Relationship to linopy

Everything about linopy in one place, for a reader who arrives from linopy or PyPSA. There are three separate relationships, and keeping them apart keeps the rest of the docs quiet:

	What	Where it matters
Not a dependency	solving a model never imports it	packaging
The oracle	how we know the answers are right	testing
The lane	the second thing a file can be built as	what a caller chooses

1. It is not a runtime dependency

lps.solve, lps.build, lps.write and lps.check go YAML → polars → HiGHS or file, and import nothing from linopy, xarray or pandas. The bare-install job runs the whole suite with none of the three present.

pip install "lpspec[linopy]" adds linopy, xarray and pandas. The extra buys the lane below and the to_pandas / to_dataarray bridges out of a result, nothing else. The lane is a peer, not a fallback: nothing routes to it, and a bare install is a complete one.

Nothing a bare install can reach names linopy, including a traceback. The public exception tree is rooted at LpspecError, with no alias (#389).

2. It is the oracle

Correctness here is the same YAML, built both ways, produces the same model. The differential suite builds a model through the relational engine and through linopy, and compares the two.

The comparison means something only because both paths consume the same resolved AST, the narrow waist in the architecture notes. If each path resolved names on its own, the suite would compare two dialects rather than check one language.

The oracle has one blind spot: a shared misreading passes the differential suite green. Only a published optimum from outside catches it, and docs/examples/index.md is where those live.

Where a concept is already linopy's, lpspec copies its name. Solve statuses; status and termination_condition as two axes with is_ok as the rollup; the shape of a result. A second vocabulary for one fact taxes a reader arriving from linopy or PyPSA. But copy it, do not import it. The engine may not import linopy, so the tables live here. A test imports linopy to assert the copy still matches. A copy nobody checks is a copy that rots.

3. It is a lane

A lane is one of the two ways a spec is executed. This one builds the same file as a linopy.Model instead of attaching data relationally, and the caller picks it by an import. The call is the one lps.build takes: the same first argument (a path, a mapping, or a spec the language has already read), the same sources, the same index sources.

from lpspec import linopy as lpspec_linopy

m = lpspec_linopy.build('spec.yaml', {...})  # -> linopy.Model
m.solve(...)
lpspec_linopy.expression(m, 'spec.yaml', 'co2', {...})  # a named quantity, read back

Both calls are pure: YAML in, a model or a value out, nothing retained. build returns a plain linopy.Model with no accessor, no attached schema and no patched attributes, so nothing is lost across pickle, deepcopy or to_netcdf. To inspect the math, re-read the file with to_spec. expression is the reader, and the same purity forces it to take sources again. It evaluates a declared named expression on the solved model and hands back linopy's native .solution. That is the eager half of result.expression(name), so the differential suite can hold the two lanes to one answer.

This lane constructs; it does not attach. Math for a linopy.Model that something else built, a PyPSA network say, has no verb here (#845). Such a verb would be the one file allowed to reference names it did not declare. That exception costs the whole language layer for one use case. Build a second model and merge it.

What a construct becomes

What lpspec.linopy.build calls for each thing a file can say. Each row lives in linopy/builder.py, one section per group below.

Declaration	linopy
variables:	Model.add_variables(lower, upper, coords, name, mask, binary, integer)
sos:	Model.add_sos_constraints(variable, sos_type, sos_dim, big_m), the block handed over rather than a formulation rebuilt
constraints:	Model.add_constraints(lhs, sign, rhs, name, mask), one rule per declaration
objective:	Model.add_objective(expr, sense), each additive term summed over the dims it carries
expressions:	evaluated at the solution as xarray arithmetic, every variable its .solution and every dual(c) the constraint's .dual; an entry the math never reads is read at whatever degree it was written

In an expression	linopy or xarray
x — a variable	Model.variables['x'], .fillna(0) under absence: zero
p — a parameter	its xr.DataArray, .fillna(0.0) where it stands as a coefficient
+ - * /	the Python operators linopy overloads
sum(x, over=t)	.sum('t')
sum(x, by=lk)	the lookup attached as a coordinate, then .groupby(), reindexed onto the target dimension's declared labels; by=[lk1, lk2] groups by both at once, and a lookup declaring per: groups by those dims too, so they pass through
at(p, by=lk)	.sel({into: lookup}), xarray's vectorised selection; one entry per lookup reads a tuple of labels at once, and a lookup declaring per: is an indexer over those dims as well, read pointwise on them
shift(x, over=t, offset=n)	.shift({t: n}); .roll({t: n}) under edge: wrap; a .sel() gather where the offset differs per entity or by= groups it
sum_back(x, over=t, within=w)	a sum of w scalar gathers, each unreachable position contributing zero; under by= each gather reads inside the group, so the window stops at its edge
dual(c)	Model.constraints['c'].dual, at a read only; the language keeps a dual out of the math, and a solve that stored none refuses the read

