def execute_sql_statement(
sql_statement, query, user_name, session,
cursor, return_results=False):
"""Executes a single SQL statement"""
database = query.database
db_engine_spec = database.db_engine_spec
parsed_query = ParsedQuery(sql_statement)
sql = parsed_query.stripped()
SQL_MAX_ROWS = app.config.get('SQL_MAX_ROW')
if not parsed_query.is_readonly() and not database.allow_dml:
raise SqlLabSecurityException(
_('Only `SELECT` statements are allowed against this database'))
if query.select_as_cta:
if not parsed_query.is_select():
raise SqlLabException(_(
'Only `SELECT` statements can be used with the CREATE TABLE '
'feature.'))
if not query.tmp_table_name:
start_dttm = datetime.fromtimestamp(query.start_time)
query.tmp_table_name = 'tmp_{}_table_{}'.format(
query.user_id, start_dttm.strftime('%Y_%m_%d_%H_%M_%S'))
sql = parsed_query.as_create_table(query.tmp_table_name)
query.select_as_cta_used = True
if parsed_query.is_select():
if SQL_MAX_ROWS and (not query.limit or query.limit > SQL_MAX_ROWS):
query.limit = SQL_MAX_ROWS
if query.limit:
sql = database.apply_limit_to_sql(sql, query.limit)
# Hook to allow environment-specific mutation (usually comments) to the SQL
SQL_QUERY_MUTATOR = config.get('SQL_QUERY_MUTATOR')
if SQL_QUERY_MUTATOR:
sql = SQL_QUERY_MUTATOR(sql, user_name, security_manager, database)
try:
if log_query:
log_query(
query.database.sqlalchemy_uri,
query.executed_sql,
query.schema,
user_name,
__name__,
security_manager,
)
query.executed_sql = sql
with stats_timing('sqllab.query.time_executing_query', stats_logger):
logging.info('Running query: \n{}'.format(sql))
db_engine_spec.execute(cursor, sql, async_=True)
logging.info('Handling cursor')
db_engine_spec.handle_cursor(cursor, query, session)
with stats_timing('sqllab.query.time_fetching_results', stats_logger):
logging.debug('Fetching data for query object: {}'.format(query.to_dict()))
data = db_engine_spec.fetch_data(cursor, query.limit)
except SoftTimeLimitExceeded as e:
logging.exception(e)
raise SqlLabTimeoutException(
"SQL Lab timeout. This environment's policy is to kill queries "
'after {} seconds.'.format(SQLLAB_TIMEOUT))
except Exception as e:
logging.exception(e)
raise SqlLabException(db_engine_spec.extract_error_message(e))
logging.debug('Fetching cursor description')
cursor_description = cursor.description
return dataframe.SupersetDataFrame(data, cursor_description, db_engine_spec)
def execute_sql_statements(
ctask, query_id, rendered_query, return_results=True, store_results=False,
user_name=None, session=None, start_time=None,
):
"""Executes the sql query returns the results."""
if store_results and start_time:
# only asynchronous queries
stats_logger.timing(
'sqllab.query.time_pending', now_as_float() - start_time)
query = get_query(query_id, session)
payload = dict(query_id=query_id)
database = query.database
db_engine_spec = database.db_engine_spec
db_engine_spec.patch()
if store_results and not results_backend:
raise SqlLabException("Results backend isn't configured.")
