Performance: only check independence for filled component scenarios

[mgovers]

In the current main, for every component, every scenario in a batch was checked to determine the update independence, even if there are no components of that type in the entire batch. That is, of course, not needed.

This PR changes the iteration over all scenarios to an iteration over all filled scenarios. This is much more efficient. In particular, there is an early-out if there are no components of a type in the update dataset, and another one if the batch update is not uniform for that component.

Test

Script

import numpy as np

from power_grid_model import (
    AttributeType as AT,
    ComponentType as CT,
    DatasetType as DT,
    PowerGridModel,
    initialize_array,
)

node = initialize_array(DT.input, CT.node, 1)
node[AT.id] = 0
node[AT.u_rated] = 100.0

source = initialize_array(DT.input, CT.source, 1)
source[AT.id] = 1
source[AT.node] = 0
source[AT.status] = 1
source[AT.u_ref] = 1.0
source[AT.sk] = 1e20
source[AT.rx_ratio] = 0.0

sym_load = initialize_array(DT.input, CT.sym_load, 1)
sym_load[AT.id] = 2
sym_load[AT.node] = 0
sym_load[AT.status] = 1
sym_load[AT.type] = 2
sym_load[AT.p_specified] = 0.0
sym_load[AT.q_specified] = 500.0

input_data = {
    CT.node: node,
    CT.source: source,
    CT.sym_load: sym_load,
}

source_update = initialize_array(DT.update, CT.source, (100_000, 1))
source_update[AT.id] = 1
source_update[AT.u_ref] = 1
source_update[AT.u_ref][0] = 1.1

gigantic_update_data = {CT.source: source_update}

model = PowerGridModel(input_data)
print("Starting calculation...")
result = model.calculate_power_flow(update_data=gigantic_update_data)
print("Done.")

np.testing.assert_allclose(result[CT.node][AT.u][1:], 100.0, rtol=0.0, atol=1e-8)
np.testing.assert_allclose(result[CT.node][AT.u][0], 110.0, rtol=0.0, atol=1e-8)

Current main

All the different components all take up some amount of time in the calculations. While small for one component, the sheer amount of component types adds up. This is even more so for large batch calculations

This branch

Only the source component (the only component that is updated in the update data) has any significant time being spent on

[Copilot]

Pull request overview

This PR optimizes update-independence checking by iterating only over scenarios that actually contain component data (“filled scenarios”), avoiding unnecessary per-scenario work when a component is absent/empty in the batch.

Changes:

• Replaced “all scenarios” buffer-span APIs with “filled scenarios” variants and updated call sites accordingly.
• Updated independence checking to use filled-scenario spans (with early-outs) instead of scanning every scenario.
• Adjusted unit tests to validate the new filled-scenario span behavior for row-based and columnar datasets.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 3 comments.

File	Description
tests/cpp_unit_tests/test_dataset.cpp	Updates tests to validate filled-scenario span enumeration instead of all-scenario enumeration.
power_grid_model_c/power_grid_model/include/power_grid_model/main_core/update.hpp	Uses filled-scenario spans to reduce work during independence checks.
power_grid_model_c/power_grid_model/include/power_grid_model/auxiliary/dataset.hpp	Introduces filled-scenario span views and refactors span retrieval helpers.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[TonyXiang8787]

@mgovers can you put some demostrative example in your PR description to show what exactly is improved? By looking at code change it is not that clear.

[mgovers]

@mgovers can you put some demostrative example in your PR description to show what exactly is improved? By looking at code change it is not that clear.

done

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
97.9% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud