tests/testthat/helper-methods.R

In rxode2: Facilities for Simulating from ODE-Based Models

.rxode2 <- function(...) {
  suppressMessages(rxode2::rxode2(...))
}

# Helper functions for the test cases of the methods.
# general ode methods used
# euler excluded: 1st-order accuracy requires hmin << 0.01 to pass covariate tests
# indLin excluded for now
.methods0 <- c("liblsoda", "lsoda", "dop853", "f78", "rk4", "ck54", "ab", "abm", "dop5", "bs", "ros4", "iem", "sem", "sb3a", "sb3am4", "mm", "em", "cvode", "trapz", "ssp3", "f32", "rk43", "dop54", "vern65", "vern76", "dop87", "vern98", "ros43", "ros6", "backwardEuler", "gauss6", "iiic6", "radauiia5", "geng5", "sdirk43", "euler", "midpoint", "heun", "ssp22", "rk3", "ssp53", "s4", "r4", "ls44", "ls54", "ssp54", "s5", "rk5", "c5", "l5", "lk5a", "lk5b", "b6", "s7", "s8_10", "cv8", "s8_12", "s10", "z10", "o10", "h10", "dp54", "v65e", "v76e", "dp87", "v98e", "ssp33", "bs32", "ssp43", "f45", "t54", "s54", "pp54", "pp54b", "bs54", "ss54", "dp65", "c65", "tp64", "v65r", "v65", "dverk65", "tf65", "tp75", "tmy7", "tmy7s", "v76r", "ss76", "v78", "dverk78", "dp85", "tp86", "v87e", "v87r", "ev87", "k87", "f89", "v89", "t98a", "v98r", "s98", "f108", "c108", "b109", "s1110a", "f1210", "o129", "f1412", "lsode", "bdf", "t54+sdirk43")
# AutoSwitch composite methods: dop853 primary paired with each stiff
# secondary that needs a Jacobian.  1st-order secondaries (iem,
# backwardEuler) are intentionally excluded -- when dop853 falls back to
# them they cannot meet the accuracy tolerances, which defeats the purpose
# of a high-order primary.  This set exercises the non-dense composite
# autoswitch dispatch for every higher-order stiff secondary.
.methods0 <- c("dop853+ros4", "dop853+ros43", "dop853+ros6",
               "dop853+gauss6", "dop853+iiic6",
               "dop853+radauiia5", "dop853+geng5", "dop853+sdirk43")

.methods0 <- c("lsode", "bdf", "liblsoda", "dop853", "t54+sdirk43", "cvode")
.methods1 <- c("ddop853", "ddop5", "dbs", "dros4", "ddop853+dros4", "dcvode")
# All dense composite methods (ros4 is the only dense-output stiff secondary).
.methods1 <- c("ddop853+dros4")
.methods2 <- c("A", "B", "Ao", "Bo", "As", "Bs", "Ad", "Bd", "Al", "Bl")

# ---------------------------------------------------------------------------
# Discrete-adjoint solver variants (method codes 200-304).  On a PLAIN model
# (no rx__sens_* expansion) the explicit/multistep variants run FORWARD-ONLY
# -- so pointing the covariate (test-cov) and dosing (test-nmtest) regression
# suites at them exercises exactly the covariate-interpolation + event/dosing
# handling of every adjoint forward path.  This has historically been a
# regression miss.  The implicit/Rosenbrock/CVODES adjoint variants (ros4s,
# ros43s, ros6s, backwardEulers, gauss6s, iiic6s, radauiia5s, geng5s,
# sdirk43s, cvodesadj) size their LU factorization from the adjoint-augmented
# system and cannot run forward-only, so they are intentionally omitted here
# (their forward IS the well-tested base implicit method; the adjoint sweep is
# validated on the expanded models in test-adjoint-discrete.R).
# .methods1/.methods2 are cleared so nmtest cycles over .methods0 alone.
.methods0 <- c("dop853s", "liblsodaadj", "f78s", "rk4s", "ck54s", "abs",
               "dop5s", "rk43s", "dop54s", "vern65s", "vern76s", "dop87s",
               "vern98s", "eulers", "midpoints", "heuns", "rk3s", "bs32s",
               "f45s", "t54s", "pp54s", "pp54bs", "bs54s", "ss54s", "dp65s",
               "c65s", "tp64s", "v65rs", "dverk65s", "tf65s", "tp75s",
               "tmy7adj", "tmy7sadj", "v76rs", "ss76s", "v78s", "dverk78s",
               "dp85s", "tp86s", "v87es", "v87rs", "ev87s", "k87s", "v89s",
               "t98as", "v98rs", "s98s", "c108s", "b109s", "s1110as", "o129s")
.methods1 <- NULL
.methods2 <- NULL

# test-cov validates the (method-independent) covariate-interpolation logic by
# exact-matching each integrator's RHS-evaluation trace against an approxfun
# reference.  That comparison is tied to the integrator's step schedule, so it
# can only run on the base integrators it was tuned for -- the adjoint variants
# record extra internal steps and would produce spurious knot-boundary
# mismatches (their covariate handling is the same shared code and is verified
# via solution-parity + the nmtest dosing sweep above).
.methodsCov <- c("lsode", "bdf", "liblsoda", "dop853", "t54+sdirk43", "cvode")