AS `fid_x`, " \
"t1.`isMale`, t1.`scale`, t1.`birth`, t1.`ds`, t2.`name` AS `name_y`, " \
"t2.`id` AS `id_y`, t2.`fid` AS `fid_y`, t2.`part1`, t2.`part2` \n" \
" FROM mocked_project.`pyodps_test_expr_table` t1 \n" \
" FULL OUTER JOIN \n" \
" mocked_project.`pyodps_test_expr_table2` t2\n" \
" ON t1.`name` == t2.`name` \n" \
") t3 \n" \
"WHERE ((t3.`id_x` + t3.`id_y`) < 10) AND (t3.`id_x` > 3)"
self.assertEqual(to_str(expected), to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr.join(self.expr3, on=['name', self.expr.id == self.expr3.id,
self.expr.id < 10, self.expr3.name == 'name1',
self.expr.id > 5])
expected = 'SELECT t2.`name`, t2.`id`, t2.`fid` AS `fid_x`, t2.`isMale`, ' \
't2.`scale`, t2.`birth`, t2.`ds`, t4.`fid` AS `fid_y`, t4.`part1`, t4.`part2` \n' \
'FROM (\n' \
' SELECT * \n' \
' FROM mocked_project.`pyodps_test_expr_table` t1 \n' \
' WHERE (t1.`id` < 10) AND (t1.`id` > 5)\n' \
') t2 \n' \
'INNER JOIN \n' \
' (\n' \
' SELECT * \n' \
' FROM mocked_project.`pyodps_test_expr_table2` t3 \n' \
' WHERE t3.`name` == \'name1\'\n' \
' ) t4\n' \
'ON (t2.`name` == t4.`name`) AND (t2.`id` == t4.`id`)'
self.assertEqual(to_str(expected), to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr = self.expr.left_join(self.expr3, on=['name', self.expr.id == self.expr3.id,
self.expr.id < 10, self.expr3.name == 'name1',
self.expr.id > 5])
expected = 'SELECT t1.`name` AS `name_x`, t1.`id` AS `id_x`, t1.`fid` AS `fid_x`, t1.`isMale`, ' \
't1.`scale`, t1.`birth`, t1.`ds`, t2.`name` AS `name_y`, t2.`id` AS `id_y`, ' \
't2.`fid` AS `fid_y`, t2.`part1`, t2.`part2` \n' \
'FROM mocked_project.`pyodps_test_expr_table` t1 \n' \
'LEFT OUTER JOIN \n' \
' mocked_project.`pyodps_test_expr_table2` t2\n' \
'ON ((((t1.`name` == t2.`name`) AND (t1.`id` == t2.`id`)) ' \
"AND (t1.`id` < 10)) AND (t2.`name` == 'name1')) AND (t1.`id` > 5)"
self.assertEqual(to_str(expected), to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
def testFilterPushdownThroughUnion(self):
expr = self.expr['name', 'id'].union(self.expr2['id', 'name'])
expr = expr.filter(expr.id + 1 < 3)
expected = 'SELECT * \n' \
'FROM (\n' \
' SELECT t1.`name`, t1.`id` \n' \
' FROM mocked_project.`pyodps_test_expr_table` t1 \n' \
' WHERE (t1.`id` + 1) < 3 \n' \
' UNION ALL\n' \
' SELECT t2.`name`, t2.`id` \n' \
' FROM mocked_project.`pyodps_test_expr_table2` t2 \n' \
' WHERE (t2.`id` + 1) < 3\n' \
') t3'
self.assertEqual(to_str(expected), to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
expr1 = self.expr.filter(self.expr.id == 1)['name', 'id']
expr2 = self.expr.filter(self.expr.id == 0)['id', 'name']
expr = expr1.union(expr2)
expected = 'SELECT * \n' \
'FROM (\n' \
' SELECT t1.`name`, t1.`id` \n' \
' FROM mocked_project.`pyodps_test_expr_table` t1 \n' \
' WHERE t1.`id` == 1 \n' \
' UNION ALL\n' \
' SELECT t2.`name`, t2.`id` \n' \
' FROM mocked_project.`pyodps_test_expr_table` t2 \n' \
' WHERE t2.`id` == 0\n' \
') t3'
self.assertEqual(to_str(expected), to_str(ODPSEngine(self.odps).compile(expr, prettify=False)))
