def col_split(dataset, request: Request,
column_split_spec: ColumnSplitSpec, *args, **kwargs):
"""
Split query in 2 steps if a large number of columns is being selected.
- First query only selects event_id and project_id.
- Second query selects all fields for only those events.
- Shrink the date range.
"""
# The query function may mutate the request body during query
# evaluation, so we need to copy the body to ensure that the query has
# not been modified by the time we're ready to run the full query.
minimal_request = copy.deepcopy(request)
minimal_request.query.set_selected_columns(
column_split_spec.get_min_columns())
result = query_func(dataset, minimal_request, *args, **kwargs)
del minimal_request
if result.result["data"]:
request = copy.deepcopy(request)
event_ids = list(
set([
event[column_split_spec.id_column]
for event in result.result["data"]
]))
request.query.add_conditions([(column_split_spec.id_column, "IN",
event_ids)])
request.query.set_offset(0)
request.query.set_limit(len(event_ids))
project_ids = list(
set([
event[column_split_spec.project_column]
for event in result.result["data"]
]))
request.extensions["project"]["project"] = project_ids
timestamp_field = column_split_spec.timestamp_column
timestamps = [
event[timestamp_field] for event in result.result["data"]
]
request.extensions[
"timeseries"]["from_date"] = util.parse_datetime(
min(timestamps)).isoformat()
# We add 1 second since this gets translated to ('timestamp', '<', to_date)
# and events are stored with a granularity of 1 second.
request.extensions["timeseries"]["to_date"] = (
util.parse_datetime(max(timestamps)) +
timedelta(seconds=1)).isoformat()
return query_func(dataset, request, *args, **kwargs)
def get_time_limit(
cls, timeseries_extension: Mapping[str, Any]
) -> Tuple[datetime, datetime]:
max_days, date_align = state.get_configs(
[("max_days", None), ("date_align_seconds", 1)]
)
to_date = parse_datetime(timeseries_extension["to_date"], date_align)
from_date = parse_datetime(timeseries_extension["from_date"], date_align)
assert from_date <= to_date
if max_days is not None and (to_date - from_date).days > max_days:
from_date = to_date - timedelta(days=max_days)
return (from_date, to_date)
def process_condition(self, condition) -> Tuple[str, str, Any]:
lhs, op, lit = condition
if (lhs in self.__time_parse_columns
and op in (">", "<", ">=", "<=", "=", "!=")
and isinstance(lit, str)):
lit = parse_datetime(lit)
return lhs, op, lit
def process_condition(self, condition) -> Tuple[str, str, any]:
lhs, op, lit = condition
if (lhs in self.__time_parse_columns
and op in ('>', '<', '>=', '<=', '=', '!=')
and isinstance(lit, str)):
lit = parse_datetime(lit)
return lhs, op, lit
def parse(exp: Expression) -> Expression:
result = DATETIME_MATCH.match(exp)
if result is not None:
date_string = result.expression("date_string")
assert isinstance(date_string, Literal) # mypy
assert isinstance(date_string.value, str) # mypy
return Literal(exp.alias, parse_datetime(date_string.value))
return exp
def col_split(dataset, request: Request, *args, **kwargs):
"""
Split query in 2 steps if a large number of columns is being selected.
- First query only selects event_id and project_id.
- Second query selects all fields for only those events.
- Shrink the date range.
"""
# The query function may mutate the request body during query
# evaluation, so we need to copy the body to ensure that the query has
# not been modified by the time we're ready to run the full query.
minimal_request = copy.deepcopy(request)
minimal_request.query.set_selected_columns(MIN_COLS)
result, status = query_func(dataset, minimal_request, *args, **kwargs)
del minimal_request
# If something failed, just return
if status != 200:
return result, status
if result['data']:
request = copy.deepcopy(request)
event_ids = list(
set([event['event_id'] for event in result['data']]))
request.query.add_conditions([('event_id', 'IN', event_ids)])
request.query.set_offset(0)
request.query.set_limit(len(event_ids))
project_ids = list(
set([event['project_id'] for event in result['data']]))
request.extensions['project']['project'] = project_ids
timestamps = [event['timestamp'] for event in result['data']]
request.extensions[
'timeseries']['from_date'] = util.parse_datetime(
min(timestamps)).isoformat()
