dbt-selly/dbt-env/lib/python3.8/site-packages/dbt/adapters/sql/impl.py

251 lines
8.2 KiB
Python

import agate
from typing import Any, Optional, Tuple, Type, List
import dbt.clients.agate_helper
from dbt.contracts.connection import Connection
import dbt.exceptions
from dbt.adapters.base import BaseAdapter, available
from dbt.adapters.sql import SQLConnectionManager
from dbt.logger import GLOBAL_LOGGER as logger
from dbt.adapters.base.relation import BaseRelation
LIST_RELATIONS_MACRO_NAME = 'list_relations_without_caching'
GET_COLUMNS_IN_RELATION_MACRO_NAME = 'get_columns_in_relation'
LIST_SCHEMAS_MACRO_NAME = 'list_schemas'
CHECK_SCHEMA_EXISTS_MACRO_NAME = 'check_schema_exists'
CREATE_SCHEMA_MACRO_NAME = 'create_schema'
DROP_SCHEMA_MACRO_NAME = 'drop_schema'
RENAME_RELATION_MACRO_NAME = 'rename_relation'
TRUNCATE_RELATION_MACRO_NAME = 'truncate_relation'
DROP_RELATION_MACRO_NAME = 'drop_relation'
ALTER_COLUMN_TYPE_MACRO_NAME = 'alter_column_type'
class SQLAdapter(BaseAdapter):
"""The default adapter with the common agate conversions and some SQL
methods implemented. This adapter has a different much shorter list of
methods to implement, but some more macros that must be implemented.
To implement a macro, implement "${adapter_type}__${macro_name}". in the
adapter's internal project.
Methods to implement:
- date_function
Macros to implement:
- get_catalog
- list_relations_without_caching
- get_columns_in_relation
"""
ConnectionManager: Type[SQLConnectionManager]
connections: SQLConnectionManager
@available.parse(lambda *a, **k: (None, None))
def add_query(
self,
sql: str,
auto_begin: bool = True,
bindings: Optional[Any] = None,
abridge_sql_log: bool = False,
) -> Tuple[Connection, Any]:
"""Add a query to the current transaction. A thin wrapper around
ConnectionManager.add_query.
:param sql: The SQL query to add
:param auto_begin: If set and there is no transaction in progress,
begin a new one.
:param bindings: An optional list of bindings for the query.
:param abridge_sql_log: If set, limit the raw sql logged to 512
characters
"""
return self.connections.add_query(sql, auto_begin, bindings,
abridge_sql_log)
@classmethod
def convert_text_type(cls, agate_table: agate.Table, col_idx: int) -> str:
return "text"
@classmethod
def convert_number_type(
cls, agate_table: agate.Table, col_idx: int
) -> str:
decimals = agate_table.aggregate(agate.MaxPrecision(col_idx))
return "float8" if decimals else "integer"
@classmethod
def convert_boolean_type(
cls, agate_table: agate.Table, col_idx: int
) -> str:
return "boolean"
@classmethod
def convert_datetime_type(
cls, agate_table: agate.Table, col_idx: int
) -> str:
return "timestamp without time zone"
@classmethod
def convert_date_type(cls, agate_table: agate.Table, col_idx: int) -> str:
return "date"
@classmethod
def convert_time_type(cls, agate_table: agate.Table, col_idx: int) -> str:
return "time"
@classmethod
def is_cancelable(cls) -> bool:
return True
def expand_column_types(self, goal, current):
reference_columns = {
c.name: c for c in
self.get_columns_in_relation(goal)
}
target_columns = {
c.name: c for c
in self.get_columns_in_relation(current)
}
for column_name, reference_column in reference_columns.items():
target_column = target_columns.get(column_name)
if target_column is not None and \
target_column.can_expand_to(reference_column):
col_string_size = reference_column.string_size()
new_type = self.Column.string_type(col_string_size)
logger.debug("Changing col type from {} to {} in table {}",
target_column.data_type, new_type, current)
self.alter_column_type(current, column_name, new_type)
def alter_column_type(
self, relation, column_name, new_column_type
) -> None:
"""
1. Create a new column (w/ temp name and correct type)
2. Copy data over to it
3. Drop the existing column (cascade!)
4. Rename the new column to existing column
"""
kwargs = {
'relation': relation,
'column_name': column_name,
'new_column_type': new_column_type,
}
self.execute_macro(
ALTER_COLUMN_TYPE_MACRO_NAME,
kwargs=kwargs
)
def drop_relation(self, relation):
if relation.type is None:
dbt.exceptions.raise_compiler_error(
'Tried to drop relation {}, but its type is null.'
.format(relation))
self.cache_dropped(relation)
self.execute_macro(
DROP_RELATION_MACRO_NAME,
kwargs={'relation': relation}
)
def truncate_relation(self, relation):
self.execute_macro(
TRUNCATE_RELATION_MACRO_NAME,
kwargs={'relation': relation}
)
def rename_relation(self, from_relation, to_relation):
self.cache_renamed(from_relation, to_relation)
kwargs = {'from_relation': from_relation, 'to_relation': to_relation}
self.execute_macro(
RENAME_RELATION_MACRO_NAME,
kwargs=kwargs
)
def get_columns_in_relation(self, relation):
return self.execute_macro(
GET_COLUMNS_IN_RELATION_MACRO_NAME,
kwargs={'relation': relation}
)
def create_schema(self, relation: BaseRelation) -> None:
relation = relation.without_identifier()
logger.debug('Creating schema "{}"', relation)
kwargs = {
'relation': relation,
}
self.execute_macro(CREATE_SCHEMA_MACRO_NAME, kwargs=kwargs)
self.commit_if_has_connection()
# we can't update the cache here, as if the schema already existed we
# don't want to (incorrectly) say that it's empty
def drop_schema(self, relation: BaseRelation) -> None:
relation = relation.without_identifier()
logger.debug('Dropping schema "{}".', relation)
kwargs = {
'relation': relation,
}
self.execute_macro(DROP_SCHEMA_MACRO_NAME, kwargs=kwargs)
# we can update the cache here
self.cache.drop_schema(relation.database, relation.schema)
def list_relations_without_caching(
self, schema_relation: BaseRelation,
) -> List[BaseRelation]:
kwargs = {'schema_relation': schema_relation}
results = self.execute_macro(
LIST_RELATIONS_MACRO_NAME,
kwargs=kwargs
)
relations = []
quote_policy = {
'database': True,
'schema': True,
'identifier': True
}
for _database, name, _schema, _type in results:
try:
_type = self.Relation.get_relation_type(_type)
except ValueError:
_type = self.Relation.External
relations.append(self.Relation.create(
database=_database,
schema=_schema,
identifier=name,
quote_policy=quote_policy,
type=_type
))
return relations
def quote(self, identifier):
return '"{}"'.format(identifier)
def list_schemas(self, database: str) -> List[str]:
results = self.execute_macro(
LIST_SCHEMAS_MACRO_NAME,
kwargs={'database': database}
)
return [row[0] for row in results]
def check_schema_exists(self, database: str, schema: str) -> bool:
information_schema = self.Relation.create(
database=database,
schema=schema,
identifier='INFORMATION_SCHEMA',
quote_policy=self.config.quoting
).information_schema()
kwargs = {'information_schema': information_schema, 'schema': schema}
results = self.execute_macro(
CHECK_SCHEMA_EXISTS_MACRO_NAME,
kwargs=kwargs
)
return results[0][0] > 0