Operations and output queries
FEA expressions resolve through RunMat's canonical operation catalog. One public operation name
may have numeric .m overloads and FEA field overloads, but both use one identity, documentation
entry, completion item, and lowering registry.
Field operations
magnitude converts a typed vector field to a scalar field on the same domain. Complex fields
require an explicit representation such as magnitude, phase, real component, or imaginary
component before an operation that assumes real ordering or sign.
component selects a typed component. recover performs an explicit supported element-to-node
recovery and records the recovery method. Neither operation changes the underlying physical
quantity.
Spatial reductions
min and max over a region return attributed extrema. sum, mean, integral, resultant, norm,
and moment operations enforce field location, physical quantity, geometric measure, weighting, and
reference-position rules. A boundary integral and a nodal sum are different overloads with
different evidence.
Named argument over selects the typed reduction domain. Moment operations require an explicit
reference point. Vector and tensor extrema are rejected until reduced to an ordered scalar
representation.
Two port operations read a pair of face regions on one step. pressure_drop(step, from = inlet, to = outlet) returns the area-weighted mean fluid pressure on from minus the mean on
to, in Pa; it requires a fluid step. transmission_loss(step, incident = inlet, transmitted = outlet) returns in dB at each frequency of an
acoustic harmonic step, where is the incident power at the incident port and
the power leaving through the transmitted port. The incident region must carry
an incident wave and the transmitted region a radiation (anechoic) boundary; a run with no
positive power at either port fails with that reason. Both operations refuse a region that is not
a face region.
output pressure_drop = pressure_drop(analysis.flow, from = geometry.channel.region.inlet, to = geometry.channel.region.outlet)In this example the channel is 3 m long and its walls are 0.1 m apart (L/h = 30). The fully developed drop is Pa. The run reads within 5% of that value. The uniform inlet adds an entrance loss of about 0.9 h of equivalent channel length, which is about 3% of the developed drop.
Sampling and series
sample selects explicit positions on a field's independent domain. Interpolation requires one
point or a nonempty typed path. History and series reductions preserve sample coordinates,
weighting, event attribution, and source identity. Output policy blocks bound retained samples and
storage.
Sweep operations
collect(case.output.<name>) retains compact case values. retain_each keeps each referenced
output object. Case min and max retain the governing case coordinates. Sweep operations accept
only outputs declared by the source study.
Acceptance assessments
An assessment inside a study compares a declared real scalar output with a
dimensionally compatible bound. Use maximum, minimum, or both for an inclusive
range. The evidence policy is required.
output spatial_peak = max(
magnitude(analysis.response.field.node.displacement),
over = geometry.bracket.region.body,
)
output peak_history = sample(
output.spatial_peak,
over = time,
at = [0 s, 0.05 s, 0.1 s],
)
output peak = max(output.peak_history, over = time)
assessment deflection {
value = output.peak
maximum = 1 mm
evidence = authored_samples
}This example checks the three declared times. Choose sampling coordinates that cover the engineering requirement; three samples alone do not establish the peak throughout a transient. Spatial reduction occurs before sampling, and the time reduction preserves the governing sample and entity.
authored_samples compares the declared output when its numerical result is
acceptable. certified_bound additionally requires a certified absolute error
bound. Unavailable evidence or uncertainty crossing the limit produces an
indeterminate result. Acceptance status is separate from solver convergence.
Study setup should preserve a user's criterion while checking output coverage and available execution evidence. A final-time output cannot answer a peak-over-time requirement. Do not replace such a requirement with a final-time comparison or claim continuous coverage from a sampled maximum.
Imported assessment functions
Use an imported function when a comparison needs custom logic. The function receives
the scalar output first, followed by the declared parameters. Values are passed in
canonical SI units: 1 mm is passed as 0.001, for example. A function assessment
cannot also declare minimum or maximum.
import function within_limit from "./within_limit.m"
// Inside a study, after declaring output.peak:
assessment deflection {
value = output.peak
function = function.within_limit
parameters = [1 mm]
evidence = authored_samples
}function result = within_limit(value, limit)
arguments
value (1,1) double
limit (1,1) double
end
result = struct('evaluated', true, 'accepted', value <= limit);
endThe checked function must have a complete signature with real scalar inputs and a
scalar record containing exactly two logical fields, evaluated and accepted.
It cannot capture workspace values or require ambient capabilities such as file
access or randomness. The executable must match the frozen imported source.
evaluated = false produces an indeterminate verdict. A thrown exception or an
invalid return also produces an indeterminate verdict, with an execution diagnostic.
A function that has not run is unavailable, not failed. The logical return does not
certify an error bound or establish coverage between authored samples. Checking the
source establishes its language contract; execution still requires a host that
admits the portable function and its bounded task resources.
Reassessing saved results
To compare existing results against revised criteria, save the revised assessment declarations and reassess the original result publication. This produces a separate immutable assessment report. It does not solve again, replace the original report, or apply revised geometry, materials or loads to the old numerical result.
runmat reassess bracket.fea --study bracket \
--original-definition sha256:<original-definition-digest> \
--result-manifest <saved-study-result-manifest> --jsonUse the original definition digest and content-addressed manifest recorded by the solve. Changed bounds or compatible custom assessment functions can reuse that publication. Each referenced output must retain its original quantity, region, sampling and producer. Changing an output while keeping its name requires new numerical evidence. Imported functions still require checked frozen sources and bounded executable admission. Missing evidence remains indeterminate.
Diagnostics
Operation errors report the resolved argument types, physical dimensions, locations, available overloads, and invalid named argument. There is no fallback to a string-dispatched solver helper. An operation is not part of the language unless its catalog descriptor, type checker, lowering, editor metadata, and documentation are all present.
Examples
Field operations compose inside an output expression while preserving quantity and location:
output displacement = analysis.solve.field.node.displacement
output displacement_magnitude = magnitude(output.displacement)
output peak_displacement = max(
output.displacement_magnitude,
over = geometry.bracket.region.body,
)Use a block when an output also needs retention policy:
output tip_history {
value = sample(
source = output.tip_displacement,
over = time,
at = [0 s, 0.5 s, 1 s],
)
precision = real64
compression = lossless
retention = explicit_samples
maximum_samples = 2000
partial_access = manifest_and_chunks
purpose = publication
}Sweep expressions consume outputs declared by the source study:
sweep thickness_study over study.static_check {
axis wall_thickness {
thin = 2 mm
nominal = 3 mm
thick = 4 mm
}
output peak_by_case = collect case.output.peak_stress.value
output governing_case = max(case.output.peak_stress.value, over = cases)
output stress_fields = retain_each case.output.stress
}