Reactor

Reactor loads a reactor.yaml specification and orchestrates solving.

Core fields - path, id, name, organization, country, year, doi, notes - tags: list of strings used for relation filtering - solving_order: optional ordered list of domains/relations to solve in sequence - solver_mode: "overwrite" or "check" - verbose: bool - relations: list of Relation - default_relations: list of Relation (defaults applied at load) - variables_dict: dict of Variable objects keyed by name

Example

from fusdb import Reactor, Variable

reactor = Reactor(
    id="EXAMPLE",
    name="Example Reactor",
    tags=["tokamak", "hmode"],
    variables_dict={
        "R": Variable(name="R", values=[3.2], unit="m", input_source="explicit"),
        "a": Variable(name="a", values=[1.1], unit="m", input_source="explicit"),
    },
)

Methods - __post_init__() initializes the per-instance logger context. - from_yaml(path) loads YAML, parses variables, applies defaults, filters relations. - _relation_filter_inputs() returns variable names and method overrides for filtering. - _ordered_relations() yields relations in solving order (domains or names). - solve(mode=None, verbose=None) runs one or more RelationSystem passes and updates variables_dict. - diagnose() returns {"violated_relations": [...], "likely_culprits": {...}, "variable_issues": [...]} using "check" mode. - popcon(...) evaluates POPCON-style scans over one or more axes (grid or point-solve). - plot_popcon(...) plots masked fills + contour overlays from POPCON results. - __repr__() returns a compact summary string. - plot_cross_sections() plots the plasma cross-section using R, a, kappa_95, delta_95.

Profiles - Profile-capable variables accept a dict payload with coord, x, y, optional meta. - n_e, n_i, T_e, and T_i are profile-valued; scalar inputs are treated as flat profiles. - If n_avg/T_avg are provided without profiles, a flat profile on r_minor is created by defaults. - If profiles are provided without averages, n_avg/T_avg are computed as simple means.

Reactor YAML (example)

metadata:
  id: REACTOR_ID        # required
  name: Reactor Name    # optional, defaults to id
  organization: Org     # optional
  country: Country      # optional
  year: 2025            # optional
  doi: 10.0000/example  # optional
  notes: Optional notes # optional

tags: # used to filter relations.
  - device_tag
  - regime_tag

solver_tags:
  mode: overwrite   # overwrite | check
  verbosity: false  # optional
  solving_order: # optional, controls the order of `RelationSystem` runs and may warn on overlapping outputs. Each entry can be a domain tag or an exact `Relation.name`.
    - domain_a
    - domain_b
    - "Specific Relation Name"

variables:
  R: 3.2            # shorthand value (unit defaults to registry)
  a:
    value: 1.1      # explicit value
    unit: m         # optional unit (converted to registry default using pint)
    method: relation_name # optional, selects a specific relation when multiple are available.
    rel_tol: 0.02   # optional, tolerance override
    abs_tol: 0.0    # optional, tolerance override
    fixed: false    # optional, prevents solver changes (useful for certain values such as geometry)
  n_e:
    value:
      coord: r_minor
      x: [0.0, 0.5, 1.0]
      y: [1.0e20, 0.9e20, 0.7e20]