Resolve some deprecation warnings

.github/workflows/pytest.yml

@@ -16,7 +16,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        python-version: ["3.11", "3.12"]
+        python-version: ["3.12", "3.13"]
         os: [ubuntu-latest, macOS-latest, windows-latest]
 
     steps:

CHANGES.rst

@@ -8,6 +8,8 @@ Version 0.2.2   (unreleased)
 
 • Removed deprecated sporco.fista modules
 • Removed deprecation warning redirects for functions renamed in 0.2.0
+• Resolved a number of deprecation warnings
+• Fixed bugs in a number of example scripts
 
 
 Version 0.2.1   (2022-02-17)

docs/requirements.txt

@@ -1,5 +1,6 @@
 future
 scipy
+filetype
 matplotlib
 sphinx>=5.0.0
 numpydoc

docs/source/docntbk.py

@@ -8,6 +8,7 @@
 
 import os
 import os.path
+from pathlib import Path
 from glob import glob
 import re
 import pickle
@@ -68,11 +69,14 @@ def fetch_intersphinx_inventory(uri):
 
     # See https://stackoverflow.com/a/30981554
     class MockConfig(object):
+        intersphinx_cache_limit = None
         intersphinx_timeout = None
         tls_verify = False
+        tls_cacerts = None
+        user_agent = None
 
     class MockApp(object):
-        srcdir = ''
+        srcdir = Path('')
         config = MockConfig()
 
         def warn(self, msg):
@@ -126,6 +130,8 @@ def preprocess_script_string(str):
 
     # Remove header comment
     str = re.sub(r'^(#[^#\n]+\n){5}\n*', r'', str)
+    # Remove r from r""" ... """
+    str = re.sub('^r"""', '"""', str, flags=re.MULTILINE)
     # Insert notebook plotting configuration function
     str = re.sub(r'from sporco import plot', r'from sporco import plot'
                  '\nplot.config_notebook_plotting()',
@@ -500,7 +506,7 @@ def write_notebook_rst(txt, res, fnm, pth):
         # Partial path for current output image
         rpth = os.path.join(extfnm, rnew)
         # In RST text, replace old output name with the new one
-        txt = re.sub('\.\. image:: ' + r, '.. image:: ' + rpth, txt, re.M)
+        txt = re.sub(r'\.\. image:: ' + r, '.. image:: ' + rpth, txt, re.M)
         # Full path of the current output image
         fullrpth = os.path.join(pth, rpth)
         # Write the current output image to disk
@@ -818,7 +824,7 @@ def __init__(self, env, inv, baseurl):
         self.citeid = {}
         if not hasattr(env, 'bibtex_cache'):
             for cite in env.domaindata['cite']['citations']:
-                self.citenum[cite.key] = cite.label
+                self.citenum[cite.key] = cite.citation_id[2:]
                 self.citeid[cite.key] = cite.citation_id
 
 

examples/scripts/cdl/cbpdndl_grd.py

@@ -8,7 +8,7 @@
 Convolutional Dictionary Learning
 =================================
 
-This example demonstrating the use of :class:`.dictlrn.DictLearn` to construct a dictionary learning algorithm with the flexibility of choosing the sparse coding and dictionary update classes. In this case they are :class:`.cbpdn.ConvBPDNGradReg` and :func:`.admm.ccmod.ConvCnstrMOD` respectively, so the resulting dictionary learning algorithm is not equivalent to :class:`.dictlrn.cbpdndl.ConvBPDNDictLearn`. Sparse coding with a CBPDN variant that includes a gradient regularization term on one of the coefficient maps :cite:`wohlberg-2016-convolutional2` enables CDL without the need for the usual lowpass/highpass filtering as a pre-processing of the training images.
+This example demonstrates the use of :class:`.dictlrn.DictLearn` to construct a dictionary learning algorithm with the flexibility of choosing the sparse coding and dictionary update classes. In this case they are :class:`.cbpdn.ConvBPDNGradReg` and :func:`.admm.ccmod.ConvCnstrMOD` respectively, so the resulting dictionary learning algorithm is not equivalent to :class:`.dictlrn.cbpdndl.ConvBPDNDictLearn`. Sparse coding with a CBPDN variant that includes a gradient regularization term on one of the coefficient maps :cite:`wohlberg-2016-convolutional2` enables CDL without the need for the usual lowpass/highpass filtering as a pre-processing of the training images.
 """
 
 
@@ -53,15 +53,15 @@
 cri = cnvrep.CDU_ConvRepIndexing(D0.shape, S)
 
 
-"""
+r"""
 Set up weights for the $\ell_1$ norm to disable regularization of the coefficient map corresponding to the impulse filter.
 """
 
 wl1 = np.ones((1,)*4 + (D0.shape[2:]), dtype=np.float32)
 wl1[..., 0] = 0.0
 
 
-"""
+r"""
 Set of weights for the $\ell_2$ norm of the gradient to disable regularization of all coefficient maps except for the one corresponding to the impulse filter.
 """
 

examples/scripts/cdl/cbpdndl_jnt_clr.py

@@ -8,7 +8,7 @@
 Convolutional Dictionary Learning
 =================================
 
-This example demonstrating the use of :class:`.dictlrn.DictLearn` to construct a dictionary learning algorithm with the flexibility of choosing the sparse coding and dictionary update classes. In this case they are :class:`.cbpdn.ConvBPDNJoint` and :func:`.admm.ccmod.ConvCnstrMOD` respectively, so the resulting dictionary learning algorithm is not equivalent to :class:`.dictlrn.cbpdndl.ConvBPDNDictLearn`. The example uses colour input images and a greyscale dictionary :cite:`wohlberg-2016-convolutional`.
+This example demonstrates the use of :class:`.dictlrn.DictLearn` to construct a dictionary learning algorithm with the flexibility of choosing the sparse coding and dictionary update classes. In this case they are :class:`.cbpdn.ConvBPDNJoint` and :func:`.admm.ccmod.ConvCnstrMOD` respectively, so the resulting dictionary learning algorithm is not equivalent to :class:`.dictlrn.cbpdndl.ConvBPDNDictLearn`. The example uses colour input images and a greyscale dictionary :cite:`wohlberg-2016-convolutional`.
 """
 
 

examples/scripts/cdl/cbpdndl_md_clr.py

@@ -53,7 +53,7 @@
 Sw = W * S
 
 
-"""
+r"""
 $\ell_2$-TV denoising with a spatial mask as a non-linear lowpass filter.
 """
 

examples/scripts/cdl/cbpdndl_md_gry.py

@@ -53,7 +53,7 @@
 Sw = W * S
 
 
-"""
+r"""
 $\ell_2$-TV denoising with a spatial mask as a non-linear lowpass filter.
 """
 

examples/scripts/cdl/cbpdndl_video.py

@@ -34,7 +34,7 @@
 frmlst = []
 for i, frm in enumerate(reader):
     if i >= 250:
-        frm = zoom(signal.rgb2gray(frm.astype(np.float32)/255.0), 0.25)
+        frm = zoom(signal.rgb2gray(np.array(frm, dtype=np.float32)/255.0), 0.25)
         frmlst.append(frm[20:-20, 70:-70])
 vid = np.stack(frmlst, axis=2)
 

examples/scripts/csc/cbpdn_ams_clr.py

@@ -63,7 +63,7 @@
 imgwp = spad(imgw)
 
 
-"""
+r"""
 $\ell_2$-TV denoising with a spatial mask as a non-linear lowpass filter. The highpass component is the difference between the test image and the lowpass component, multiplied by the mask for faster convergence of the convolutional sparse coding (see :cite:`wohlberg-2017-convolutional3`).
 """
 

examples/scripts/csc/cbpdn_ams_grd_gry.py

@@ -78,7 +78,7 @@
 Di = np.dstack((di, D))
 
 
-"""
+r"""
 Set up weights for the $\ell_1$ norm to disable regularization of the coefficient map corresponding to the impulse filter intended to represent lowpass image components (not to be confused with the AMS impulse filter used to implement spatial masking).
 """
 
@@ -87,7 +87,7 @@
 wl1i = np.concatenate((wl1, np.zeros(wl1.shape[0:-1] + (1,))), axis=-1)
 
 
-"""
+r"""
 When representing lowpass image components using an impulse filter together with an $\ell_2$ norm on the gradient of its coefficient map, we usually want to set the weight array for this norm (specified by the ``GradWeight`` option) to disable regularization of all coefficient maps except for the one corresponding to that impulse filter (not to be confused with the AMS impulse filter used to implement spatial masking). In this case set a non-zero value for the weights of the other coefficient maps size this improves performance in this inpainting problem.
 """
 

examples/scripts/csc/cbpdn_ams_gry.py

@@ -69,7 +69,7 @@
 imgwp = spad(imgw)
 
 
-"""
+r"""
 $\ell_2$-TV denoising with a spatial mask as a non-linear lowpass filter. The highpass component is the difference between the test image and the lowpass component, multiplied by the mask for faster convergence of the convolutional sparse coding (see :cite:`wohlberg-2017-convolutional3`).
 """
 

examples/scripts/csc/cbpdn_clr_cd.py

@@ -4,7 +4,7 @@
 # and user license can be found in the 'LICENSE.txt' file distributed
 # with the package.
 
-"""
+r"""
 Multi-channel CSC
 =================
 

examples/scripts/csc/cbpdn_clr_gd.py

@@ -4,7 +4,7 @@
 # and user license can be found in the 'LICENSE.txt' file distributed
 # with the package.
 
-"""
+r"""
 Multi-channel CSC
 =================
 

examples/scripts/csc/cbpdn_clr_pd.py

@@ -4,7 +4,7 @@
 # and user license can be found in the 'LICENSE.txt' file distributed
 # with the package.
 
-"""
+r"""
 Multi-channel CSC
 =================
 

examples/scripts/csc/cbpdn_grd_cuda.py

@@ -48,15 +48,15 @@
 D = np.concatenate((di, Db), axis=2)
 
 
-"""
+r"""
 Set up weights for the $\ell_1$ norm to disable regularization of the coefficient map corresponding to the impulse filter.
 """
 
 wl1 = np.ones((1,)*2 + (D.shape[2:]), dtype=np.float32)
 wl1[..., 0] = 0.0
 
 
-"""
+r"""
 Set of weights for the $\ell_2$ norm of the gradient to disable regularization of all coefficient maps except for the one corresponding to the impulse filter.
 """
 

examples/scripts/csc/cbpdn_gry.py

@@ -4,7 +4,7 @@
 # and user license can be found in the 'LICENSE.txt' file distributed
 # with the package.
 
-"""
+r"""
 Single-channel CSC
 ==================
 

examples/scripts/csc/cbpdn_jnt_clr.py

@@ -4,7 +4,7 @@
 # and user license can be found in the 'LICENSE.txt' file distributed
 # with the package.
 
-"""
+r"""
 Multi-channel CSC
 =================
 

examples/scripts/csc/cbpdn_md_gry.py

@@ -63,7 +63,7 @@
 imgwp = spad(imgw)
 
 
-"""
+r"""
 $\ell_2$-TV denoising with a spatial mask as a non-linear lowpass filter. The highpass component is the difference between the test image and the lowpass component, multiplied by the mask for faster convergence of the convolutional sparse coding (see :cite:`wohlberg-2017-convolutional3`).
 """
 

examples/scripts/csc/cbpdn_pgm_clr.py

@@ -4,7 +4,7 @@
 # and user license can be found in the 'LICENSE.txt' file distributed
 # with the package.
 
-"""
+r"""
 PGM CBPDN Solver
 ==================
 

examples/scripts/csc/cbpdn_pgm_gry.py

@@ -4,7 +4,7 @@
 # and user license can be found in the 'LICENSE.txt' file distributed
 # with the package.
 
-"""
+r"""
 PGM CBPDN Solver
 ==================
 

examples/scripts/csc/cbpdnin_gry.py

@@ -4,7 +4,7 @@
 # and user license can be found in the 'LICENSE.txt' file distributed
 # with the package.
 
-"""
+r"""
 Single-channel CSC With Lateral Inhibition / No Self Inhibition
 ===============================================================
 

examples/scripts/csc/cbpdnin_wgt_gry.py

@@ -4,7 +4,7 @@
 # and user license can be found in the 'LICENSE.txt' file distributed
 # with the package.
 
-"""
+r"""
 Single-channel CSC With Weighted Lateral Inhibition / No Self Inhibition
 ========================================================================
 

examples/scripts/csc/cminl1_gry.py

@@ -4,7 +4,7 @@
 # and user license can be found in the 'LICENSE.txt' file distributed
 # with the package.
 
-"""
+r"""
 Single-channel CSC (Constrained Data Fidelity)
 ==============================================
 

examples/scripts/csc/cprjl1_gry.py

@@ -4,7 +4,7 @@
 # and user license can be found in the 'LICENSE.txt' file distributed
 # with the package.
 
-"""
+r"""
 Single-channel CSC (Constrained Penalty Term)
 =============================================
 

examples/scripts/csc/gwnden_clr.py

@@ -4,8 +4,7 @@
 # and user license can be found in the 'LICENSE.txt' file distributed
 # with the package.
 
-"""
-
+r"""
 Gaussian White Noise Restoration via CSC
 ========================================
 
@@ -50,7 +49,7 @@ def crop(x, n=8):
     return x[n:-n, n:-n]
 
 
-"""
+r"""
 Load a reference image and corrupt it with Gaussian white noise with $\sigma = 0.1$. (The call to ``numpy.random.seed`` ensures that the pseudo-random noise is reproducible.)
 """
 
@@ -76,7 +75,7 @@ def crop(x, n=8):
 D = util.convdicts()['G:8x8x128']
 
 
-"""
+r"""
 Set solver options. See Section 8 of :cite:`wohlberg-2017-convolutional2` for details of construction of $\ell_1$ weighting matrix $W$.
 """
 

examples/scripts/csc/gwnden_gry.py

@@ -4,7 +4,7 @@
 # and user license can be found in the 'LICENSE.txt' file distributed
 # with the package.
 
-"""
+r"""
 Gaussian White Noise Restoration via CSC
 ========================================
 
@@ -47,7 +47,7 @@ def crop(x, n=8):
     return x[n:-n, n:-n]
 
 
-"""
+r"""
 Load a reference image and corrupt it with Gaussian white noise with $\sigma = 0.1$. (The call to ``numpy.random.seed`` ensures that the pseudo-random noise is reproducible.)
 """
 
@@ -73,7 +73,7 @@ def crop(x, n=8):
 D = util.convdicts()['G:8x8x128']
 
 
-"""
+r"""
 Set solver options. See Section 8 of :cite:`wohlberg-2017-convolutional2` for details of construction of $\ell_1$ weighting matrix $W$.
 """
 

examples/scripts/csc/implsden_clr.py

@@ -4,7 +4,7 @@
 # and user license can be found in the 'LICENSE.txt' file distributed
 # with the package.
 
-"""
+r"""
 Impulse Noise Restoration via CSC
 =================================
 
@@ -67,7 +67,7 @@ def crop(x, n=8):
 D = np.concatenate((Di, Di, D0), axis=3)
 
 
-"""
+r"""
 The problem is solved using class :class:`.admm.cbpdn.ConvBPDNGradReg`, which implements the form of CBPDN with an additional gradient regularization term, as defined above. The regularization parameters for the $\ell_1$ and gradient terms are ``lmbda`` and ``mu`` respectively. Setting correct weighting arrays for these regularization terms is critical to obtaining good performance. For the $\ell_1$ norm, the weights on the filters that are intended to represent the impulse noise are tuned to an appropriate value for the impulse noise density (this value sets the relative cost of representing an image feature by one of the impulses or by one of the filters in the learned dictionary), the weights on the filters that are intended to represent low frequency components are set to zero (we only want them penalised by the gradient term), and the weights of the remaining filters are set to zero. For the gradient penalty, all weights are set to zero except for those corresponding to the filters intended to represent low frequency components, which are set to unity.
 """
 
@@ -128,7 +128,7 @@ def crop(x, n=8):
 
 
 
-"""
+r"""
 The previous method gave good results, but the weight on the filter representing the impulse noise is an additional parameter that has to be tuned. This parameter can be avoided by switching to an $\ell_1$ data fidelity term instead of including dictionary filters to represent the impulse noise, as in the problem :cite:`wohlberg-2016-convolutional2`
 
   $$\mathrm{argmin}_{\{\mathbf{x}_m\}} \;