# Breaking down into multiple statements
parsed_query = ParsedQuery(rendered_query)
statements = parsed_query.get_statements()
logging.info(f'Executing {len(statements)} statement(s)')
logging.info("Set query to 'running'")
query.status = QueryStatus.RUNNING
query.start_running_time = now_as_float()
engine = database.get_sqla_engine(
schema=query.schema,
nullpool=True,
user_name=user_name,
)
# Sharing a single connection and cursor across the
# execution of all statements (if many)
with closing(engine.raw_connection()) as conn:
with closing(conn.cursor()) as cursor:
statement_count = len(statements)
for i, statement in enumerate(statements):
# TODO CHECK IF STOPPED
msg = f'Running statement {i+1} out of {statement_count}'
logging.info(msg)
query.set_extra_json_key('progress', msg)
session.commit()
is_last_statement = i == len(statements) - 1
try:
cdf = execute_sql_statement(
statement, query, user_name, session, cursor,
return_results=is_last_statement and return_results)
msg = f'Running statement {i+1} out of {statement_count}'
except Exception as e:
msg = str(e)
if statement_count > 1:
msg = f'[Statement {i+1} out of {statement_count}] ' + msg
payload = handle_query_error(msg, query, session, payload)
return payload
# Success, updating the query entry in database
query.rows = cdf.size
query.progress = 100
query.set_extra_json_key('progress', None)
query.status = QueryStatus.SUCCESS
if query.select_as_cta:
query.select_sql = database.select_star(
query.tmp_table_name,
limit=query.limit,
schema=database.force_ctas_schema,
show_cols=False,
latest_partition=False)
query.end_time = now_as_float()
session.commit()
payload.update({
'status': query.status,
'data': cdf.data if cdf.data else [],
'columns': cdf.columns if cdf.columns else [],
'query': query.to_dict(),
})
if store_results:
key = str(uuid.uuid4())
logging.info(f'Storing results in results backend, key: {key}')
with stats_timing('sqllab.query.results_backend_write', stats_logger):
json_payload = json.dumps(
payload, default=json_iso_dttm_ser, ignore_nan=True)
cache_timeout = database.cache_timeout
if cache_timeout is None:
cache_timeout = config.get('CACHE_DEFAULT_TIMEOUT', 0)
results_backend.set(key, zlib_compress(json_payload), cache_timeout)
query.results_key = key
session.commit()
if return_results:
return payload
def execute_sql_statement(sql_statement, query, user_name, session, cursor):
"""Executes a single SQL statement"""
query_id = query.id
database = query.database
db_engine_spec = database.db_engine_spec
parsed_query = ParsedQuery(sql_statement)
sql = parsed_query.stripped()
SQL_MAX_ROWS = app.config["SQL_MAX_ROW"]
if not parsed_query.is_readonly() and not database.allow_dml:
raise SqlLabSecurityException(
_("Only `SELECT` statements are allowed against this database")
)
if query.select_as_cta:
if not parsed_query.is_select():
raise SqlLabException(
_(
"Only `SELECT` statements can be used with the CREATE TABLE "
"feature."
)
)
if not query.tmp_table_name:
start_dttm = datetime.fromtimestamp(query.start_time)
query.tmp_table_name = "tmp_{}_table_{}".format(
query.user_id, start_dttm.strftime("%Y_%m_%d_%H_%M_%S")
)
sql = parsed_query.as_create_table(query.tmp_table_name)
query.select_as_cta_used = True
if parsed_query.is_select():
if SQL_MAX_ROWS and (not query.limit or query.limit > SQL_MAX_ROWS):
query.limit = SQL_MAX_ROWS
if query.limit:
sql = database.apply_limit_to_sql(sql, query.limit)
# Hook to allow environment-specific mutation (usually comments) to the SQL
SQL_QUERY_MUTATOR = config["SQL_QUERY_MUTATOR"]
if SQL_QUERY_MUTATOR:
sql = SQL_QUERY_MUTATOR(sql, user_name, security_manager, database)
try:
if log_query:
log_query(
query.database.sqlalchemy_uri,
query.executed_sql,
query.schema,
user_name,
__name__,
security_manager,
)
query.executed_sql = sql
session.commit()
with stats_timing("sqllab.query.time_executing_query", stats_logger):
logging.info(f"Query {query_id}: Running query: \n{sql}")
db_engine_spec.execute(cursor, sql, async_=True)
logging.info(f"Query {query_id}: Handling cursor")
db_engine_spec.handle_cursor(cursor, query, session)
with stats_timing("sqllab.query.time_fetching_results", stats_logger):
logging.debug(
"Query {}: Fetching data for query object: {}".format(
query_id, query.to_dict()
)
)
data = db_engine_spec.fetch_data(cursor, query.limit)
except SoftTimeLimitExceeded as e:
logging.exception(f"Query {query_id}: {e}")
raise SqlLabTimeoutException(
"SQL Lab timeout. This environment's policy is to kill queries "
"after {} seconds.".format(SQLLAB_TIMEOUT)
)
except Exception as e:
logging.exception(f"Query {query_id}: {e}")
raise SqlLabException(db_engine_spec.extract_error_message(e))
logging.debug(f"Query {query_id}: Fetching cursor description")
cursor_description = cursor.description
return SupersetDataFrame(data, cursor_description, db_engine_spec)
def execute_sql_statements(
ctask,
query_id,
rendered_query,
return_results=True,
store_results=False,
user_name=None,
session=None,
start_time=None,
expand_data=False,
):
"""Executes the sql query returns the results."""
