pyextal.jaxJoint.engine

Keeping the Fortran engine on the dataset being worked on.

bloch_parse loads one .dat into the engine’s module state, and several things the refinement needs read that state rather than anything passed to them: get_bloch_constants, the symmetry operations (spgr, used by GroupSymmetry and appliedSymmetry), lookupSF/defaultsf, and beamcheck – which calculate_tilt_dependent_values calls per tilt, at setup and at every geometry rebuild. With several datasets the engine must hold the right one whenever any of these runs, so every such step goes through EngineSwitch.activate.

Re-parsing is safe: the engine’s allocators release the previous arrays first. terminate() is not: it ends the engine for the life of the process, so it must never run between datasets. jaxJoint never calls bloch_run, so there is nothing to terminate at the end either.

Attributes

MAX_DAT_PATH

Classes

EngineSwitch

Parses each dataset's .dat into the engine on demand.

Module Contents

pyextal.jaxJoint.engine.MAX_DAT_PATH = 120

class pyextal.jaxJoint.engine.EngineSwitch

Parses each dataset’s .dat into the engine on demand.

Datasets are added in order; activate re-parses only when the engine holds a different .dat, so two datasets built against the same file never re-parse between them.

_dats = []

_params

_beams

_loaded = None

_tmp = None

parses = 0

add(dat)

Registers a dataset’s .dat and returns its index.

_short(dat)

record(d, constants)

Remembers the beam list dataset d’s setup was built with.

activate(d)

Makes the engine hold dataset d’s .dat.

Parameters:

d (int) – Dataset index.

Returns:

From that dataset’s bloch_parse.

Return type:

SimParams

Raises:

RuntimeError – If the re-parsed beam list differs from the one the dataset was set up with.

_check(d)