# We add 1 second since this gets translated to ('timestamp', '<', to_date)
# and events are stored with a granularity of 1 second.
request.extensions['timeseries']['to_date'] = (
util.parse_datetime(max(timestamps)) +
timedelta(seconds=1)).isoformat()
return query_func(dataset, request, *args, **kwargs)
def __process_condition(self, exp: Expression) -> Expression:
result = self.condition_match.match(exp)
if result is not None:
literal = result.expression("literal")
assert isinstance(exp, FunctionCall) # mypy
assert isinstance(literal, Literal) # mypy
try:
value = parse_datetime(str(literal.value))
except ValueError as err:
column_name = result.string("column_name")
raise InvalidQueryException(
f"Illegal datetime in condition on column {column_name}: '{literal.value}''"
) from err
return FunctionCall(
exp.alias,
exp.function_name,
(exp.parameters[0], Literal(literal.alias, value)),
)
return exp
Column("my_time", None, "time"),
Literal(None, "2020-01-01"),
),
FunctionCall(
"my_time",
"toStartOfHour",
(Column(None, None, "finish_ts"), Literal(None, "Universal")),
),
binary_condition(
ConditionFunctions.EQ,
FunctionCall(
"my_time",
"toStartOfHour",
(Column(None, None, "finish_ts"), Literal(None, "Universal")),
),
Literal(None, parse_datetime("2020-01-01")),
),
"(toStartOfHour(finish_ts, 'Universal') AS my_time)",
"equals((toStartOfHour(finish_ts, 'Universal') AS my_time), toDateTime('2020-01-01T00:00:00', 'Universal'))",
id="granularity-3600-simple-condition",
),
pytest.param(
60,
binary_condition(
BooleanFunctions.AND,
binary_condition(
ConditionFunctions.EQ,
Column("my_time", None, "time"),
Literal(None, "2020-01-01"),
),
binary_condition(
def process_query(self, query: Query,
request_settings: RequestSettings) -> None:
conditions = query.get_conditions()
if not conditions:
return
# Enable the processor only if we have enough data in the flattened
# columns. Which have been deployed at BEGINNING_OF_TIME. If the query
# starts earlier than that we do not apply the optimization.
if self.__beginning_of_time:
apply_optimization = False
for condition in conditions:
if (is_condition(condition) and isinstance(condition[0], str)
and condition[0] in self.__timestamp_cols
and condition[1] in (">=", ">")
and isinstance(condition[2], str)):
try:
start_ts = parse_datetime(condition[2])
if (start_ts -
self.__beginning_of_time).total_seconds() > 0:
apply_optimization = True
except Exception:
# We should not get here, it means the from timestamp is malformed
# Returning here is just for safety
logger.error(
"Cannot parse start date for NestedFieldOptimizer: %r",
condition,
)
return
if not apply_optimization:
return
# Do not use flattened tags if tags are being unpacked anyway. In that case
# using flattened tags only implies loading an additional column thus making
# the query heavier and slower
if self.__has_tags(query.get_arrayjoin_from_ast()):
return
if query.get_groupby_from_ast():
for expression in query.get_groupby_from_ast():
if self.__has_tags(expression):
return
if self.__has_tags(query.get_having_from_ast()):
return
if query.get_orderby_from_ast():
for orderby in query.get_orderby_from_ast():
if self.__has_tags(orderby.expression):
return
new_conditions = []
positive_like_expression: List[str] = []
negative_like_expression: List[str] = []
for c in conditions:
keyvalue = self.__is_optimizable(c, self.__nested_col)
if not keyvalue:
new_conditions.append(c)
else:
expression = f"{escape_field(keyvalue.nested_col_key)}={escape_field(keyvalue.value)}"
if keyvalue.operand == Operand.EQ:
positive_like_expression.append(expression)
else:
negative_like_expression.append(expression)
if positive_like_expression:
# Positive conditions "=" are all merged together in one LIKE expression
positive_like_expression = sorted(positive_like_expression)
like_formatted = f"%|{'|%|'.join(positive_like_expression)}|%"
new_conditions.append(
[self.__flattened_col, "LIKE", like_formatted])
for expression in negative_like_expression:
# Negative conditions "!=" cannot be merged together. We can still transform
# them into NOT LIKE statements, but each condition has to be one
# statement.
not_like_formatted = f"%|{expression}|%"
new_conditions.append(
[self.__flattened_col, "NOT LIKE", not_like_formatted])
query.set_conditions(new_conditions)
Column("my_time", None, "time"),
Literal(None, "2020-01-01"),
),
FunctionCall(
"my_time",
"toStartOfHour",
(Column(None, None, "finish_ts"), Literal(None, "Universal")),
),
binary_condition(
ConditionFunctions.EQ,
FunctionCall(
"my_time",
"toStartOfHour",
(Column(None, None, "finish_ts"), Literal(None, "Universal")),
),
Literal(None, parse_datetime("2020-01-01")),
),
"(toStartOfHour(finish_ts, 'Universal') AS my_time)",
"equals((toStartOfHour(finish_ts, 'Universal') AS my_time), toDateTime('2020-01-01T00:00:00', 'Universal'))",
id="granularity-3600-simple-condition",
),
pytest.param(
60,
binary_condition(
BooleanFunctions.AND,
binary_condition(
ConditionFunctions.EQ,
Column("my_time", None, "time"),
Literal(None, "2020-01-01"),
),
binary_condition(
def time_split(dataset, request: Request, *args, **kwargs):
"""
If a query is:
- ORDER BY timestamp DESC
- has no grouping
- has an offset/limit
- has a large time range
We know we have to reverse-sort the entire set of rows to return the small
chunk at the end of the time range, so optimistically split the time range
into smaller increments, and start with the last one, so that we can potentially
avoid querying the entire range.
"""
date_align, split_step = state.get_configs([("date_align_seconds", 1),
("split_step", 3600)
] # default 1 hour
)
query_limit = request.query.get_limit()
limit = query_limit if query_limit is not None else 0
remaining_offset = request.query.get_offset()
to_date = util.parse_datetime(
request.extensions["timeseries"]["to_date"], date_align)
from_date = util.parse_datetime(
request.extensions["timeseries"]["from_date"], date_align)
overall_result = None
split_end = to_date
split_start = max(split_end - timedelta(seconds=split_step), from_date)
total_results = 0
while split_start < split_end and total_results < limit:
request.extensions["timeseries"][
"from_date"] = split_start.isoformat()
request.extensions["timeseries"]["to_date"] = split_end.isoformat()
# Because its paged, we have to ask for (limit+offset) results
# and set offset=0 so we can then trim them ourselves.
request.query.set_offset(0)
request.query.set_limit(limit - total_results + remaining_offset)
# The query function may mutate the request body during query
# evaluation, so we need to copy the body to ensure that the query
# has not been modified in between this call and the next loop
# iteration, if needed.
# XXX: The extra data is carried across from the initial response
# and never updated.
result = query_func(dataset, copy.deepcopy(request), *args,
**kwargs)
if overall_result is None:
overall_result = result
else:
overall_result.result["data"].extend(result.result["data"])
if remaining_offset > 0 and len(overall_result.result["data"]) > 0:
to_trim = min(remaining_offset,
len(overall_result.result["data"]))
overall_result.result["data"] = overall_result.result["data"][
to_trim:]
remaining_offset -= to_trim
total_results = len(overall_result.result["data"])
if total_results < limit:
if len(result.result["data"]) == 0:
# If we got nothing from the last query, expand the range by a static factor
split_step = split_step * STEP_GROWTH
else:
# If we got some results but not all of them, estimate how big the time
# range should be for the next query based on how many results we got for
# our last query and its time range, and how many we have left to fetch.
remaining = limit - total_results
split_step = split_step * math.ceil(
remaining / float(len(result.result["data"])))
# Set the start and end of the next query based on the new range.
split_end = split_start
try:
split_start = max(
split_end - timedelta(seconds=split_step), from_date)
except OverflowError:
split_start = from_date
return overall_result
def execute(
self,
query: Query,
request_settings: RequestSettings,
runner: SplitQueryRunner,
) -> Optional[QueryResult]:
"""
If a query is:
- ORDER BY timestamp DESC
- has no grouping
- has an offset/limit
- has a large time range
We know we have to reverse-sort the entire set of rows to return the small
chunk at the end of the time range, so optimistically split the time range
into smaller increments, and start with the last one, so that we can potentially
avoid querying the entire range.