if store_results and start_time:
# only asynchronous queries
stats_logger.timing("sqllab.query.time_pending", now_as_float() - start_time)
query = get_query(query_id, session)
payload = dict(query_id=query_id)
database = query.database
db_engine_spec = database.db_engine_spec
db_engine_spec.patch()
if database.allow_run_async and not results_backend:
raise SqlLabException("Results backend isn't configured.")
# Breaking down into multiple statements
parsed_query = ParsedQuery(rendered_query)
statements = parsed_query.get_statements()
logging.info(f"Query {query_id}: Executing {len(statements)} statement(s)")
logging.info(f"Query {query_id}: Set query to 'running'")
query.status = QueryStatus.RUNNING
query.start_running_time = now_as_float()
session.commit()
engine = database.get_sqla_engine(
schema=query.schema,
nullpool=True,
user_name=user_name,
source=sources.get("sql_lab", None),
)
# Sharing a single connection and cursor across the
# execution of all statements (if many)
with closing(engine.raw_connection()) as conn:
with closing(conn.cursor()) as cursor:
statement_count = len(statements)
for i, statement in enumerate(statements):
# Check if stopped
query = get_query(query_id, session)
if query.status == QueryStatus.STOPPED:
return
# Run statement
msg = f"Running statement {i+1} out of {statement_count}"
logging.info(f"Query {query_id}: {msg}")
query.set_extra_json_key("progress", msg)
session.commit()
try:
cdf = execute_sql_statement(
statement, query, user_name, session, cursor
)
except Exception as e:
msg = str(e)
if statement_count > 1:
msg = f"[Statement {i+1} out of {statement_count}] " + msg
payload = handle_query_error(msg, query, session, payload)
return payload
# Success, updating the query entry in database
query.rows = cdf.size
query.progress = 100
query.set_extra_json_key("progress", None)
if query.select_as_cta:
query.select_sql = database.select_star(
query.tmp_table_name,
limit=query.limit,
schema=database.force_ctas_schema,
show_cols=False,
latest_partition=False,
)
query.end_time = now_as_float()
data, selected_columns, all_columns, expanded_columns = _serialize_and_expand_data(
cdf, db_engine_spec, store_results and results_backend_use_msgpack, expand_data
)
payload.update(
{
"status": QueryStatus.SUCCESS,
"data": data,
"columns": all_columns,
"selected_columns": selected_columns,
"expanded_columns": expanded_columns,
"query": query.to_dict(),
}
)
payload["query"]["state"] = QueryStatus.SUCCESS
if store_results and results_backend:
key = str(uuid.uuid4())
logging.info(
f"Query {query_id}: Storing results in results backend, key: {key}"
)
with stats_timing("sqllab.query.results_backend_write", stats_logger):
with stats_timing(
"sqllab.query.results_backend_write_serialization", stats_logger
):
serialized_payload = _serialize_payload(
payload, results_backend_use_msgpack
)
cache_timeout = database.cache_timeout
if cache_timeout is None:
cache_timeout = config["CACHE_DEFAULT_TIMEOUT"]
compressed = zlib_compress(serialized_payload)
logging.debug(
f"*** serialized payload size: {getsizeof(serialized_payload)}"
)
logging.debug(f"*** compressed payload size: {getsizeof(compressed)}")
results_backend.set(key, compressed, cache_timeout)
query.results_key = key
query.status = QueryStatus.SUCCESS
session.commit()
if return_results:
return payload
def execute_sql_statements( # pylint: disable=too-many-arguments, too-many-locals, too-many-statements, too-many-branches
query_id: int,
rendered_query: str,
return_results: bool,
store_results: bool,
user_name: Optional[str],
session: Session,
start_time: Optional[float],
expand_data: bool,
log_params: Optional[Dict[str, Any]],
) -> Optional[Dict[str, Any]]:
"""Executes the sql query returns the results."""
