Drop lint ignores except in demos (#266)

demos/burgers-goal_oriented.py

@@ -1,11 +1,12 @@
 # Burgers equation with Goal-oriented mesh adaptation
 # ===================================================
 
-# In the `Hessian-based adaptation  <./burgers-hessian.py.html>`__, we applied a Hessian-based
-# mesh adaptation to the time-dependent Burgers equation. Here, we alternatively consider
-# a goal-oriented error estimation to drive the mesh adaptation. Again, we will
-# choose to partition the problem over multiple subintervals and hence multiple meshes
-# to adapt. We also chose to apply a QoI which integrates in time as well as space.
+# In the `Hessian-based adaptation  <./burgers-hessian.py.html>`__, we applied a
+# Hessian-based mesh adaptation to the time-dependent Burgers equation. Here, we
+# alternatively consider a goal-oriented error estimation to drive the mesh adaptation.
+# Again, we will choose to partition the problem over multiple subintervals and hence
+# multiple meshes to adapt. We also chose to apply a QoI which integrates in time as
+# well as space.
 #
 # We copy over the setup as before. The only difference is that we import from
 # ``goalie_adjoint`` rather than ``goalie``. ::
@@ -115,8 +116,8 @@ def time_integrated_qoi(t):
 
 # Compared to the previous `Hessian-based adaptation <./burgers-hessian.py.html>`__,
 # this adaptor depends on adjoint solution data as well as forward solution data.
-# For simplicity, we begin by using :meth:`~.RiemannianMetric.compute_isotropic_metric()`.
-# ::
+# For simplicity, we begin by using
+# :meth:`~.RiemannianMetric.compute_isotropic_metric()`. ::
 
 
 def adaptor(mesh_seq, solutions=None, indicators=None):
@@ -257,7 +258,8 @@ def adaptor(mesh_seq, solutions=None, indicators=None):
 # (See documentation for details.) To use it, we just need to define
 # a different adaptor function. The same error indicator is used as
 # for the isotropic approach. Additionally, the Hessian of the forward
-# solution is estimated as in the `Hessian-based adaptation <./burgers-hessian.py.html>`__
+# solution is estimated as in the
+# `Hessian-based adaptation <./burgers-hessian.py.html>`__
 # to give anisotropy to the metric.
 #
 # For this driver, normalisation is handled differently than for
@@ -351,8 +353,8 @@ def adaptor(mesh_seq, solutions=None, indicators=None):
     fig.savefig(f"burgers-anisotropic_mesh{iteration + 1}.jpg")
     plt.close()
 
-    # Since we have multiple subintervals, we should check if the target complexity has been
-    # (approximately) reached on each subinterval
+    # Since we have multiple subintervals, we should check if the target complexity has
+    # been (approximately) reached on each subinterval
     continue_unconditionally = np.array(complexities) < 0.90 * target
     return continue_unconditionally
 

demos/burgers-hessian.py

@@ -13,6 +13,7 @@
 from animate.adapt import adapt
 from animate.metric import RiemannianMetric
 from firedrake import *
+from firedrake.__future__ import interpolate
 
 from goalie import *
 

demos/burgers_ee.py

@@ -122,14 +122,15 @@ def end_time_qoi():
     qoi_type="end_time",
 )
 
-# Given the description of the PDE problem in the form of a :class:`GoalOrientedMeshSeq`,
-# Goalie is able to extract all of the relevant information to automatically compute
-# error estimators. During the computation, we solve the forward and adjoint equations
-# over the mesh sequence, as before. In addition, we solve the adjoint problem again
-# in an *enriched* finite element space. Currently, Goalie supports uniform refinement
-# of the meshes (:math:`h`-refinement) or globally increasing the polynomial order
-# (:math:`p`-refinement). Choosing one (or both) of these as the ``"enrichment_method"``,
-# we are able to compute error indicator fields as follows. ::
+# Given the description of the PDE problem in the form of a
+# :class:`GoalOrientedMeshSeq`, Goalie is able to extract all of the relevant
+# information to automatically compute error estimators. During the computation, we
+# solve the forward and adjoint equations over the mesh sequence, as before. In
+# addition, we solve the adjoint problem again in an *enriched* finite element space.
+# Currently, Goalie supports uniform refinement of the meshes (:math:`h`-refinement) or
+# globally increasing the polynomial order (:math:`p`-refinement). Choosing one (or
+# both) of these as the ``"enrichment_method"``, we are able to compute error indicator
+# fields as follows. ::
 
 solutions, indicators = mesh_seq.indicate_errors(
     enrichment_kwargs={"enrichment_method": "h"}

demos/burgers_oo.py

@@ -110,7 +110,8 @@ def time_integrated_qoi(t):
     enrichment_kwargs={"enrichment_method": "h"}
 )
 
