<#6340 [BUG] Flyte Conditional is_true() goes to i...
# flyte-github
a
#6340 [BUG] Flyte Conditional is_true() goes to incorrect code branch in python testing but is_false() works. Issue created by Titash21 Describe the bug I am using Flytekit v1.15.1. When testing an entire workflow file which contains Conditionals and sub_workflows, when a variable is passed to a Flyte Conditional, passing the right parameters still calls the incorrect code block. This is the python test code:
Copy code
def test_ml_workflow(mocker):
    bucket_name = "test-bucket"
    prefix = "test-prefix"
    model_path = ""

    with task_mock(load_pretrained_model) as load_pretrained_model_mocked:

        # collect_data, no cold_start
        ml_workflow(bucket_name=bucket_name, prefix=prefix, model_path=model_path, collect_data=True, cold_start=False)
        print("-" * 30)
• In the above test, even though
cold_start=False
is passed, the flyte calls
cold_start_data_collect()
method (see code below). • It actually goes inside both the methods:
data_collect()
and
cold_start_data_collect()
. • If the
is_true()
is changed to
is_false()
only for the lowest level Conditional, the error is resolved and only
cold_start_data_collect()
is called correctly. • Workflow code is below.
Copy code
import datetime  
import joblib  
import pandas as pd  
from flytekit import workflow, conditional, task, StructuredDataset, LaunchPlan, CronSchedule  
from flytekit.core.task import Echo
@task  
def get_date_from_datetime() -> datetime.date:  
return datetime.datetime.now().date()
@task  
def cold_start_data_collect() -> bool:  
date_result = get_date_from_datetime()  
print("Inside cold_start_data_collect", date_result)  
data = {'col1': [1, 2, 3], 'col2': ['A', 'B', 'C']}  
final_df = pd.DataFrame(data)  
return True
@task  
def data_collect() -> bool:  
print("Inside data_collect")  
data = {'col1': [1, 2, 3], 'col2': ['A', 'B', 'C']}  
raw_data = pd.DataFrame(data)  
return True
@task  
def preprocess_data() -> bool:  
return True
@task  
def model_training() -> bool:  
return True
@task  
def load_pretrained_model(model_path: str) -> bool:  
print("Inside load_pretrained_model")  
loaded_model = joblib.load(model_path)  
return True
echo = Echo(name="no_data_collection", inputs={"data_collect": bool})
@workflow  
def data_collection_workflow(bucket_name: str, prefix: str, cold_start: bool) -> bool:  
level_3 = (conditional("cold_start?")  
.if_(cold_start.is_true())  
.then(data_collect())  
.else_()  
.then(cold_start_data_collect()))  
return True
@workflow  
def old_workflow(collect_data:bool, bucket_name:str, prefix:str, cold_start:bool) -> bool:  
print("Inside old_workflow")  
level_2 = (conditional("cold_start?")  
.if_(collect_data.is_true())  
.then(data_collection_workflow(bucket_name=bucket_name, prefix=prefix, cold_start=cold_start))  
.else_()  
.then(echo(data_collect=False)))  
return True
@workflow  
def ml_workflow(bucket_name: str, prefix: str, model_path: str, collect_data: bool = True, cold_start: bool = False):  
level_1 = (  
conditional("collect_data ?")  
.if_(collect_data.is_true())  
.then(old_workflow(  
collect_data=collect_data,  
bucket_name=bucket_name,  
prefix=prefix,  
cold_start=cold_start  
))  
.else_()  
.then(load_pretrained_model(model_path))  
)
cron_lp = LaunchPlan.get_or_create(  
name="test_workflow_bug_lp",  
workflow=ml_workflow,  
default_inputs={  
"bucket_name": "",  
"prefix": "",  
"model_path": "",  
"collect_data": True,  
"cold_start": False  
}  
)
if *name* == "*main*":  
ml_workflow()
Expected behavior For this method, call the correct code branch ie
cold_start_data_collect()
when
cold_start=True
.
Copy code
@workflow
def data_collection_workflow(bucket_name: str, prefix: str, cold_start: bool) -> bool:
    level_3 = (conditional("cold_start?")
                .if_(cold_start.is_true())
                .then(data_collect())
                .else_()
                .then(cold_start_data_collect()))
    return True
Additional context to reproduce No response Screenshots I am attaching two screenshots with the outputs. Image Image Are you sure this issue hasn't been raised already? • Yes Have you read the Code of Conduct? • Yes flyteorg/flyte