Step 4: Cycles¶
Cycles have been introduced to reflect business situations our customers frequently encounter.
For instance, a large Fast Food chain wants to generate sales forecasts for its stores every week. When creating a given scenario, it will need to be attached to a given week. And often, a single one will be published amongst all the scenarios generated for a given week. This kind of 'official' scenario will be referred to as the 'Primary' scenario in Taipy Core.
Note that Cycles can be ignored entirely if the business problem has no time frequency.
In this step, scenarios are attached to a MONTHLY cycle. Using Cycles, the developer will benefit from specific Taipy's functions to navigate through these Cycles. For instance, by providing the Cycle, Taipy can get all the scenarios created in a month. You can also easily get every primary scenario generated for the past X months to monitor KPIs over time.
Let’s slightly change the filter function by passing the month as an argument to get started. You must create a new Data Node representing the month (see the steps below).
def filter_by_month(df, month):
df['Date'] = pd.to_datetime(df['Date'])
df = df[df['Date'].dt.month == month]
return df
Then to introduce Cycles, you need to set the frequency (predefined attribute) of the scenario to Monthly (as described below).
Configuration
-
Recreate the config of the previous step but change the task accordingly with a new input Data Node (month).
-
Add the frequency property for the scenario and put "MONTHLY:FREQUENCY" (DAYLY, WEEKLY, MONTHLY, YEARLY)
-
Load the new configuration in the code
The configuration is the same as the last step except for the scenario and task configuration. A new parameter is added for the frequency.
from taipy.config import Scope
month_cfg = Config.configure_data_node(id="month")
task_filter_cfg = Config.configure_task(id="filter_by_month",
function=filter_by_month,
input=[historical_data_cfg, month_cfg],
output=month_values_cfg)
...
scenario_cfg = Config.configure_scenario(id="my_scenario",
pipeline_configs=[pipeline_cfg],
frequency=Frequency.MONTHLY)
As you can see, a Cycle is activated once you have set the desired frequency on the scenario. In this code snippet, since we have specified frequency=Frequency.MONTHLY, the corresponding scenario will be automatically attached to the correct period (month) once it is created. The creation_date here is artificially given to the scenarios.
tp.Core().run()
scenario_1 = tp.create_scenario(scenario_cfg,
creation_date=dt.datetime(2022,10,7),
name="Scenario 2022/10/7")
scenario_2 = tp.create_scenario(scenario_cfg,
creation_date=dt.datetime(2022,10,5),
name="Scenario 2022/10/5")
Scenario 1 and Scenario 2 are two scenario entities/instances created from the same scenario configuration. They belong to the same Cycle but don't share the same Data Nodes. By default, each scenario instance has its own data node instances. They are not shared with any other scenario. The Scope concept can modify this behavior, which will be covered in the next step.
scenario_1.month.write(10)
scenario_2.month.write(10)
print("Month Data Node of Scenario 1", scenario_1.month.read())
print("Month Data Node of Scenario 2", scenario_2.month.read())
scenario_1.submit()
scenario_2.submit()
Results:
Month Data Node of Scenario 1 10
Month Data Node of Scenario 2 10
[2022-12-22 16:20:04,746][Taipy][INFO] job JOB_filter_by_month_a4d3c4a7-5ec9-4cca-8a1b-578c910e255a is completed.
[2022-12-22 16:20:04,833][Taipy][INFO] job JOB_count_values_a81b2f60-e9f9-4848-aa58-272810a0b755 is completed.
[2022-12-22 16:20:05,026][Taipy][INFO] job JOB_filter_by_month_22a3298b-ac8d-4b55-b51f-5fab0971cc9e is completed.
[2022-12-22 16:20:05,084][Taipy][INFO] job JOB_count_values_a52b910a-4024-443e-8ea2-f3cdda6c1c9d is completed.
[2022-12-22 16:20:05,317][Taipy][INFO] job JOB_filter_by_month_8643e5cf-e863-434f-a1ba-18222d6faab8 is completed.
[2022-12-22 16:20:05,376][Taipy][INFO] job JOB_count_values_72ab71be-f923-4898-a8a8-95ec351c24d9 is completed.
Primary scenarios¶
In each Cycle, there is a primary scenario. A primary scenario is interesting because it represents the important scenario of the Cycle, the reference. By default, the first scenario created for a cycle will be primary.
tp.set_primary(<Scenario>) allows changing the primary scenario in a Cycle.
<Scenario>.is_primary identifies as a boolean whether the scenario is primary or not.
print("Scenario 1 before", scenario_1.is_primary)
print("Scenario 2 before", scenario_2.is_primary)
tp.set_primary(scenario_2)
print("Scenario 1 after", scenario_1.is_primary)
print("Scenario 2 after", scenario_2.is_primary)
Results:
Scenario 1 before True
Scenario 2 before False
Scenario 1 after False
Scenario 2 after True
Scenario 3 is the only scenario in another Cycle due to its creation date and is the default primary scenario.
scenario_3 = tp.create_scenario(scenario_cfg,
creation_date=dt.datetime(2021,9,1),
name="Scenario 2022/9/1")
scenario_3.month.write(9)
scenario_3.submit()
print("Is scenario 3 primary?", scenario_3.is_primary)
Results:
[2022-12-22 16:20:05,317][Taipy][INFO] job JOB_filter_by_month_8643e5cf-e863-434f-a1ba-18222d6faab8 is completed.
[2022-12-22 16:20:05,376][Taipy][INFO] job JOB_count_values_72ab71be-f923-4898-a8a8-95ec351c24d9 is completed.
Is scenario 3 primary? True
Also, as you can see, every scenario has been submitted and executed entirely. However, the results for these tasks are all the same. Skipping Tasks (defined in subsequent steps) will help optimize your executions by skipping the execution of redundant tasks.
Useful functions on cycles¶
-
tp.get_primary_scenarios(): returns a list of all primary scenarios -
tp.get_scenarios(cycle=<Cycle>): returns all the scenarios in the Cycle -
tp.get_cycles(): returns the list of Cycles -
tp.get_primary(<Cycle>): returns the primary scenario of the Cycle