if store_results and start_time:
# only asynchronous queries
stats_logger.timing("sqllab.query.time_pending",
now_as_float() - start_time)
query = get_query(query_id, session)
payload: Dict[str, Any] = dict(query_id=query_id)
database = query.database
db_engine_spec = database.db_engine_spec
db_engine_spec.patch()
if database.allow_run_async and not results_backend:
raise SqlLabException("Results backend isn't configured.")
# Breaking down into multiple statements
parsed_query = ParsedQuery(rendered_query, strip_comments=True)
if not db_engine_spec.run_multiple_statements_as_one:
statements = parsed_query.get_statements()
logger.info("Query %s: Executing %i statement(s)", str(query_id),
len(statements))
else:
statements = [rendered_query]
logger.info("Query %s: Executing query as a single statement",
str(query_id))
logger.info("Query %s: Set query to 'running'", str(query_id))
query.status = QueryStatus.RUNNING
query.start_running_time = now_as_float()
session.commit()
# Should we create a table or view from the select?
if (query.select_as_cta and query.ctas_method == CtasMethod.TABLE
and not parsed_query.is_valid_ctas()):
raise SqlLabException(
_("CTAS (create table as select) can only be run with a query where "
"the last statement is a SELECT. Please make sure your query has "
"a SELECT as its last statement. Then, try running your query again."
))
if (query.select_as_cta and query.ctas_method == CtasMethod.VIEW
and not parsed_query.is_valid_cvas()):
raise SqlLabException(
_("CVAS (create view as select) can only be run with a query with "
"a single SELECT statement. Please make sure your query has only "
"a SELECT statement. Then, try running your query again."))
engine = database.get_sqla_engine(
schema=query.schema,
nullpool=True,
user_name=user_name,
source=QuerySource.SQL_LAB,
)
# Sharing a single connection and cursor across the
# execution of all statements (if many)
with closing(engine.raw_connection()) as conn:
# closing the connection closes the cursor as well
cursor = conn.cursor()
statement_count = len(statements)
for i, statement in enumerate(statements):
# Check if stopped
query = get_query(query_id, session)
if query.status == QueryStatus.STOPPED:
return None
# For CTAS we create the table only on the last statement
apply_ctas = query.select_as_cta and (
query.ctas_method == CtasMethod.VIEW or
(query.ctas_method == CtasMethod.TABLE
and i == len(statements) - 1))
# Run statement
msg = f"Running statement {i+1} out of {statement_count}"
logger.info("Query %s: %s", str(query_id), msg)
query.set_extra_json_key("progress", msg)
session.commit()
try:
result_set = execute_sql_statement(
statement,
query,
user_name,
session,
cursor,
log_params,
apply_ctas,
)
except Exception as ex: # pylint: disable=broad-except
msg = str(ex)
if statement_count > 1:
msg = f"[Statement {i+1} out of {statement_count}] " + msg
payload = handle_query_error(msg, query, session, payload)
