def process_operations(db, job, phenotype: PhenotypeModel, phenotype_id, phenotype_owner, c: PhenotypeOperations,
final=False):
try:
evaluator = util.expression_evaluator
except:
evaluator = 'mongo'
# the NLPQL expression to be evaluated
expression = c['raw_text']
# the NLPQL feature name to assign to the result
nlpql_feature = c['name']
mongo_failed = False
if 'mongo' == evaluator:
log('Using mongo evaluator for expression "{0}"'.format(expression))
# The validate_phenotype function parses the expression and checks it
# for various errors. A normalized version of the expression is then
# stored in the operations dict for that expression. If validation has
# not been performed for some reason, the normalized expression will
# contain the empty string.
parse_result = c['normalized_expr']
if 0 == len(parse_result):
# get the names of the phenotype's data_entities and operations
names = get_all_names(phenotype)
# Parse the expression and return a fully-parenthesized version
# that uses mnemonics for the operators. The evaluator will attempt
# to resolve any unknown tokens into concatenated names and logic
# operators. If it finds a token that it cannot resolve into known
# names it returns an empty string. An empty string is also
# returned if the expression cannot be evaluated for some other
# reason.
parse_result = expr_eval.parse_expression(expression, names)
if 0 == len(parse_result):
log('\n\t*** Expression cannot be evaluated. ***\n')
mongo_failed = True
else:
# generate a list of expr_eval.ExpressionObject items
expr_list = expr_eval.generate_expressions(nlpql_feature, parse_result)
if 0 == len(expr_list):
log('\t\n*** No subexpressions found! ***\n')
mongo_failed = True
else:
mongo_process_operations(expr_list, db, job, phenotype,
phenotype_id, phenotype_owner, c, final)
if 'pandas' == evaluator or mongo_failed:
log('Using pandas evaluator for expression "{0}"'.format(expression))
pandas_process_operations(db, job, phenotype, phenotype_id, phenotype_owner, c, final)
def validate_phenotype(p_cfg: PhenotypeModel):
error = None
try:
if not error:
if not p_cfg:
error = "Empty phenotype object"
if not error and len(p_cfg.data_entities) == 0:
error = "Must have at least one data entity (define)"
if not error and len(p_cfg.operations) > 0 and len(
p_cfg.data_entities) == 0:
error = "Operations (define) require at least one data entity (define)"
except Exception as ex:
log(ex)
error = ''.join(traceback.format_stack())
# Run validity and syntax checks on all expressions, ensure that only
# defined names are used as variables, etc.
name_list = get_all_names(p_cfg)
# get raw text of all expressions
KEY_RAW = 'raw_text'
expression_list = []
for i, op in enumerate(p_cfg.operations):
if KEY_RAW in op:
# save expression index and raw text
expression_list.append((i, op[KEY_RAW]))
for i, expr in expression_list:
# The 'parse_result' is a string of whitespace-separated expression
# tokens. Invalid expressions cause an empty string to be returned.
parse_result = expr_eval.parse_expression(expr, name_list)
if 0 == len(parse_result):
error = 'Invalid expression: {0}'.format(expr)
break
else:
# saved the parse result for later use
p_cfg.operations[i]['normalized_expr'] = parse_result
if not error:
return {"success": True}
else:
return {"success": False, "error": error}
def process_operations(db, job, phenotype: PhenotypeModel, phenotype_id, phenotype_owner, c: PhenotypeOperations,
final=False):
try:
evaluator = util.expression_evaluator
except:
evaluator = 'pandas'
# the NLPQL expression to be evaluated
expression = c['raw_text']
# the NLPQL feature name to assign to the result
nlpql_feature = c['name']
mongo_failed = False
if 'mongo' == evaluator:
print('Using mongo evaluator for expression "{0}"'.format(expression))
# get the names of the phenotype's data_entities and operations
names = get_all_names(phenotype)
# Parse the expression and return a fully-parenthesized version
# that uses mnemonics for the operators. The evaluator will attempt
# to resolve any unknown tokens into concatenated names and logic
# operators. If it finds a token that it cannot resolve into known
# names it returns an empty list. An empty list is also returned
# if the expression cannot be evaluated for some other reason.
parse_result = expr_eval.parse_expression(expression, names)
if 0 == len(parse_result):
print('\n\t*** Expression cannot be evaluated. ***\n')
mongo_failed = True
else:
# generate a list of expr_eval.ExpressionObject items
expr_list = expr_eval.generate_expressions(nlpql_feature, parse_result)
if 0 == len(expr_list):
print('\t\n*** No subexpressions found! ***\n')
mongo_failed = True
else:
mongo_process_operations(expr_list, db, job, phenotype,
phenotype_id, phenotype_owner,c, final)
if 'pandas' == evaluator or mongo_failed:
print('Using pandas evaluator for expression "{0}"'.format(expression))
pandas_process_operations(db, job, phenotype, phenotype_id, phenotype_owner, c, final)