"""
limit = query.get_limit()
if limit is None or query.get_groupby():
return None
if query.get_offset() >= 1000:
return None
orderby = query.get_orderby()
if not orderby or orderby[0] != f"-{self.__timestamp_col}":
return None
conditions = query.get_conditions() or []
from_date_str = next(
(condition[2] for condition in conditions
if _identify_condition(condition, self.__timestamp_col, ">=")),
None,
)
to_date_str = next(
(condition[2] for condition in conditions
if _identify_condition(condition, self.__timestamp_col, "<")),
None,
)
from_date_ast, to_date_ast = get_time_range(query,
self.__timestamp_col)
if not from_date_str or not to_date_str:
return None
date_align, split_step = state.get_configs([("date_align_seconds", 1),
("split_step", 3600)
] # default 1 hour
)
to_date = util.parse_datetime(to_date_str, date_align)
from_date = util.parse_datetime(from_date_str, date_align)
if from_date != from_date_ast:
logger.warning(
"Mismatch in start date on time splitter.",
extra={
"ast": str(from_date_ast),
"legacy": str(from_date)
},
exc_info=True,
)
metrics.increment("mismatch.ast_from_date")
remaining_offset = query.get_offset()
overall_result = None
split_end = to_date
split_start = max(split_end - timedelta(seconds=split_step), from_date)
total_results = 0
while split_start < split_end and total_results < limit:
# We need to make a copy to use during the query execution because we replace
# the start-end conditions on the query at each iteration of this loop.
split_query = copy.deepcopy(query)
_replace_condition(split_query, self.__timestamp_col, ">=",
split_start.isoformat())
_replace_ast_condition(split_query, self.__timestamp_col, ">=",
LiteralExpr(None, split_start))
_replace_condition(split_query, self.__timestamp_col, "<",
split_end.isoformat())
_replace_ast_condition(split_query, self.__timestamp_col, "<",
LiteralExpr(None, split_end))
# Because its paged, we have to ask for (limit+offset) results
# and set offset=0 so we can then trim them ourselves.
split_query.set_offset(0)
split_query.set_limit(limit - total_results + remaining_offset)
# At every iteration we only append the "data" key from the results returned by
# the runner. The "extra" key is only populated at the first iteration of the
# loop and never changed.
result = runner(split_query, request_settings)
if overall_result is None:
overall_result = result
else:
overall_result.result["data"].extend(result.result["data"])
if remaining_offset > 0 and len(overall_result.result["data"]) > 0:
to_trim = min(remaining_offset,
len(overall_result.result["data"]))
overall_result.result["data"] = overall_result.result["data"][
to_trim:]
remaining_offset -= to_trim
total_results = len(overall_result.result["data"])
if total_results < limit:
if len(result.result["data"]) == 0:
# If we got nothing from the last query, expand the range by a static factor
split_step = split_step * STEP_GROWTH
else:
# If we got some results but not all of them, estimate how big the time
# range should be for the next query based on how many results we got for
# our last query and its time range, and how many we have left to fetch.
remaining = limit - total_results
split_step = split_step * math.ceil(
remaining / float(len(result.result["data"])))
# Set the start and end of the next query based on the new range.
split_end = split_start
try:
split_start = max(
split_end - timedelta(seconds=split_step), from_date)
except OverflowError:
split_start = from_date
return overall_result
def execute(
self,
query: Query,
request_settings: RequestSettings,
runner: SplitQueryRunner,
) -> Optional[QueryResult]:
"""
Split query in 2 steps if a large number of columns is being selected.
- First query only selects event_id, project_id and timestamp.
- Second query selects all fields for only those events.
- Shrink the date range.
