Unexpected empty cell with a where condition

[jbe456]

Steps to reproduce

import pandas as pd

dummy_df = pd.DataFrame(
    data={
        "id": [0, 1, 2, 3],
        "Name": ["John", "John", "Thomas", "Mike"],
        "Score": [10, 8, 5, 12],
    }
)

import atoti

session = atoti.create_session()
dummy_store = session.read_pandas(
    dummy_df, keys=["id"], types={"id": atoti.types.STRING, "Score": atoti.types.DOUBLE}
)
cube = session.create_cube(dummy_store)

cube.measures["underlying"] = atoti.agg.single_value(
    atoti.where(cube.measures["Score.SUM"] == 12, 1000),
    scope=atoti.scope.origin(cube.levels["Name"]),
)

cube.measures["final"] = atoti.where(cube.measures["underlying"] == None, 'NOTHING', 'SOMETHING')

Actual Result

The measure final returns an empty cell for Total Name although underlying has a value:

Expected Result

I would expect the final measure to return "SOMETHING" for "Total Name"

Environment

• Atoti: 0.4.0.20200512194154
• Python: 3.8.2
• Operating System: linux

[fabiencelier]

This is related to the issue #17 and the comment about the where: single_value adds Name as a required level and where uses this required level to do a "per" which make it null when Name is not expressed.
The issue is that singe value should not have required levels.

This can be reproduced like that:

m["test"] = atoti.agg.sum(
    m["Score.SUM"],
    scope=atoti.scope.origin(lvl["Name"])
)
m["test"]._required_levels # should be empty but isn't

[fabiencelier]

My bad, I was wrong.
All the aggregation functions add required levels when using scope, but they are only transmit it to the next function in the case of single_value. If you do your use case with a tt.agg.sum instead of the single value you will have the expected result.

I will change the behavior of single value so it works as the other aggregation functions. It will only transmit the required levels if it is a store measure (m["test"] = store["test"])

[fabiencelier]

Fixed in 0.4.0