Inconsistent behavior when multiplying with subset coordinates

[FabianHofmann]

Version Checks

• I have confirmed this bug exists on the current master branch of Linopy.

Issue Description

Multiplying variables/expressions with constants that have a subset of coordinates produces inconsistent behavior:

1. Variable * subset_const: Creates NaN coefficients with valid variable labels
2. LinearExpression * subset_const: Fails with AssertionError in _multiply_by_constant
3. Variable + subset_const: Works correctly (fills missing with 0)

The root cause is that addition uses merge() with controlled join='outer' and fill_value=0, while multiplication uses uncontrolled xarray operations.

Reproducible Example

import xarray as xr
import pandas as pd
import linopy

m = linopy.Model()
x = m.add_variables(coords=[pd.Index(range(5), name='i')])
subset = xr.DataArray([10.0, 30.0], dims=['i'], coords={'i': [1, 3]})

# Works but produces NaN coefficients with valid var labels
result1 = x * subset
print(result1.coeffs.squeeze().values)  # [nan, 10., nan, 30., nan]
print(result1.data.vars.squeeze().values)  # [0, 1, 2, 3, 4] - all valid!

# Fails with AssertionError
expr = 1 * x
result2 = expr * subset  # AssertionError in _multiply_by_constant

# Addition works correctly
result3 = x + subset
print(result3.const.values)  # [0., 10., 0., 30., 0.]

Expected Behavior

Multiplication should behave consistently:

• Either fill missing coordinates with 0 (like addition)
• Or raise an error for misaligned coordinates
• Both x * c and (1*x) * c should behave the same

Installed Versions

Details

linopy @ master (commit 59f92ae)

[FBumann]

Check again in #591

[FBumann]

Re-tested on linopy 0.7.0: the inconsistency reported here — Variable * subset producing NaN coefficients, LinearExpression * subset raising AssertionError, the two paths disagreeing — is fixed. Both now give [0, 10, 0, 30, 0] consistently, no crash:

import pandas as pd, xarray as xr, linopy
m = linopy.Model()
x = m.add_variables(coords=[pd.RangeIndex(5, name="i")])
subset = xr.DataArray([10.0, 30.0], dims=["i"], coords={"i": [1, 3]})
(x * subset).coeffs.squeeze().values        # [ 0. 10.  0. 30.  0.]
((1 * x) * subset).coeffs.squeeze().values  # [ 0. 10.  0. 30.  0.]  — same

What remains is the fill itself: multiplying by a subset constant silently sets the coefficient to 0 on the uncovered coordinates (i = 0, 2, 4), dropping those variables from the expression.

Suggest keeping this issue open, reframed around that residual — a subset constant in multiplication silently drops variables — which the v1 arithmetic convention (#591) fixes by raising instead. (#569, purely about the now-fixed inconsistency, can be closed.)