Use vulture to detect dead code and remove what the command found (#788)

docs/examples_solver_config/conditioning-on-zero-residual.py

@@ -132,11 +132,6 @@ def vector_field(y, t):  # noqa: ARG001
 }
 
 
-def log_residual(*args):
-    """Evaluate the log-ODE-residual."""
-    return jnp.log10(jnp.abs(residual(*args)))
-
-
 def residual(x, t):
     """Evaluate the ODE residual."""
     return x[1] - jax.vmap(jax.vmap(vector_field), in_axes=(0, None))(x[0], t)

makefile

@@ -47,3 +47,6 @@ doc:
 	make example
 	make benchmarks-plot-results
 	JUPYTER_PLATFORM_DIRS=1 mkdocs build
+
+find-dead-code:
+	vulture . --ignore-names case*,fixture*,*jvp --exclude probdiffeq/_version.py

probdiffeq/backend/numpy.py

@@ -44,10 +44,6 @@ def diff(arr, /):
     return jnp.diff(arr)
 
 
-def diff_along_axis(arr, /, *, axis):
-    return jnp.diff(arr, axis=axis)
-
-
 def reshape(arr, /, new_shape, order="C"):
     return jnp.reshape(arr, new_shape, order=order)
 

probdiffeq/backend/ode.py

@@ -49,23 +49,6 @@ def solution(t):
     return solution
 
 
-def ivp_logistic():
-    # Local imports because diffeqzoo is not an official dependency
-    from diffeqzoo import backend, ivps
-
-    if not backend.has_been_selected:
-        backend.select("jax")
-
-    f, u0, (t0, _), f_args = ivps.logistic()
-    t1 = 0.75
-
-    @jax.jit
-    def vf(x, *, t):  # noqa: ARG001
-        return f(x, *f_args)
-
-    return vf, (u0,), (t0, t1)
-
-
 def ivp_lotka_volterra():
     # Local imports because diffeqzoo is not an official dependency
     from diffeqzoo import backend, ivps
@@ -83,34 +66,6 @@ def vf(x, *, t):  # noqa: ARG001
     return vf, (u0,), (t0, t1)
 
 
-def ivp_affine_multi_dimensional():
-    t0, t1 = 0.0, 2.0
-    u0 = jnp.ones((2,))
-
-    @jax.jit
-    def vf(x, *, t):  # noqa: ARG001
-        return 2 * x
-
-    def solution(t):
-        return jnp.exp(2 * t) * jnp.ones((2,))
-
-    return vf, (u0,), (t0, t1), solution
-
-
-def ivp_affine_scalar():
-    t0, t1 = 0.0, 2.0
-    u0 = 1.0
-
-    @jax.jit
-    def vf(x, *, t):  # noqa: ARG001
-        return 2 * x
-
-    def solution(t):
-        return jnp.exp(2 * t)
-
-    return vf, (u0,), (t0, t1), solution
-
-
 def ivp_three_body_1st():
     # Local imports because diffeqzoo is not an official dependency
     from diffeqzoo import backend, ivps

probdiffeq/util/filter_util.py

@@ -7,8 +7,6 @@
 from probdiffeq.backend.typing import Any
 
 
-# TODO: fixedpointsmoother and kalmanfilter should be estimate()
-#  with two different methods. This saves a lot of code.
 def estimate_fwd(data, /, init, prior_transitions, observation_model, *, estimator):
     """Estimate forward-in-time."""
     initialise, step = estimator
@@ -91,7 +89,7 @@ def _initialise(rv, data, model) -> _KFState:
         observed, conditional = ssm.conditional.revert(rv, model)
         corrected = ssm.conditional.apply(data, conditional)
         logpdf = ssm.stats.logpdf(data, observed)
-        return _KFState(corrected, 1.0, logpdf)
+        return _KFState(corrected, num_data_points=0.0, logpdf=logpdf)
 
     def _step(state: _KFState, cond_and_data_and_obs) -> tuple[_KFState, _KFState]:
         conditional, data, observation = cond_and_data_and_obs
@@ -105,7 +103,7 @@ def _step(state: _KFState, cond_and_data_and_obs) -> tuple[_KFState, _KFState]:
         # Update logpdf
         logpdf_new = ssm.stats.logpdf(data, observed)
         logpdf_mean = (logpdf * num_data + logpdf_new) / (num_data + 1)
-        state = _KFState(corrected, num_data + 1.0, logpdf_mean)
+        state = _KFState(corrected, num_data_points=num_data + 1.0, logpdf=logpdf_mean)
 
         # Scan-compatible output
         return state, state