return payload
# Commit the connection so CTA queries will create the table.
conn.commit()
# Success, updating the query entry in database
query.rows = result_set.size
query.progress = 100
query.set_extra_json_key("progress", None)
if query.select_as_cta:
query.select_sql = database.select_star(
query.tmp_table_name,
schema=query.tmp_schema_name,
limit=query.limit,
show_cols=False,
latest_partition=False,
)
query.end_time = now_as_float()
use_arrow_data = store_results and cast(bool, results_backend_use_msgpack)
data, selected_columns, all_columns, expanded_columns = _serialize_and_expand_data(
result_set, db_engine_spec, use_arrow_data, expand_data)
# TODO: data should be saved separately from metadata (likely in Parquet)
payload.update({
"status": QueryStatus.SUCCESS,
"data": data,
"columns": all_columns,
"selected_columns": selected_columns,
"expanded_columns": expanded_columns,
"query": query.to_dict(),
})
payload["query"]["state"] = QueryStatus.SUCCESS
if store_results and results_backend:
key = str(uuid.uuid4())
logger.info("Query %s: Storing results in results backend, key: %s",
str(query_id), key)
with stats_timing("sqllab.query.results_backend_write", stats_logger):
with stats_timing(
"sqllab.query.results_backend_write_serialization",
stats_logger):
serialized_payload = _serialize_payload(
payload, cast(bool, results_backend_use_msgpack))
cache_timeout = database.cache_timeout
if cache_timeout is None:
cache_timeout = config["CACHE_DEFAULT_TIMEOUT"]
compressed = zlib_compress(serialized_payload)
logger.debug("*** serialized payload size: %i",
getsizeof(serialized_payload))
logger.debug("*** compressed payload size: %i",
getsizeof(compressed))
results_backend.set(key, compressed, cache_timeout)
query.results_key = key
query.status = QueryStatus.SUCCESS
session.commit()
if return_results:
# since we're returning results we need to create non-arrow data
if use_arrow_data:
(
data,
selected_columns,
all_columns,
expanded_columns,
) = _serialize_and_expand_data(result_set, db_engine_spec, False,
expand_data)
payload.update({
"data": data,
"columns": all_columns,
"selected_columns": selected_columns,
"expanded_columns": expanded_columns,
})
return payload
return None
def execute_sql_statement(
sql_statement: str,
query: Query,
user_name: Optional[str],
session: Session,
cursor: Any,
log_params: Optional[Dict[str, Any]],
apply_ctas: bool = False,
) -> SupersetResultSet:
"""Executes a single SQL statement"""
database = query.database
db_engine_spec = database.db_engine_spec
parsed_query = ParsedQuery(sql_statement)
sql = parsed_query.stripped()
# This is a test to see if the query is being
# limited by either the dropdown or the sql.
# We are testing to see if more rows exist than the limit.
increased_limit = None if query.limit is None else query.limit + 1
if not db_engine_spec.is_readonly_query(
parsed_query) and not database.allow_dml:
raise SqlLabSecurityException(
_("Only `SELECT` statements are allowed against this database"))
if apply_ctas:
if not query.tmp_table_name:
start_dttm = datetime.fromtimestamp(query.start_time)
query.tmp_table_name = "tmp_{}_table_{}".format(
query.user_id, start_dttm.strftime("%Y_%m_%d_%H_%M_%S"))
sql = parsed_query.as_create_table(
query.tmp_table_name,
schema_name=query.tmp_schema_name,
method=query.ctas_method,
)
query.select_as_cta_used = True
# Do not apply limit to the CTA queries when SQLLAB_CTAS_NO_LIMIT is set to true
if parsed_query.is_select() and not (query.select_as_cta_used
and SQLLAB_CTAS_NO_LIMIT):
if SQL_MAX_ROW and (not query.limit or query.limit > SQL_MAX_ROW):
query.limit = SQL_MAX_ROW
if query.limit:
# We are fetching one more than the requested limit in order
# to test whether there are more rows than the limit.
# Later, the extra row will be dropped before sending
# the results back to the user.
sql = database.apply_limit_to_sql(sql, increased_limit, force=True)
# Hook to allow environment-specific mutation (usually comments) to the SQL
sql = SQL_QUERY_MUTATOR(sql, user_name, security_manager, database)
try:
if log_query:
log_query(
query.database.sqlalchemy_uri,
query.executed_sql,
query.schema,
user_name,
__name__,
security_manager,
log_params,
)
query.executed_sql = sql
session.commit()
with stats_timing("sqllab.query.time_executing_query", stats_logger):
logger.debug("Query %d: Running query: %s", query.id, sql)
db_engine_spec.execute(cursor, sql, async_=True)
logger.debug("Query %d: Handling cursor", query.id)
db_engine_spec.handle_cursor(cursor, query, session)
with stats_timing("sqllab.query.time_fetching_results", stats_logger):
logger.debug(
"Query %d: Fetching data for query object: %s",
query.id,
str(query.to_dict()),
)
data = db_engine_spec.fetch_data(cursor, increased_limit)
if query.limit is None or len(data) <= query.limit:
query.limiting_factor = LimitingFactor.NOT_LIMITED
else:
# return 1 row less than increased_query
data = data[:-1]
except Exception as ex:
logger.error("Query %d: %s", query.id, type(ex), exc_info=True)
logger.debug("Query %d: %s", query.id, ex)
raise SqlLabException(db_engine_spec.extract_error_message(ex))
logger.debug("Query %d: Fetching cursor description", query.id)
cursor_description = cursor.description
return SupersetResultSet(data, cursor_description, db_engine_spec)
def execute_sql_statement( # pylint: disable=too-many-arguments,too-many-statements
sql_statement: str,
query: Query,
session: Session,
cursor: Any,
log_params: Optional[Dict[str, Any]],
apply_ctas: bool = False,
) -> SupersetResultSet:
"""Executes a single SQL statement"""
database: Database = query.database
db_engine_spec = database.db_engine_spec
parsed_query = ParsedQuery(sql_statement)
if is_feature_enabled("RLS_IN_SQLLAB"):
# Insert any applicable RLS predicates
parsed_query = ParsedQuery(
str(
insert_rls(
parsed_query._parsed[0], # pylint: disable=protected-access
database.id,
query.schema,
username=get_username(),
)))
sql = parsed_query.stripped()
# This is a test to see if the query is being
# limited by either the dropdown or the sql.
# We are testing to see if more rows exist than the limit.
increased_limit = None if query.limit is None else query.limit + 1
if not db_engine_spec.is_readonly_query(
parsed_query) and not database.allow_dml:
raise SupersetErrorException(
SupersetError(
message=__(
"Only SELECT statements are allowed against this database."
),
error_type=SupersetErrorType.DML_NOT_ALLOWED_ERROR,
level=ErrorLevel.ERROR,
))
if apply_ctas:
if not query.tmp_table_name:
start_dttm = datetime.fromtimestamp(query.start_time)
query.tmp_table_name = "tmp_{}_table_{}".format(
query.user_id, start_dttm.strftime("%Y_%m_%d_%H_%M_%S"))
sql = parsed_query.as_create_table(
query.tmp_table_name,
schema_name=query.tmp_schema_name,
method=query.ctas_method,
)
query.select_as_cta_used = True
# Do not apply limit to the CTA queries when SQLLAB_CTAS_NO_LIMIT is set to true
if db_engine_spec.is_select_query(parsed_query) and not (
query.select_as_cta_used and SQLLAB_CTAS_NO_LIMIT):
if SQL_MAX_ROW and (not query.limit or query.limit > SQL_MAX_ROW):
query.limit = SQL_MAX_ROW
sql = apply_limit_if_exists(database, increased_limit, query, sql)
# Hook to allow environment-specific mutation (usually comments) to the SQL
sql = SQL_QUERY_MUTATOR(
sql,
user_name=get_username(), # TODO(john-bodley): Deprecate in 3.0.