"""
limit = query.get_limit()
if (limit is None or limit == 0 or query.get_groupby()
or query.get_aggregations()
or not query.get_selected_columns()):
return None
if limit > settings.COLUMN_SPLIT_MAX_LIMIT:
metrics.increment("column_splitter.query_above_limit")
return None
# Do not split if there is already a = or IN condition on an ID column
id_column_matcher = FunctionCall(
Or([String(ConditionFunctions.EQ),
String(ConditionFunctions.IN)]),
(
Column(None, String(self.__id_column)),
AnyExpression(),
),
)
for expr in query.get_condition_from_ast() or []:
match = id_column_matcher.match(expr)
if match:
return None
# We need to count the number of table/column name pairs
# not the number of distinct Column objects in the query
# so to avoid counting aliased columns multiple times.
total_columns = {(col.table_name, col.column_name)
for col in query.get_all_ast_referenced_columns()}
minimal_query = copy.deepcopy(query)
minimal_query.set_selected_columns(
[self.__id_column, self.__project_column, self.__timestamp_column])
# TODO: provide the table alias name to this splitter if we ever use it
# in joins.
minimal_query.set_ast_selected_columns([
SelectedExpression(self.__id_column,
ColumnExpr(None, None, self.__id_column)),
SelectedExpression(self.__project_column,
ColumnExpr(None, None, self.__project_column)),
SelectedExpression(
self.__timestamp_column,
ColumnExpr(None, None, self.__timestamp_column),
),
])
for exp in minimal_query.get_all_expressions():
if exp.alias in (
self.__id_column,
self.__project_column,
self.__timestamp_column,
) and not (isinstance(exp, ColumnExpr)
and exp.column_name == exp.alias):
logger.warning(
"Potential alias shadowing due to column splitter",
extra={"expression": exp},
exc_info=True,
)
minimal_columns = {
(col.table_name, col.column_name)
for col in minimal_query.get_all_ast_referenced_columns()
}
if len(total_columns) <= len(minimal_columns):
return None
# Ensures the AST minimal query is actually runnable on its own.
if not minimal_query.validate_aliases():
return None
legacy_references = set(minimal_query.get_all_referenced_columns())
ast_column_names = {
c.column_name
for c in minimal_query.get_all_ast_referenced_columns()
}
# Ensures the legacy minimal query (which does not expand alias references)
# does not contain alias references we removed when creating minimal_query.
if legacy_references - ast_column_names:
metrics.increment("columns.skip_invalid_legacy_query")
return None
result = runner(minimal_query, request_settings)
del minimal_query
if not result.result["data"]:
return None
# Making a copy just in case runner returned None (which would drive the execution
# strategy to ignore the result of this splitter and try the next one).
query = copy.deepcopy(query)
event_ids = list(
set([event[self.__id_column] for event in result.result["data"]]))
if len(event_ids) > settings.COLUMN_SPLIT_MAX_RESULTS:
# We may be runing a query that is beyond clickhouse maximum query size,
# so we cowardly abandon.
metrics.increment(
"column_splitter.intermediate_results_beyond_limit")
return None
query.add_conditions([(self.__id_column, "IN", event_ids)])
query.add_condition_to_ast(
in_condition(
None,
ColumnExpr(None, None, self.__id_column),
[LiteralExpr(None, e_id) for e_id in event_ids],
))
query.set_offset(0)
# TODO: This is technically wrong. Event ids are unique per project, not globally.
# So, if the minimal query only returned the same event_id from two projects, we
# would be underestimating the limit here.
query.set_limit(len(event_ids))
project_ids = list(
set([
event[self.__project_column] for event in result.result["data"]
]))
_replace_condition(
query,
self.__project_column,
"IN",
project_ids,
)
_replace_ast_condition(
query,
self.__project_column,
"IN",
literals_tuple(None,
[LiteralExpr(None, p_id) for p_id in project_ids]),
)
timestamps = [
event[self.__timestamp_column] for event in result.result["data"]
]
_replace_condition(
query,
self.__timestamp_column,
">=",
util.parse_datetime(min(timestamps)).isoformat(),
)
_replace_ast_condition(
query,
self.__timestamp_column,
">=",
LiteralExpr(None, util.parse_datetime(min(timestamps))),
)
# We add 1 second since this gets translated to ('timestamp', '<', to_date)
# and events are stored with a granularity of 1 second.
_replace_condition(
query,
self.__timestamp_column,
"<",
(util.parse_datetime(max(timestamps)) +
timedelta(seconds=1)).isoformat(),
)
_replace_ast_condition(
query,
self.__timestamp_column,
"<",
LiteralExpr(
None,
(util.parse_datetime(max(timestamps)) + timedelta(seconds=1)),
),
)
return runner(query, request_settings)
def execute(
self,
query: Query,
query_settings: QuerySettings,
runner: SplitQueryRunner,
) -> Optional[QueryResult]:
"""
Split query in 2 steps if a large number of columns is being selected.