A where:	linopy
on a declaration	the mask= argument; a mask that excludes nothing is passed as None
defined(x)	Model.variables['x'].labels != -1, linopy's own marker for an absent slot
a comparison	the Python comparison operators element-wise, absence reading as false

Absence has no single row. It is positional: a missing parameter row is zero in a coefficient, an error in bounds:, and false in a where operand. linopy/absence.py holds all four spellings, and the builder calls them qualified, as absence.coefficient(...), so a reader meets the name at the call.

The same language, and the same data

The lane accepts exactly the same language, which is what makes the oracle an oracle. The equality is structural: both lanes run the same to_program gate. A construct one lane refuses, the other refuses in the same sentence, never with a redirection to the other lane.

Accepting is not building, and two constructs part the lanes, one in each direction. Neither is a language limit: both files pass check, and each is built by the lane the other cannot. A LaneError names the wall and the route around it, and that is what parts it from a language error.

The first is this lane's wall: an objective carrying a constant. The expression setter of linopy.Objective rejects any expression whose const is nonzero: "Constant values in objective function not supported." There is no slot to put one in, which is why PyPSA carries n.objective_constant out of band. So examples/ports/osemosys_utopia.yaml, whose objective owes a fixed cost on capacity that already stood in 1990, builds relationally and not here. Dropping the constant is the one repair that must not happen. The lane is the oracle, and a quietly shortened objective would recalibrate every differential test on such a model to the wrong number. Adding the constant back as a variable pinned to [1, 1] reaches the right answer and was refused too. It puts a column on the caller's model that the other lane does not have. So builder.py checks for a constant before linopy is asked and raises LaneError, naming the wall and the route that does build the model. tests/test_corpus_parity.py carries the strict xfail, typed to that error rather than to any ValueError. The day linopy grows a slot, the test XPASSes and the check comes out with it (#894).

The second is the relational lane's wall, and it is the mirror: an operator acting along a dimension that a constant part does not carry, beside a term that does. Take sum(x * k + d, over=t) where d is a scalar. The relational lane compiles a constant part as its own table. A fragment with no rows for t has no slots for the operator to act on. Under a mask, only the rows know which slots those are. This lane has no such split. The operand is one masked expression, so the constant is dropped wherever the term is, and the lane builds the file as written.

It is one wall, reached by all four operators that act along a dimension (sum(over=), sum(by=), shift, sum_back). That is why they share one refusal rather than each wording its own: a fix for one that left the others would fix a symptom. The relational lane names the rewrite that reaches the same number: declare the parameter over the dimension and supply it there (#1137).

Finding that wall turned up a real disagreement behind it. sum_back read a constant at a slot the variable was absent from, where every other operator drops it. So the two lanes answered 2.5 and 3.0 on a file neither refused. A reduction consumes its operand before any row exists, so absence has to be pushed into the operand first. sum and sum(by=) did that and the window did not. The fix gave the window the same pass, with a differential test over every operator that moves along a dimension (#1142).

The lane takes the same data too (#60). It reads every shape the data contract accepts and follows every index rule in where coordinates come from. A malformed source gets the same refusal from both lanes, in the same sentence. So one sources mapping goes to either lane, and an import alone decides which lane builds a file.

Parts of linopy not taken

lpspec does not take array operations (merge, reindex, stack), the Python modeling API, or the solver layer. The first is data prep (the limits). The second is hard rule 5: the model is the file you review and diff. The third is #106, where lpspec adopts linopy's design for declared solver capabilities without adopting its code.

The modeling API is what a reader arriving from linopy misses first. Two notebook pages replace it. Change a model covers the loops: update for new numbers, a longer table for more rows, a patched dict for new math. Fix, relax, remove covers the verbs, the same loops aimed at fix, relax and remove_constraints. Neither replaces the debugging: an IIS. A built row is read with row, in linopy's own form.

Where linopy is ahead, and why none of it is a ceiling question, is the roadmap. What is owed to linopy rather than merely true of it is prior art and credit. The same page credits Calliope, whose math language this surface is derived from.