security_manager=security_manager,
database=database,
)
try:
query.executed_sql = sql
if log_query:
log_query(
query.database.sqlalchemy_uri,
query.executed_sql,
query.schema,
get_username(),
__name__,
security_manager,
log_params,
)
session.commit()
with stats_timing("sqllab.query.time_executing_query", stats_logger):
logger.debug("Query %d: Running query: %s", query.id, sql)
db_engine_spec.execute(cursor, sql, async_=True)
logger.debug("Query %d: Handling cursor", query.id)
db_engine_spec.handle_cursor(cursor, query, session)
with stats_timing("sqllab.query.time_fetching_results", stats_logger):
logger.debug(
"Query %d: Fetching data for query object: %s",
query.id,
str(query.to_dict()),
)
data = db_engine_spec.fetch_data(cursor, increased_limit)
if query.limit is None or len(data) <= query.limit:
query.limiting_factor = LimitingFactor.NOT_LIMITED
else:
# return 1 row less than increased_query
data = data[:-1]
except SoftTimeLimitExceeded as ex:
logger.warning("Query %d: Time limit exceeded", query.id)
logger.debug("Query %d: %s", query.id, ex)
raise SupersetErrorException(
SupersetError(
message=__(
"The query was killed after %(sqllab_timeout)s seconds. It might "
"be too complex, or the database might be under heavy load.",
sqllab_timeout=SQLLAB_TIMEOUT,
),
error_type=SupersetErrorType.SQLLAB_TIMEOUT_ERROR,
level=ErrorLevel.ERROR,
)) from ex
except Exception as ex:
# query is stopped in another thread/worker
# stopping raises expected exceptions which we should skip
session.refresh(query)
if query.status == QueryStatus.STOPPED:
raise SqlLabQueryStoppedException() from ex
logger.error("Query %d: %s", query.id, type(ex), exc_info=True)
logger.debug("Query %d: %s", query.id, ex)
raise SqlLabException(db_engine_spec.extract_error_message(ex)) from ex
logger.debug("Query %d: Fetching cursor description", query.id)
cursor_description = cursor.description
return SupersetResultSet(data, cursor_description, db_engine_spec)
def execute_sql_statement(
sql_statement: str,
query: Query,
user_name: Optional[str],
session: Session,
cursor: Any,
log_params: Optional[Dict[str, Any]],
) -> SupersetResultSet:
"""Executes a single SQL statement"""
database = query.database
db_engine_spec = database.db_engine_spec
parsed_query = ParsedQuery(sql_statement)
sql = parsed_query.stripped()
if not db_engine_spec.is_readonly_query(
parsed_query) and not database.allow_dml:
raise SqlLabSecurityException(
_("Only `SELECT` statements are allowed against this database"))
if query.select_as_cta:
if not parsed_query.is_select():
raise SqlLabException(
_("Only `SELECT` statements can be used with the CREATE TABLE "
"feature."))
if not query.tmp_table_name:
start_dttm = datetime.fromtimestamp(query.start_time)
query.tmp_table_name = "tmp_{}_table_{}".format(
query.user_id, start_dttm.strftime("%Y_%m_%d_%H_%M_%S"))
sql = parsed_query.as_create_table(
query.tmp_table_name,
schema_name=query.tmp_schema_name,
method=query.ctas_method,
)
query.select_as_cta_used = True
# Do not apply limit to the CTA queries when SQLLAB_CTAS_NO_LIMIT is set to true
if parsed_query.is_select() and not (query.select_as_cta_used
and SQLLAB_CTAS_NO_LIMIT):
if SQL_MAX_ROW and (not query.limit or query.limit > SQL_MAX_ROW):
query.limit = SQL_MAX_ROW
if query.limit:
sql = database.apply_limit_to_sql(sql, query.limit)
# Hook to allow environment-specific mutation (usually comments) to the SQL
if SQL_QUERY_MUTATOR:
sql = SQL_QUERY_MUTATOR(sql, user_name, security_manager, database)
try:
if log_query:
log_query(
query.database.sqlalchemy_uri,
query.executed_sql,
query.schema,
user_name,
__name__,
security_manager,
log_params,
)
query.executed_sql = sql
session.commit()
with stats_timing("sqllab.query.time_executing_query", stats_logger):
logger.debug("Query %d: Running query: %s", query.id, sql)
db_engine_spec.execute(cursor, sql, async_=True)
logger.debug("Query %d: Handling cursor", query.id)
db_engine_spec.handle_cursor(cursor, query, session)
with stats_timing("sqllab.query.time_fetching_results", stats_logger):
logger.debug(
"Query %d: Fetching data for query object: %s",
query.id,
str(query.to_dict()),
)
data = db_engine_spec.fetch_data(cursor, query.limit)
except SoftTimeLimitExceeded as ex:
logger.error("Query %d: Time limit exceeded", query.id)
logger.debug("Query %d: %s", query.id, ex)
raise SqlLabTimeoutException(
"SQL Lab timeout. This environment's policy is to kill queries "
"after {} seconds.".format(SQLLAB_TIMEOUT))
except Exception as ex:
logger.error("Query %d: %s", query.id, type(ex))
logger.debug("Query %d: %s", query.id, ex)
raise SqlLabException(db_engine_spec.extract_error_message(ex))
logger.debug("Query %d: Fetching cursor description", query.id)
cursor_description = cursor.description
return SupersetResultSet(data, cursor_description, db_engine_spec)
def execute_sql_statements(
ctask,
query_id,
rendered_query,
return_results=True,
store_results=False,
user_name=None,
session=None,
start_time=None,
):
"""Executes the sql query returns the results."""