- First query only selects event_id, project_id and timestamp.
- Second query selects all fields for only those events.
- Shrink the date range.
"""
limit = query.get_limit()
if (limit is None or limit == 0 or query.get_groupby()
or not query.get_selected_columns()):
return None
if limit > settings.COLUMN_SPLIT_MAX_LIMIT:
metrics.increment("column_splitter.query_above_limit")
return None
# Do not split if there is already a = or IN condition on an ID column
id_column_matcher = FunctionCall(
Or([String(ConditionFunctions.EQ),
String(ConditionFunctions.IN)]),
(
Column(None, String(self.__id_column)),
AnyExpression(),
),
)
for expr in query.get_condition() or []:
match = id_column_matcher.match(expr)
if match:
return None
# We need to count the number of table/column name pairs
# not the number of distinct Column objects in the query
# so to avoid counting aliased columns multiple times.
selected_columns = {
(col.table_name, col.column_name)
for col in query.get_columns_referenced_in_select()
}
if len(selected_columns) < settings.COLUMN_SPLIT_MIN_COLS:
metrics.increment("column_splitter.main_query_min_threshold")
return None
minimal_query = copy.deepcopy(query)
# TODO: provide the table alias name to this splitter if we ever use it
# in joins.
minimal_query.set_ast_selected_columns([
SelectedExpression(
self.__id_column,
ColumnExpr(self.__id_column, None, self.__id_column),
),
SelectedExpression(
self.__project_column,
ColumnExpr(self.__project_column, None, self.__project_column),
),
SelectedExpression(
self.__timestamp_column,
ColumnExpr(self.__timestamp_column, None,
self.__timestamp_column),
),
])
for exp in minimal_query.get_all_expressions():
if exp.alias in (
self.__id_column,
self.__project_column,
self.__timestamp_column,
) and not (isinstance(exp, ColumnExpr)
and exp.column_name == exp.alias):
logger.warning(
"Potential alias shadowing due to column splitter",
extra={"expression": exp},
exc_info=True,
)
# Ensures the AST minimal query is actually runnable on its own.
if not minimal_query.validate_aliases():
return None
# There is a Clickhouse bug where if functions in the ORDER BY clause are not in the SELECT,
# they fail on distributed tables. For that specific case, skip the query splitter.
for orderby in minimal_query.get_orderby():
if isinstance(orderby.expression,
(FunctionCallExpr, CurriedFunctionCallExpr)):
metrics.increment("column_splitter.orderby_has_a_function")
return None
result = runner(minimal_query, query_settings)
del minimal_query
if not result.result["data"]:
metrics.increment("column_splitter.no_data_from_minimal_query")
return None
# Making a copy just in case runner returned None (which would drive the execution
# strategy to ignore the result of this splitter and try the next one).
query = copy.deepcopy(query)
event_ids = list(
set([event[self.__id_column] for event in result.result["data"]]))
if len(event_ids) > settings.COLUMN_SPLIT_MAX_RESULTS:
# We may be runing a query that is beyond clickhouse maximum query size,
# so we cowardly abandon.
metrics.increment(
"column_splitter.intermediate_results_beyond_limit")
return None
query.add_condition_to_ast(
in_condition(
ColumnExpr(None, None, self.__id_column),
[LiteralExpr(None, e_id) for e_id in event_ids],
))
query.set_offset(0)
query.set_limit(len(result.result["data"]))
project_ids = list(
set([
event[self.__project_column] for event in result.result["data"]
]))
_replace_ast_condition(
query,
self.__project_column,
"IN",
literals_tuple(None,
[LiteralExpr(None, p_id) for p_id in project_ids]),
)
timestamps = [
event[self.__timestamp_column] for event in result.result["data"]
]
_replace_ast_condition(
query,
self.__timestamp_column,
">=",
LiteralExpr(None, util.parse_datetime(min(timestamps))),
)
