ENH: Improve importing optional packages and modules

Improve importing optional packages and modules: - Add helper classes and utils to take care of trying to import optional packages, and to provide meaningful information when such packages cannot be imported. - Add the corresponding tests. Prefer using all lowercases for the flags that contain the package/module import success value. Move the `whitematteranalysis/utils.py` module to the `whitematteranalysis/utils/utils.py` module, and remove the `utils` module import from `whitematteranalysis/__init__.py` to avoid module clashes.
SlicerDMRI · Oct 21, 2023 · cf4a2ae · cf4a2ae
1 parent 4c93d41
commit cf4a2ae
Show file tree

Hide file tree

Showing 16 changed files with 256 additions and 72 deletions.
diff --git a/bin/wm_assess_cluster_location_by_hemisphere.py b/bin/wm_assess_cluster_location_by_hemisphere.py
@@ -21,6 +21,7 @@
 import vtk
 
 import whitematteranalysis as wma
+from whitematteranalysis.utils.utils import hemisphere_loc_name_typo_warn_msg
 
 
 def _build_arg_parser():
@@ -79,7 +80,7 @@ def write_mask_location_to_vtk(inpd, mask_location):
                 array = inpointdata.GetArray(idx)
                 if array.GetName() == 'HemisphereLocataion':
                     warnings.warn(
-                        wma.utils.hemisphere_loc_name_typo_warn_msg,
+                        hemisphere_loc_name_typo_warn_msg,
                         PendingDeprecationWarning)
                     print('  -- HemisphereLocataion is in the input data: skip updating the vtk file.')
                     return inpd
@@ -126,7 +127,7 @@ def _read_location(_inpd):
                 array = inpointdata.GetArray(idx)
                 if array.GetName() == 'HemisphereLocataion':
                     warnings.warn(
-                        wma.utils.hemisphere_loc_name_typo_warn_msg,
+                        hemisphere_loc_name_typo_warn_msg,
                         PendingDeprecationWarning)
                     flag_location = _read_location(inpd)
                     break

diff --git a/bin/wm_quality_control_after_clustering.py b/bin/wm_quality_control_after_clustering.py
@@ -4,21 +4,22 @@
 import argparse
 import glob
 import os
+import warnings
 
 import numpy
 
 import whitematteranalysis as wma
+from whitematteranalysis.utils.opt_pckg import optional_package
 
-HAVE_PLT = 1
-try:
-    import matplotlib
+matplotlib, have_mpl, _ = optional_package("matplotlib")
+plt, _, _ = optional_package("matplotlib.pyplot")
 
+if have_mpl:
     # Force matplotlib to not use any Xwindows backend.
-    matplotlib.use('Agg')
-    import matplotlib.pyplot as plt
-except:
-    print(f"<{os.path.basename(__file__)}> Error importing matplotlib.pyplot package, can't plot quality control data.\n")
-    HAVE_PLT = 0
+    matplotlib.use("Agg")
+else:
+    warnings.warn(matplotlib._msg)
+    warnings.warn("Cannot plot quality control data.")
 
 
 def _build_arg_parser():
@@ -113,7 +114,7 @@ def main():
         print(cidx + 1,'\t', subjects_per_cluster[cidx],'\t', percent_subjects_per_cluster[cidx] * 100.0, file=clusters_qc_file)
     clusters_qc_file.close()
 
-    if HAVE_PLT:
+    if have_mpl:
         print(f"<{os.path.basename(__file__)}> Saving subjects per cluster histogram.")
         fig, ax = plt.subplots()
         counts = numpy.zeros(num_of_subjects+1)

diff --git a/bin/wm_quality_control_tract_overlap.py b/bin/wm_quality_control_tract_overlap.py
@@ -5,19 +5,20 @@
 import argparse
 import os
 import time
+import warnings
 
 import numpy
 import vtk
 
 import whitematteranalysis as wma
+from whitematteranalysis.utils.opt_pckg import optional_package
 
-HAVE_PLT = 1
+matplotlib, have_mpl, _ = optional_package("matplotlib")
+plt, _, _ = optional_package("matplotlib.pyplot")
 
-try:
-    import matplotlib.pyplot as plt
-except:
-    print(f"<{os.path.basename(__file__)}> Error importing matplotlib.pyplot package, can't plot quality control data.\n")
-    HAVE_PLT = 0    
+if not have_mpl:
+    warnings.warn(matplotlib._msg)
+    warnings.warn("Cannot plot quality control data.")
 
 
 def _build_arg_parser():
@@ -69,7 +70,7 @@ def main():
 
     number_of_subjects = len(input_polydatas)
 
-    if HAVE_PLT:
+    if have_mpl:
         plt.figure(1)
 
     # Loop over subjects and check each

diff --git a/bin/wm_quality_control_tractography.py b/bin/wm_quality_control_tractography.py
@@ -5,18 +5,20 @@
 import os
 import sys
 import time
+import warnings
 
 import numpy
 import vtk
 
 import whitematteranalysis as wma
+from whitematteranalysis.utils.opt_pckg import optional_package
 
-HAVE_PLT = 1
-try:
-    import matplotlib.pyplot as plt
-except:
-    print(f"<{os.path.basename(__file__)}> Error importing matplotlib.pyplot package, can't plot quality control data.\n")
-    HAVE_PLT = 0
+matplotlib, have_mpl, _ = optional_package("matplotlib")
+plt, _, _ = optional_package("matplotlib.pyplot")
+
+if not have_mpl:
+    warnings.warn(matplotlib._msg)
+    warnings.warn("Cannot plot quality control data.")
 
 
 def _build_arg_parser():
@@ -175,7 +177,7 @@ def main():
     spatial_qc_file.write(outstr)
     spatial_qc_file.close()
 
-    if HAVE_PLT:
+    if have_mpl:
         plt.figure(1)
 
     # Loop over subjects and check each
@@ -272,7 +274,7 @@ def main():
         spatial_qc_file.write(outstr)
 
         # Save the subject's fiber lengths  
-        if HAVE_PLT:
+        if have_mpl:
             plt.figure(1)
             if lengths.size > 1:
                 plt.hist(lengths, bins=100, histtype='step', label=subject_id)
@@ -329,7 +331,7 @@ def main():
         del pd3
         subject_idx += 1
 
-    if HAVE_PLT:
+    if have_mpl:
         plt.figure(1)
         plt.title('Histogram of fiber lengths for all subjects')
         plt.xlabel('fiber length (mm)')

diff --git a/bin/wm_separate_clusters_by_hemisphere.py b/bin/wm_separate_clusters_by_hemisphere.py
@@ -21,6 +21,7 @@
 import vtk
 
 import whitematteranalysis as wma
+from whitematteranalysis.utils.utils import hemisphere_loc_name_typo_warn_msg
 
 
 def _build_arg_parser():
@@ -90,7 +91,7 @@ def write_mask_location_to_vtk(inpd, mask_location):
                 array = inpointdata.GetArray(idx)
                 if array.GetName() == 'HemisphereLocataion':
                     warnings.warn(
-                        wma.utils.hemisphere_loc_name_typo_warn_msg,
+                        hemisphere_loc_name_typo_warn_msg,
                         PendingDeprecationWarning)
                     print('  -- HemisphereLocataion is in the input data: skip updating the vtk file.')
                     return inpd
@@ -137,7 +138,7 @@ def _read_location(_inpd):
                 array = inpointdata.GetArray(idx)
                 if array.GetName() == 'HemisphereLocataion':
                     warnings.warn(
-                        wma.utils.hemisphere_loc_name_typo_warn_msg,
+                        hemisphere_loc_name_typo_warn_msg,
                         PendingDeprecationWarning)
                     flag_location = _read_location(inpd)
                     break

diff --git a/whitematteranalysis/__init__.py b/whitematteranalysis/__init__.py
@@ -3,4 +3,4 @@
                io, laterality, mrml, register_two_subjects,
                register_two_subjects_nonrigid,
                register_two_subjects_nonrigid_bsplines, relative_distance,
-               render, similarity, tract_measurement, utils)
+               render, similarity, tract_measurement)
diff --git a/whitematteranalysis/cluster.py b/whitematteranalysis/cluster.py
@@ -30,19 +30,21 @@
 
 from pprint import pprint
 
+from whitematteranalysis.utils.opt_pckg import optional_package
+
 from . import fibers, filter, io, mrml, render, similarity
 
-HAVE_PLT = 1
-try:
-    import matplotlib
+matplotlib, have_mpl, _ = optional_package("matplotlib")
+plt, _, _ = optional_package("matplotlib.pyplot")
 
+if have_mpl:
     # Force matplotlib to not use any Xwindows backend.
-    matplotlib.use('Agg')
-    import matplotlib.pyplot as plt
-except:
-    print(f"<{os.path.basename(__file__)}> Error importing matplotlib.pyplot package, can't plot quality control data.\n")
-    HAVE_PLT = 0
-    
+    matplotlib.use("Agg")
+else:
+    warnings.warn(matplotlib._msg)
+    warnings.warn("Cannot plot quality control data.")
+
+
 # This did not work better. Leave here for future testing if of interest
 if 0:
     try:    
@@ -986,7 +988,7 @@ def output_and_quality_control_cluster_atlas(atlas, output_polydata_s, subject_f
 
     clusters_qc_file.close()
 
-    if HAVE_PLT:
+    if have_mpl:
         print(f"<{os.path.basename(__file__)}> Saving subjects per cluster histogram.")
         fig, ax = plt.subplots()
         counts = numpy.zeros(number_of_subjects+1)

diff --git a/whitematteranalysis/congeal_multisubject.py b/whitematteranalysis/congeal_multisubject.py
@@ -11,23 +11,24 @@ class MultiSubjectRegistration
 
 import os
 import time
+import warnings
 
 import numpy
 import vtk
 from joblib import Parallel, delayed
 
-HAVE_PLT = 1
-try:
-    import matplotlib
+import whitematteranalysis as wma
+from whitematteranalysis.utils.opt_pckg import optional_package
 
-    # Force matplotlib to not use any Xwindows backend.
-    matplotlib.use('Agg')
-    import matplotlib.pyplot as plt
-except:
-    print(f"<{os.path.basename(__file__)}> Error importing matplotlib.pyplot package, can't plot objectives.\n")
-    HAVE_PLT = 0
+matplotlib, have_mpl, _ = optional_package("matplotlib")
+plt, _, _ = optional_package("matplotlib.pyplot")
 
-import whitematteranalysis as wma
+if have_mpl:
+    # Force matplotlib to not use any Xwindows backend.
+    matplotlib.use("Agg")
+else:
+    warnings.warn(matplotlib._msg)
+    warnings.warn("Cannot plot objectives.")
 
 
 class MultiSubjectRegistration:
@@ -304,7 +305,7 @@ def iterate(self):
         functions_per_subject = list()
         objective_changes_per_subject = list()
         decreases = list()
-        if HAVE_PLT:
+        if have_mpl:
             plt.close('all')
             plt.figure(0)
             plt.title('Iteration '+str(self.total_iterations)+' Objective Values for All Subjects')
@@ -326,7 +327,7 @@ def iterate(self):
                 objective_total_after += objectives[0]
             objective_changes_per_subject.append(diff)
             sidx += 1
-            if HAVE_PLT:
+            if have_mpl:
                 plt.figure(0)
                 plt.plot(objectives, 'o-', label=sidx)
 
@@ -342,7 +343,7 @@ def iterate(self):
         print("Iteration:", self.total_iterations, "TOTAL objective change:",  total_change)
         print("Iteration:", self.total_iterations, "PERCENT objective change:",  percent_change)
 
-        if HAVE_PLT:
+        if have_mpl:
             plt.figure(0)
             if self.mode == "Nonrigid":
                 fname_fig_base = "iteration_%05d_sigma_%03d_grid_%03d" % (self.total_iterations, self.sigma, self.nonrigid_grid_resolution)

diff --git a/whitematteranalysis/filter.py b/whitematteranalysis/filter.py
@@ -20,20 +20,24 @@
 """
 
 import os
+import warnings
 
 import numpy
 import vtk
 
-try:
-    from joblib import Parallel, delayed
-    USE_PARALLEL = 1
-except ImportError:
-    USE_PARALLEL = 0
-    print(f"<{os.path.basename(__file__)}> Failed to import joblib, cannot multiprocess.")
-    print(f"<{os.path.basename(__file__)}> Please install joblib for this functionality.")
+from whitematteranalysis.utils.opt_pckg import optional_package
 
 from . import fibers, similarity
 
+joblib, have_joblib, _ = optional_package("joblib")
+Parallel, _, _ = optional_package("joblib.Parallel")
+delayed, _, _ = optional_package("joblib.delayed")
+
+if not have_joblib:
+    warnings.warn(joblib._msg)
+    warnings.warn("Cannot multiprocess.")
+
+
 verbose = 0
 
 
@@ -604,7 +608,7 @@ def remove_outliers(inpd, min_fiber_distance, n_jobs=0, distance_method ='Mean')
     min_fiber_distance = min_fiber_distance * min_fiber_distance
 
     # pairwise distance matrix
-    if USE_PARALLEL and n_jobs > 0:
+    if have_joblib and n_jobs > 0:
         distances = Parallel(n_jobs=n_jobs, verbose=1)(
             delayed(similarity.fiber_distance)(
                 fiber_array.get_fiber(lidx),
@@ -678,7 +682,7 @@ def smooth(inpd, fiber_distance_sigma = 25, points_per_fiber=30, n_jobs=2, upper
     print(f"<{os.path.basename(__file__)}> Computing pairwise distances...")
 
     # pairwise distance matrix
-    if USE_PARALLEL:
+    if have_joblib:
         distances = Parallel(n_jobs=n_jobs, verbose=1)(
             delayed(similarity.fiber_distance)(
             current_fiber_array.get_fiber(lidx),
@@ -802,7 +806,7 @@ def anisotropic_smooth(inpd, fiber_distance_threshold, points_per_fiber=30, n_jo
         done[numpy.nonzero(numpy.array(curr_count) >= cluster_max)] = 1
 
         # pairwise distance matrix
-        if USE_PARALLEL:
+        if have_joblib:
             distances = Parallel(n_jobs=n_jobs, verbose=1)(
                 delayed(similarity.fiber_distance)(
                 current_fiber_array.get_fiber(lidx),
@@ -967,7 +971,7 @@ def laplacian_of_gaussian(inpd, fiber_distance_sigma = 25, points_per_fiber=30,
     fiber_indices = list(range(0, fiber_array.number_of_fibers))
 
     # pairwise distance matrix
-    if USE_PARALLEL:
+    if have_joblib:
         distances = Parallel(n_jobs=n_jobs, verbose=1)(
             delayed(similarity.fiber_distance)(
             fiber_array.get_fiber(lidx),

diff --git a/whitematteranalysis/laterality.py b/whitematteranalysis/laterality.py
@@ -21,22 +21,25 @@ class ComputeWhiteMatterLaterality
 """
 
 import os
+import warnings
 
 import numpy
 import vtk
 
-try:
-    from joblib import Parallel, delayed
-    USE_PARALLEL = 1
-except ImportError:
-    USE_PARALLEL = 0
-    print(f"<{os.path.basename(__file__)}> Failed to import joblib, cannot multiprocess.")
-    print(f"<{os.path.basename(__file__)}> Please install joblib for this functionality.")
+from whitematteranalysis.utils.opt_pckg import optional_package
 
 from . import filter, similarity
 from .fibers import FiberArray
 from .io import LateralityResults
 
+joblib, have_joblib, _ = optional_package("joblib")
+Parallel, _, _ = optional_package("joblib.Parallel")
+delayed, _, _ = optional_package("joblib.delayed")
+
+if not have_joblib:
+    warnings.warn(joblib._msg)
+    warnings.warn("Cannot multiprocess.")
+
 
 def compute_laterality_index(left, right, idx=None):
     ''' Compute laterality index from left and right hemisphere quantities.'''
@@ -171,7 +174,7 @@ def compute(self, input_vtk_polydata):
             print(f"<{os.path.basename(__file__)}> Starting to compute laterality indices")
 
         # run the computation, either in parallel or not
-        if (USE_PARALLEL & (self.parallel_jobs > 1)):
+        if (have_joblib & (self.parallel_jobs > 1)):
             if self.verbose:
                 print(f"<{os.path.basename(__file__)}> Starting parallel code. Processes:", \
                     self.parallel_jobs)

diff --git a/whitematteranalysis/utils/__init__.py b/whitematteranalysis/utils/__init__.py