if store_results and start_time:
# only asynchronous queries
stats_logger.timing('sqllab.query.time_pending',
now_as_float() - start_time)
query = get_query(query_id, session)
payload = dict(query_id=query_id)
database = query.database
db_engine_spec = database.db_engine_spec
db_engine_spec.patch()
if store_results and not results_backend:
raise SqlLabException("Results backend isn't configured.")
# Breaking down into multiple statements
parsed_query = ParsedQuery(rendered_query)
statements = parsed_query.get_statements()
logging.info(f'Executing {len(statements)} statement(s)')
logging.info("Set query to 'running'")
query.status = QueryStatus.RUNNING
query.start_running_time = now_as_float()
engine = database.get_sqla_engine(
schema=query.schema,
nullpool=True,
user_name=user_name,
source=sources.get('sql_lab', None),
)
# Sharing a single connection and cursor across the
# execution of all statements (if many)
with closing(engine.raw_connection()) as conn:
with closing(conn.cursor()) as cursor:
statement_count = len(statements)
for i, statement in enumerate(statements):
# TODO CHECK IF STOPPED
msg = f'Running statement {i+1} out of {statement_count}'
logging.info(msg)
query.set_extra_json_key('progress', msg)
session.commit()
try:
cdf = execute_sql_statement(statement, query, user_name,
session, cursor)
msg = f'Running statement {i+1} out of {statement_count}'
except Exception as e:
msg = str(e)
if statement_count > 1:
msg = f'[Statement {i+1} out of {statement_count}] ' + msg
payload = handle_query_error(msg, query, session, payload)
return payload
# Success, updating the query entry in database
query.rows = cdf.size
query.progress = 100
query.set_extra_json_key('progress', None)
query.status = QueryStatus.SUCCESS
if query.select_as_cta:
query.select_sql = database.select_star(
query.tmp_table_name,
limit=query.limit,
schema=database.force_ctas_schema,
show_cols=False,
latest_partition=False)
query.end_time = now_as_float()
selected_columns = cdf.columns or []
data = cdf.data or []
all_columns, data, expanded_columns = db_engine_spec.expand_data(
selected_columns, data)
payload.update({
'status': query.status,
'data': data,
'columns': all_columns,
'selected_columns': selected_columns,
'expanded_columns': expanded_columns,
'query': query.to_dict(),
})
if store_results:
key = str(uuid.uuid4())
logging.info(f'Storing results in results backend, key: {key}')
with stats_timing('sqllab.query.results_backend_write', stats_logger):
json_payload = json.dumps(payload,
default=json_iso_dttm_ser,
ignore_nan=True)
cache_timeout = database.cache_timeout
if cache_timeout is None:
cache_timeout = config.get('CACHE_DEFAULT_TIMEOUT', 0)
results_backend.set(key, zlib_compress(json_payload),
cache_timeout)
query.results_key = key
session.commit()
if return_results:
return payload