-# Plotting this, we find that the results are consistent with those generated previously. ::
+# Plotting this, we find that the results are consistent with those generated
+# previously. ::
 
 fig, axes, tcs = plot_indicator_snapshots(indicators, time_partition, "u", levels=50)
 fig.savefig("burgers-oo_ee.jpg")

demos/burgers_time_integrated.py

@@ -1,19 +1,20 @@
 # Adjoint Burgers equation with a time integrated QoI
 # ======================================================
 #
-# So far, we only considered a quantity of interest
-# corresponding to a spatial integral at the end time.
-# For some problems, it is more suitable to have a QoI
-# which integrates in time as well as space.
+# So far, we only considered a quantity of interest corresponding to a spatial integral
+# at the end time. For some problems, it is more suitable to have a QoI which integrates
+# in time as well as space.
 #
-# Begin by importing from Goalie and the first Burgers demo. ::
+# Begin by importing from Firedrake and Goalie. Note that we use the *future* version
+# Firedrake's `interpolate` function: :function:`firedrake.__future__.interpolate`. ::
 
 from firedrake import *
+from firedrake.__future__ import interpolate
 
 from goalie_adjoint import *
 
-# Redefine the ``get_initial_condition`` and ``get_function_spaces``,
-# functions as in the first Burgers demo. ::
+# Redefine the ``get_initial_condition`` and ``get_function_spaces``, functions as in
+# the first Burgers demo. ::
 
 
 def get_function_spaces(mesh):
@@ -26,14 +27,12 @@ def get_initial_condition(mesh_seq):
     return {"u": assemble(interpolate(as_vector([sin(pi * x), 0]), fs))}
 
 
-# The solver needs to be modified slightly in order to take
-# account of time dependent QoIs. The Burgers solver
-# uses backward Euler timestepping. The corresponding
-# quadrature routine is like the midpoint rule, but takes
-# the value from the next timestep, rather than the average
-# between that and the current value. As such, the QoI may
-# be computed by simply incrementing the :attr:`J` attribute
-# of the :class:`AdjointMeshSeq` as follows. ::
+# The solver needs to be modified slightly in order to take account of time dependent
+# QoIs. The Burgers solver uses backward Euler timestepping. The corresponding
+# quadrature routine is like the midpoint rule, but takes the value from the next
+# timestep, rather than the average between that and the current value. As such, the QoI
+# may be computed by simply incrementing the :attr:`J` attribute of the
+# :class:`AdjointMeshSeq` as follows. ::
 
 
 def get_solver(mesh_seq):
@@ -69,17 +68,15 @@ def solver(index):
     return solver
 
 
-# The QoI is effectively just a time-integrated version
-# of the one previously seen.
+# The QoI is effectively just a time-integrated version of the one previously seen.
 #
 # .. math::
 #    J(u) = \int_0^{T_{\mathrm{end}}} \int_0^1
 #    \mathbf u(1,y,t) \cdot \mathbf u(1,y,t)
 #    \;\mathrm dy\;\mathrm dt.
 #
-# Note that in this case we multiply by the timestep.
-# It is wrapped in a :class:`Function` from `'R'` space to avoid
-# recompilation if the value is changed. ::
+# Note that in this case we multiply by the timestep. It is wrapped in a
+# :class:`Function` from `'R'` space to avoid recompilation if the value is changed. ::
 
 
 def get_qoi(mesh_seq, i):
@@ -106,9 +103,8 @@ def time_integrated_qoi(t):
     end_time, num_subintervals, dt, ["u"], num_timesteps_per_export=1
 )
 
