def resolve_condition(cond):
"""
When conditions have been parsed by the api.event_search module
we can end up with conditions that are not valid on the current dataset
due to how ap.event_search checks for valid field names without
being aware of the dataset.
We have the dataset context here, so we need to re-scope conditions to the
current dataset.
"""
index = get_function_index(cond)
if index is not None:
# IN conditions are detected as a function but aren't really.
if cond[index] == "IN":
cond[0] = resolve_column(cond[0])
return cond
func_args = cond[index + 1]
for (i, arg) in enumerate(func_args):
# Nested function
if isinstance(arg, (list, tuple)):
func_args[i] = resolve_condition(arg)
else:
func_args[i] = resolve_column(arg)
cond[index + 1] = func_args
return cond
# No function name found
if isinstance(cond, (list, tuple)) and len(cond):
# Condition is [col, operator, value]
if isinstance(cond[0], six.string_types) and len(cond) == 3:
cond[0] = resolve_column(cond[0])
return cond
if isinstance(cond[0], (list, tuple)):
if get_function_index(cond[0]) is not None:
cond[0] = resolve_condition(cond[0])
return cond
else:
# Nested conditions
return [resolve_condition(item) for item in cond]
raise ValueError("Unexpected condition format %s" % cond)
def parse_columns_in_functions(col, context=None, index=None):
"""
Checks expressions for arguments that should be considered a column while
ignoring strings that represent clickhouse function names
if col is a list, means the expression has functions and we need
to parse for arguments that should be considered column names.
Assumptions here:
* strings that represent clickhouse function names are always followed by a list or tuple
* strings that are quoted with single quotes are used as string literals for CH
* otherwise we should attempt to get the snuba column name (or custom tag)
"""
function_name_index = get_function_index(col)
if function_name_index is not None:
# if this is non zero, that means there are strings before this index
# that should be converted to snuba column names
# e.g. ['func1', ['column', 'func2', ['arg1']]]
if function_name_index > 0:
for i in range(0, function_name_index):
if context is not None:
context[i] = get_snuba_column_name(col[i])
args = col[function_name_index + 1]
# check for nested functions in args
if get_function_index(args):
# look for columns
return parse_columns_in_functions(args, args)
# check each argument for column names
else:
for (i, arg) in enumerate(args):
parse_columns_in_functions(arg, args, i)
else:
# probably a column name
if context is not None and index is not None:
context[index] = get_snuba_column_name(col)