# We add 1 second since this gets translated to ('timestamp', '<', to_date)
# and events are stored with a granularity of 1 second.
_replace_ast_condition(
query,
self.__timestamp_column,
"<",
LiteralExpr(
None,
(util.parse_datetime(max(timestamps)) + timedelta(seconds=1)),
),
)
return runner(query, query_settings)
def wrapper(*args, **kwargs):
body = args[0]
use_split, date_align, split_step = state.get_configs([
('use_split', 0),
('date_align_seconds', 1),
('split_step', 3600), # default 1 hour
])
to_date = util.parse_datetime(body['to_date'], date_align)
from_date = util.parse_datetime(body['from_date'], date_align)
limit = body.get('limit', 0)
remaining_offset = body.get('offset', 0)
if (use_split and limit and not body.get('groupby')
and body.get('orderby') == '-timestamp'):
overall_result = None
split_end = to_date
split_start = max(split_end - timedelta(seconds=split_step),
from_date)
total_results = 0
status = 0
while split_start < split_end and total_results < limit:
body['from_date'] = split_start.isoformat()
body['to_date'] = split_end.isoformat()
# Because its paged, we have to ask for (limit+offset) results
# and set offset=0 so we can then trim them ourselves.
body['offset'] = 0
body['limit'] = limit - total_results + remaining_offset
result, status = query_func(*args, **kwargs)
# If something failed, discard all progress and just return that
if status != 200:
overall_result = result
break
if overall_result is None:
overall_result = result
else:
overall_result['data'].extend(result['data'])
if remaining_offset > 0 and len(overall_result['data']) > 0:
to_trim = min(remaining_offset,
len(overall_result['data']))
overall_result['data'] = overall_result['data'][to_trim:]
remaining_offset -= to_trim
total_results = len(overall_result['data'])
if total_results < limit:
if len(result['data']) == 0:
# If we got nothing from the last query, jump straight to the max time range
split_end = split_start
split_start = from_date
else:
# Estimate how big the time range should be for the next query based on
# how many results we got for our last query and its time range, and how
# many we have left to fetch
remaining = limit - total_results
split_step = split_step * math.ceil(
remaining / float(len(result['data'])))
split_end = split_start
try:
split_start = max(
split_end - timedelta(seconds=split_step),
from_date)
except OverflowError:
split_start = from_date
return overall_result, status
else:
return query_func(*args, **kwargs)
def parse_and_run_query(validated_body, timer):
body = deepcopy(validated_body)
turbo = body.get('turbo', False)
max_days, table, date_align, config_sample, force_final, max_group_ids_exclude = state.get_configs([
('max_days', None),
('clickhouse_table', settings.CLICKHOUSE_TABLE),
('date_align_seconds', 1),
('sample', 1),
# 1: always use FINAL, 0: never use final, undefined/None: use project setting.
('force_final', 0 if turbo else None),
('max_group_ids_exclude', settings.REPLACER_MAX_GROUP_IDS_TO_EXCLUDE),
])
stats = {}
to_date = util.parse_datetime(body['to_date'], date_align)
from_date = util.parse_datetime(body['from_date'], date_align)
assert from_date <= to_date
if max_days is not None and (to_date - from_date).days > max_days:
from_date = to_date - timedelta(days=max_days)
where_conditions = body.get('conditions', [])
where_conditions.extend([
('timestamp', '>=', from_date),
('timestamp', '<', to_date),
('deleted', '=', 0),
])