-# The only difference when defining the :class:`AdjointMeshSeq`
-# is that we specify ``qoi_type="time_integrated"``, rather than
-# ``qoi_type="end_time"``. ::
+# The only difference when defining the :class:`AdjointMeshSeq` is that we specify
+# ``qoi_type="time_integrated"``, rather than ``qoi_type="end_time"``. ::
 
 mesh_seq = AdjointMeshSeq(
     time_partition,
@@ -128,14 +124,12 @@ def time_integrated_qoi(t):
 #    :figwidth: 90%
 #    :align: center
 #
-# With a time-integrated QoI, the adjoint problem
-# has a source term at the right-hand boundary, rather
-# than a instantaneous pulse at the terminal time. As such,
-# the adjoint solution field accumulates at the right-hand
-# boundary, as well as propagating westwards.
+# With a time-integrated QoI, the adjoint problem has a source term at the right-hand
+# boundary, rather than a instantaneous pulse at the terminal time. As such, the adjoint
+# solution field accumulates at the right-hand boundary, as well as propagating
+# westwards.
 #
-# In the `next demo <./burgers_oo.py.html>`__, we solve
-# the Burgers problem one last time, but using an
-# object-oriented approach.
+# In the `next demo <./burgers_oo.py.html>`__, we solve the Burgers problem one last
+# time, but using an object-oriented approach.
 #
 # This demo can also be accessed as a `Python script <burgers_time_integrated.py>`__.

demos/mantle_convection.py

@@ -5,8 +5,8 @@
 # involve a time derivative. Those are clearly *time-dependent* equations. However,
 # time-dependent equations need not involve a time derivative. For example, they might
 # include fields that vary with time. Examples of where this might happen are
-# in continuous pressure projection approaches, ice sheet and glaciological modelling, and
-# mantle convection modelling. In this demo, we illustrate how Goalie can be used
+# in continuous pressure projection approaches, ice sheet and glaciological modelling,
+# and mantle convection modelling. In this demo, we illustrate how Goalie can be used
 # to solve such problems.
 
 # We consider the problem of a mantle convection in a 2D unit square domain. The
@@ -19,8 +19,10 @@
 #
 # .. math::
 #    \begin{align}
-#        \nabla \cdot \mu \left[\nabla \mathbf{u} + (\nabla \mathbf{u})^T \right] - \nabla p + \mathrm{Ra}\,T\,\mathbf{k} &= 0, \\
-#        \frac{\partial T}{\partial t} \cdot \mathbf{u}\cdot\nabla T - \nabla \cdot (\kappa\nabla T) &= 0,
+#        \nabla \cdot \mu \left[\nabla \mathbf{u} + (\nabla \mathbf{u})^T \right] -
+#           \nabla p + \mathrm{Ra}\,T\,\mathbf{k} &= 0, \\
+#        \frac{\partial T}{\partial t} \cdot \mathbf{u}\cdot\nabla T
+#           - \nabla \cdot (\kappa\nabla T) &= 0,
 #    \end{align}
 #
 # where :math:`\mu`, :math:`\kappa`, and :math:`\mathrm{Ra}` are constant viscosity,

demos/ode.py

@@ -154,10 +154,10 @@ def solver(index):
 
 solutions = point_seq.solve_forward()["u"]["forward"]
 
-# Note that the solution trajectory does not include the initial value, so we prepend it.
-# We also convert the solution :class:`~.Function`\s to :class:`~.float`\s, for plotting
-# purposes. Whilst there is only one subinterval in this example, we show how to loop
-# over subintervals, as this is instructive for the general case. ::
+# Note that the solution trajectory does not include the initial value, so we prepend
+# it. We also convert the solution :class:`~.Function`\s to :class:`~.float`\s, for
+# plotting purposes. Whilst there is only one subinterval in this example, we show how
+# to loop over subintervals, as this is instructive for the general case. ::
 
 forward_euler_trajectory = [1]
 forward_euler_trajectory += [
@@ -257,7 +257,8 @@ def solver(index):
 # .. math::
 #    \frac{u_{i+1} - u_i}{\Delta t} = (\theta u_{i+1} + (1-\theta) u_i),
 #
-# where :math:`\theta\in(0,1)`. The standard choice is to take :math:`\theta=\frac12`. ::
+# where :math:`\theta\in(0,1)`. The standard choice is to take :math:`\theta=\frac12`.
+# ::
 
 
 def get_solver_crank_nicolson(point_seq):

