def create_column(self, col_name, col_type):
''' Takes the output from parse_col_desc and creates the right column type. This
method returns one of Column, ArrayColumn, MapColumn, StructColumn.'''
if isinstance(col_type, str):
if col_type.upper() == 'VARCHAR':
col_type = 'STRING'
type_name = self.TYPE_NAME_ALIASES.get(col_type.upper())
return Column(owner=None,
name=col_name.lower(),
exact_type=self.TYPES_BY_NAME[type_name])
general_class = col_type[0]
if general_class.upper() == 'ARRAY':
return ArrayColumn(owner=None,
name=col_name.lower(),
item=self.create_column(col_name='item',
col_type=col_type[1]))
if general_class.upper() == 'MAP':
return MapColumn(owner=None,
name=col_name.lower(),
key=self.create_column(col_name='key',
col_type=col_type[1]),
value=self.create_column(col_name='value',
col_type=col_type[2]))
if general_class.upper() == 'STRUCT':
struct_col = StructColumn(owner=None, name=col_name.lower())
for field_name, field_type in col_type[1:]:
struct_col.add_col(self.create_column(field_name, field_type))
return struct_col
general_class = self.TYPE_NAME_ALIASES.get(col_type[0].upper())
if general_class.upper() == 'DECIMAL':
return Column(owner=None,
name=col_name.lower(),
exact_type=get_decimal_class(int(col_type[1]),
int(col_type[2])))
if general_class.upper() == 'CHAR':
return Column(owner=None,
name=col_name.lower(),
exact_type=get_char_class(int(col_type[1])))
if general_class.upper() == 'VARCHAR':
type_size = int(col_type[1])
if type_size <= VarChar.MAX:
cur_type = get_varchar_class(type_size)
else:
cur_type = self.TYPES_BY_NAME['STRING']
return Column(owner=None,
name=col_name.lower(),
exact_type=cur_type)
raise Exception('unable to parse: {0}, type: {1}'.format(
col_name, col_type))
def test_table_model(self, cursor, hive_cursor):
table = Table("some_test_table")
cursor.drop_table(table.name, if_exists=True)
table.storage_format = 'textfile'
table.cols.append(Column(table, "bigint_col", BigInt))
table.cols.append(Column(table, "string_col", String))
cursor.create_table(table)
try:
other = hive_cursor.describe_table(table.name)
assert other.name == table.name
assert other.cols == table.cols
finally:
cursor.drop_table(table.name)
def create_join_predicate(self, parent_table, child_table):
for col in parent_table.cols:
if col.name == 'id':
parent_id_col = col
break
else:
parent_id_col = Column(parent_table, 'id', BigInt)
parent_id_col.for_flattening = True
parent_table.add_col(parent_id_col)
child_col_name = self.flat_collection_name(parent_table) + '_id'
child_col = Column(None, child_col_name, BigInt)
child_table.add_col(child_col)
return Equals.create_from_args(parent_id_col, child_col)
def create_join_predicate(self, parent_table, child_table):
for col in parent_table.cols:
if col.name == 'id':
parent_id_col = col
break
else:
parent_id_col = Column(parent_table, 'id', BigInt)
parent_id_col.for_flattening = True
parent_table.add_col(parent_id_col)
child_col_name = self.flat_collection_name(parent_table) + '_id'
child_col = Column(None, child_col_name, BigInt)
child_table.add_col(child_col)
return Equals.create_from_args(parent_id_col, child_col)
def _create_random_table(self, table_name, min_col_count, max_col_count,
allowed_storage_formats):
'''Create and return a Table with a random number of cols.'''
col_count = randint(min_col_count, max_col_count)
storage_format = choice(allowed_storage_formats)
table = Table(table_name)
table.storage_format = storage_format
allowed_types = list(TYPES)
# Avro doesn't support timestamps yet.
if table.storage_format == 'AVRO':
allowed_types.remove(Timestamp)
# TODO: 'table.cols' returns a copy of all scalar cols, so 'table.cols.append()'
# doesn't actually modify the table's columns. 'table.cols' should be changed
# to allow access to the real columns.
cols = table.cols
for col_idx in xrange(col_count):
col_type = choice(allowed_types)
col_type = choice(
filter(lambda type_: issubclass(type_, col_type), EXACT_TYPES))
if issubclass(col_type,
VarChar) and not issubclass(col_type, String):
col_type = get_varchar_class(randint(1, VarChar.MAX))
elif issubclass(col_type,
Char) and not issubclass(col_type, String):
col_type = get_char_class(randint(1, Char.MAX))
elif issubclass(col_type, Decimal):
max_digits = randint(1, Decimal.MAX_DIGITS)
col_type = get_decimal_class(max_digits,
randint(1, max_digits))
col = Column(
table, '%s_col_%s' % (col_type.__name__.lower(), col_idx + 1),
col_type)
cols.append(col)
table.cols = cols
return table
def describe_table(self, table_name):
'''Return a Table with table and col names always in lowercase.'''
rows = self.conn.execute_and_fetchall(
self.make_describe_table_sql(table_name))
table = Table(table_name.lower())
cols = table.cols # This is a copy
for row in rows:
col_name, data_type = row[:2]
if data_type == 'tinyint(1)':
# Just assume this is a boolean...
data_type = 'boolean'
if 'decimal' not in data_type and '(' in data_type:
# Strip the size of the data type
data_type = data_type[:data_type.index('(')]
cols.append(Column(table, col_name.lower(), self.parse_data_type(data_type)))
table.cols = cols
return table
def cols(self):
return ValExprList(
Column(self, item.name, item.type)
for item in self.query.select_clause.items)