fix: Variable.fix() value alignment on named dimensions

[FBumann]

This came up during working on another PR.
Small but important bugfix. Very obvious and easy since we introduced clean unternal helpers for such things.

Note

AI-assisted implementation.

Bug

Variable.fix(value) aligned the value with as_dataarray(value).broadcast_like(self.labels), which aligns only by dimension name. That happens to work only for the default dim_0. On a variable with a named dimension, a positional value (list/array) gained a spurious dim_0 and was broadcast across the real dimension instead of aligned onto it — silently producing a wrong fix constraint.

import linopy, pandas as pd

m = linopy.Model()
t = m.add_variables(lower=-5, upper=5, coords=[pd.Index([2020, 2030, 2040], name="time")], name="t")
t.fix([1.0, 2.0, 3.0])

# before: __fix__t constraint has dims (time: 3, dim_0: 3) — a 3×3 block fixing
#         every t[time] to all of 1, 2, 3 at once (infeasible); only a soft warning
# after:  __fix__t has dims (time: 3) with rhs [1, 2, 3]

This affects the released v0.7.0 fix().

Fix

Use broadcast_to_coords against the variable's own coords — the same coords-aware alignment add_variables uses for lower/upper:

• scalars broadcast across the dimension,
• positional inputs (list / ndarray) land on the right dimension,
• named pandas/xarray inputs align by coordinate value (reordered indexes reindex),
• a value carrying an unknown dimension raises an error that names the variable.

Tests

TestFixValueAlignment in test_fix_relax.py is parameterized over the value dtype — scalar (broadcast), list, ndarray, named Series (in order and reordered), and a named DataArray — asserting the fix constraint aligns to the named dimension (dims == ("time",)) with correct values, plus an unknown-dimension rejection case.

🤖 Generated with Claude Code

[FabianHofmann]

great catch, will pull this in quickly, just have to check one thing

[FabianHofmann]

all resolved - added a test class for covering more cases of alignment. (could argue it duplicates parts of the tests of broadcast_to_coords, but it's good to have them here as well to avoid regression)