demos/point_discharge2d-goal_oriented.py

@@ -242,11 +242,11 @@ def adaptor(mesh_seq, solutions, indicators):
 #
 # Looking at the final adapted mesh, we can make a few observations. Firstly, the mesh
 # elements are indeed isotropic. Secondly, there is clearly increased resolution
-# surrounding the point source, as well as the "receiver region" which the QoI integrates
-# over. There is also a band of increased resolution between these two regions. Finally,
-# the mesh has low resolution downstream of the receiver region. This is to be expected
-# because we have an advection-dominated problem, so the QoI value is independent of the
-# dynamics there.
+# surrounding the point source, as well as the "receiver region" which the QoI
+# integrates over. There is also a band of increased resolution between these two
+# regions. Finally, the mesh has low resolution downstream of the receiver region. This
+# is to be expected because we have an advection-dominated problem, so the QoI value is
+# independent of the dynamics there.
 #
 # Goalie also provides drivers for *anisotropic* goal-oriented mesh adaptation. Here,
 # we consider the ``anisotropic_dwr_metric`` driver. (See documentation for details.) To
@@ -372,4 +372,5 @@ def adaptor(mesh_seq, solutions, indicators):
 # In the `next demo <./burgers-hessian.py.html>`__, we consider mesh adaptation in the
 # time-dependent case.
 #
-# This demo can also be accessed as a `Python script <point_discharge2d-goal_oriented.py>`__.
+# This demo can also be accessed as a
+# `Python script <point_discharge2d-goal_oriented.py>`__.

demos/point_discharge2d-hessian.py

@@ -10,7 +10,7 @@
 # before progressing with this demo.
 #
 # In addition to importing from Firedrake and Goalie, we also import the mesh
-# adaptation functionality from Firedrake, which can be found in ``firedrake.meshadapt``.
+# adaptation functionality from Animate, which can be found in ``animate.adapt``.
 # ::
 
 from animate.adapt import adapt
@@ -20,8 +20,8 @@
 from goalie import *
 
 # We again consider the "point discharge with diffusion" test case from the
-# `previous demo <./point_discharge2d.py.html>`__, approximating the tracer concentration
-# :math:`c` in :math:`\mathbb P1` space. ::
+# `previous demo <./point_discharge2d.py.html>`__, approximating the tracer
+# concentration :math:`c` in :math:`\mathbb P1` space. ::
 
 field_names = ["c"]
 
@@ -187,8 +187,8 @@ def adaptor(mesh_seq, solutions):
 )
 solutions = mesh_seq.fixed_point_iteration(adaptor, parameters=params)
 
-# Mesh adaptation often gives slightly different results on difference machines. However,
-# the output should look something like
+# Mesh adaptation often gives slightly different results on difference machines.
+# However, the output should look something like
 #
 # .. code-block:: console
 #

demos/point_discharge2d.py

@@ -15,7 +15,8 @@
 #   \left\{\begin{array}{rl}
 #       \mathbf u\cdot\nabla c - \nabla\cdot(D\nabla c) = S & \text{in}\:\Omega\\
 #       c=0 & \text{on}\:\partial\Omega_{\mathrm{inflow}}\\
-#       \nabla c\cdot\widehat{\mathbf n}=0 & \text{on}\:\partial\Omega\backslash\partial\Omega_{\mathrm{inflow}}
+#       \nabla c\cdot\widehat{\mathbf n}=0 &
+#           \text{on}\:\partial\Omega\backslash\partial\Omega_{\mathrm{inflow}}
 #   \end{array}\right.,
 #
 # for a tracer concentration :math:`c`, with fluid velocity
@@ -73,7 +74,8 @@ def source(mesh):
 # With these ingredients, we can now define the :meth:`get_solver` method. Don't forget
 # to apply the corresponding `ad_block_tag` to the solve call. Additionally, we must
 # communicate the defined variational form to ``mesh_seq`` using the
-# :meth:`mesh_seq.read_form()` method for Goalie to utilise it during error indication. ::
+# :meth:`mesh_seq.read_form()` method for Goalie to utilise it during error indication.
+# ::
 
 
 def get_solver(mesh_seq):

goalie/__init__.py

@@ -10,11 +10,4 @@
 from goalie.function_data import *  # noqa
 from goalie.error_estimation import *  # noqa
 
-from animate.utility import Mesh, VTKFile  # noqa
-
-from firedrake.__future__ import interpolate  # noqa
-
-import numpy as np  # noqa
-import os  # noqa
-
 __version__ = "0.1"