[olivier-lacroix]

A minor performance improvement, discovered while profiling some LP models.

Note

The following content was generated by AI.

What

Model.assign_result builds a DataArray per variable and per constraint, then lets the public setters merge it into each container's dataset. After a solve the shape and coordinates are already correct, so writing the arrays into a shallow copy is enough.

• common.set_data_var writes one data variable without xarray's merge pass.
• Variable._assign_solution and Constraint._assign_dual use it.
• Model.assign_result calls both.

The public Variable.solution and Constraint.dual setters are unchanged, so nothing else moves.

Benchmark

Scenario: 20 variables and 20 constraints over 2000 timesteps, median of 10 assign_result calls.

checkout	assign_result
master	15.91 ms/call
this branch	0.78 ms/call

Reproduce

import statistics
import time

import pandas as pd

import linopy
from linopy.solvers import Solver

n_time, n_streams, reps = 2000, 20, 10
index = pd.RangeIndex(n_time, name="time")
model = linopy.Model()
variables = [
    model.add_variables(lower=0, coords=[index], name=f"x{i}")
    for i in range(n_streams)
]
for i, var in enumerate(variables):
    model.add_constraints(var + variables[(i + 1) % n_streams] <= 10, name=f"c{i}")
model.add_objective(sum(v.sum() for v in variables))

solver = Solver.from_name("highs", io_api="direct", options={"output_flag": False})
result = solver.solve(model=model)
model.assign_result(result)

timings = []
for _ in range(reps):
    start = time.perf_counter()
    model.assign_result(result)
    timings.append(time.perf_counter() - start)
print(f"{statistics.median(timings) * 1000:.2f} ms/call")

AI assistance

Drafted with Claude Code: the implementation and the measurement. Reviewed by hand before opening.

[codspeed]

Merging this PR will not alter performance

✅ 181 untouched benchmarks
⏩ 181 skipped benchmarks¹

---

_{Comparing olivier-lacroix:perf/assign-result-arrays (89c4c83) with master (51c4977)}

Footnotes

1. 181 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[olivier-lacroix]

Thx @FabianHofmann !