# NOTE: we rely entirely on the schema to make sure that regular snuba
# queries are required to send a project_id filter. Some other special
# internal query types do not require a project_id filter.
project_ids = util.to_list(body['project'])
if project_ids:
where_conditions.append(('project_id', 'IN', project_ids))
having_conditions = body.get('having', [])
aggregate_exprs = [
util.column_expr(col, body, alias, agg)
for (agg, col, alias) in body['aggregations']
]
groupby = util.to_list(body['groupby'])
group_exprs = [util.column_expr(gb, body) for gb in groupby]
selected_cols = [util.column_expr(util.tuplify(colname), body)
for colname in body.get('selected_columns', [])]
select_exprs = group_exprs + aggregate_exprs + selected_cols
select_clause = u'SELECT {}'.format(', '.join(select_exprs))
from_clause = u'FROM {}'.format(table)
# For now, we only need FINAL if:
# 1. The project has been marked as needing FINAL (in redis) because of recent
# replacements (and it affects too many groups for us just to exclude
# those groups from the query)
# OR
# 2. the force_final setting = 1
needs_final, exclude_group_ids = get_projects_query_flags(project_ids)
if len(exclude_group_ids) > max_group_ids_exclude:
# Cap the number of groups to exclude by query and flip to using FINAL if necessary
needs_final = True
exclude_group_ids = []
used_final = False
if force_final == 1 or (force_final is None and needs_final):
from_clause = u'{} FINAL'.format(from_clause)
used_final = True
elif exclude_group_ids:
where_conditions.append(('group_id', 'NOT IN', exclude_group_ids))
sample = body.get('sample', settings.TURBO_SAMPLE_RATE if turbo else config_sample)
if sample != 1:
from_clause = u'{} SAMPLE {}'.format(from_clause, sample)
joins = []
if 'arrayjoin' in body:
joins.append(u'ARRAY JOIN {}'.format(body['arrayjoin']))
join_clause = ' '.join(joins)
where_clause = ''
if where_conditions:
where_conditions = list(set(util.tuplify(where_conditions)))
where_clause = u'WHERE {}'.format(util.conditions_expr(where_conditions, body))
prewhere_conditions = []
if settings.PREWHERE_KEYS:
# Add any condition to PREWHERE if:
# - It is a single top-level condition (not OR-nested), and
# - Any of its referenced columns are in PREWHERE_KEYS
prewhere_candidates = [
(util.columns_in_expr(cond[0]), cond)
for cond in where_conditions if util.is_condition(cond) and
any(col in settings.PREWHERE_KEYS for col in util.columns_in_expr(cond[0]))
]
# Use the condition that has the highest priority (based on the
# position of its columns in the PREWHERE_KEYS list)
prewhere_candidates = sorted([
(min(settings.PREWHERE_KEYS.index(col) for col in cols if col in settings.PREWHERE_KEYS), cond)
for cols, cond in prewhere_candidates
])
if prewhere_candidates:
prewhere_conditions = [cond for _, cond in prewhere_candidates][:settings.MAX_PREWHERE_CONDITIONS]
prewhere_clause = ''
if prewhere_conditions:
prewhere_clause = u'PREWHERE {}'.format(util.conditions_expr(prewhere_conditions, body))
having_clause = ''
if having_conditions:
assert groupby, 'found HAVING clause with no GROUP BY'
having_clause = u'HAVING {}'.format(util.conditions_expr(having_conditions, body))
group_clause = ', '.join(util.column_expr(gb, body) for gb in groupby)
if group_clause:
if body.get('totals', False):
group_clause = 'GROUP BY ({}) WITH TOTALS'.format(group_clause)
else:
group_clause = 'GROUP BY ({})'.format(group_clause)
order_clause = ''
if body.get('orderby'):
orderby = [util.column_expr(util.tuplify(ob), body) for ob in util.to_list(body['orderby'])]
orderby = [u'{} {}'.format(
ob.lstrip('-'),
'DESC' if ob.startswith('-') else 'ASC'
) for ob in orderby]
order_clause = u'ORDER BY {}'.format(', '.join(orderby))
limitby_clause = ''
if 'limitby' in body:
limitby_clause = 'LIMIT {} BY {}'.format(*body['limitby'])
limit_clause = ''
if 'limit' in body:
limit_clause = 'LIMIT {}, {}'.format(body.get('offset', 0), body['limit'])
sql = ' '.join([c for c in [
select_clause,
from_clause,
join_clause,
prewhere_clause,
where_clause,
group_clause,
having_clause,
order_clause,
limitby_clause,
limit_clause
] if c])
timer.mark('prepare_query')
stats.update({
'clickhouse_table': table,
'final': used_final,
'referrer': request.referrer,
'num_days': (to_date - from_date).days,
'num_projects': len(project_ids),
'sample': sample,
})
return util.raw_query(
validated_body, sql, clickhouse_ro, timer, stats
)
