69 deprecation warnings need dealing with

clarity/data/HOA_tools_cec2.py

@@ -4,6 +4,7 @@
 
 import numpy as np
 from numba import njit
+from numba.typed import List as TypedList
 from scipy.signal import convolve
 from scipy.spatial.transform import Rotation as R
 from scipy.special import comb
@@ -45,9 +46,14 @@ def compute_rotation_matrix(n: int, foa_rotmat: np.ndarray) -> np.ndarray:
     sub_matrices = [np.eye(i * 2 + 1) for i in np.arange(n + 1)]
     sub_matrices[1] = foa_rotmat
 
+    typed_sub_matrices = TypedList()
+    [typed_sub_matrices.append(x) for x in sub_matrices]
+
     if n > 1:
         for i in np.arange(2, n + 1):
-            rot_mat, sub_matrices = compute_band_rotation(i, sub_matrices, rot_mat)
+            rot_mat, typed_sub_matrices = compute_band_rotation(
+                i, typed_sub_matrices, rot_mat
+            )
 
     return rot_mat
 

clarity/data/scene_renderer_cec2.py

@@ -2,6 +2,7 @@
 import json
 import logging
 import math
+import warnings
 from pathlib import Path
 
 import librosa
@@ -169,10 +170,13 @@ def load_interferer_signals(self, scene):
         """
         interferer_audio_paths = self.prepare_interferer_paths(scene)
         # NOTE: all interferer signals are assumed to by at the Clarity fs = 44100 Hz
-        signals = [
-            librosa.load(signal_path, sr=None)[0]
-            for signal_path in interferer_audio_paths
-        ]
+        with warnings.catch_warnings():
+            # Suppress annoying warning generated by librosa when reading mp3 files
+            warnings.simplefilter("ignore")
+            signals = [
+                librosa.load(signal_path, sr=None)[0]
+                for signal_path in interferer_audio_paths
+            ]
         signal_starts = [interferer["offset"] for interferer in scene["interferers"]]
         signal_lengths = [
             interferer["time_end"] - interferer["time_start"]

clarity/enhancer/gha/gainrule_camfit.py

@@ -152,7 +152,7 @@ def freq_interp_sh(f_in, y_in, f):
         y_in = y_in[0]
 
     vals = np.pad(
-        f_in.astype(np.float), 1, constant_values=((0.5 * f_in[0], 2 * f_in[-1]))
+        f_in.astype(float), 1, constant_values=((0.5 * f_in[0], 2 * f_in[-1]))
     )
     yvals = np.pad(y_in, 1, constant_values=((y_in[0], y_in[-1])))
 

clarity/predictor/torch_msbg.py

@@ -33,15 +33,15 @@
 
 def generate_key_percent(signal, thr_dB, winlen):
     if winlen != np.floor(winlen):
-        winlen = np.int(np.floor(winlen))
+        winlen = int(np.floor(winlen))
         print("\nGenerate_key_percent: \t Window length must be integer")
 
     siglen = len(signal)
     expected = thr_dB.copy()  # expected threshold
     non_zero = 10.0 ** ((expected - 30) / 10)  # put floor into histogram distribution
 
     nframes = 0
-    totframes = np.int(np.floor(siglen / winlen))
+    totframes = int(np.floor(siglen / winlen))
     every_dB = np.zeros(totframes)
 
     for ix in range(totframes):
@@ -56,12 +56,12 @@ def generate_key_percent(signal, thr_dB, winlen):
     # between the two peaks, and set a bit above that, as it heads for main peak
     frame_idx = np.where(every_dB >= expected)[0]
     valid_frames = len(frame_idx)
-    key = np.zeros(valid_frames * winlen, dtype=np.int)
+    key = np.zeros(valid_frames * winlen, dtype=int)
 
     # convert frame numbers into indices for signal
     for ix in range(valid_frames):
         key[ix * winlen : ix * winlen + winlen] = np.arange(
-            frame_idx[ix] * winlen, frame_idx[ix] * winlen + winlen, dtype=np.int
+            frame_idx[ix] * winlen, frame_idx[ix] * winlen + winlen, dtype=int
         )
     return key, used_thr_dB
 
@@ -84,7 +84,7 @@ def measure_rms(signal, sr, dB_rel_rms):
     # use this RMS to generate key threshold to more accurate RMS
     key_thr_dB = np.max([20 * np.log10(first_stage_rms) + dB_rel_rms, -80])
     key, used_thr_dB = generate_key_percent(
-        signal, key_thr_dB, np.int(np.round(win_secs * sr))
+        signal, key_thr_dB, int(np.round(win_secs * sr))
     )
     # active = 100.0 * len(key) / len(signal)
     rms = np.sqrt(np.mean(signal[key] ** 2))
@@ -94,18 +94,18 @@ def measure_rms(signal, sr, dB_rel_rms):
 
 def makesmearmat3(rl, ru, sr):
     fftsize = 512
-    nyquist = np.int(fftsize // 2)
+    nyquist = int(fftsize // 2)
     fnor = audfilt(1, 1, nyquist, sr)
     fwid = audfilt(rl, ru, nyquist, sr)
     fnext = np.hstack([fnor, np.zeros([nyquist, nyquist // 2])])
 
-    for i in np.arange(nyquist // 2 + 1, nyquist + 1, dtype=np.int):
+    for i in np.arange(nyquist // 2 + 1, nyquist + 1, dtype=int):
         fnext[i - 1, nyquist : np.min([2 * i - 1, 3 * nyquist // 2])] = np.flip(
             fnor[
                 i - 1, np.max([1, 2 * i - 3 * nyquist // 2]) - 1 : (2 * i - nyquist - 1)
             ]
         )
-    fsmear = np.linalg.lstsq(fnext, fwid)[
+    fsmear = np.linalg.lstsq(fnext, fwid, rcond=-1)[
         0
     ]  # https://stackoverflow.com/questions/33559946/numpy-vs-mldivide-matlab-operator
     fsmear = fsmear[:nyquist, :]
@@ -130,15 +130,15 @@ def audfilt(rl, ru, size, sr):
     aud_filter[0, 0] = aud_filter[0, 0] / ((rl + ru) / 2)
 
     g = np.zeros(size)
-    for i in np.arange(2, size + 1, 1, dtype=np.int):
+    for i in np.arange(2, size + 1, 1, dtype=int):
         fhz = (i - 1) * sr / (2 * size)
         erbhz = 24.7 * ((fhz * 0.00437) + 1)
         pl = 4 * fhz / (erbhz * rl)
         pu = 4 * fhz / (erbhz * ru)
-        j = np.arange(1, i, dtype=np.int)
+        j = np.arange(1, i, dtype=int)
         g[j - 1] = np.abs((i - j) / (i - 1))
         aud_filter[i - 1, j - 1] = (1 + (pl * g[j - 1])) * np.exp(-pl * g[j - 1])
-        j = np.arange(i, size + 1, dtype=np.int)
+        j = np.arange(i, size + 1, dtype=int)
         g[j - 1] = np.abs((i - j) / (i - 1))
         aud_filter[i - 1, j - 1] = (1 + (pu * g[j - 1])) * np.exp(-pu * g[j - 1])
         aud_filter[i - 1, :] = aud_filter[i - 1, :] / (erbhz * (rl + ru) / (2 * 24.7))
@@ -427,7 +427,7 @@ def __init__(
         corrn_used = np.append(corrn[ixf_useful], last_corrn)
         corrn_forward = 10 ** (0.05 * corrn_used)
         corrn_backward = 10 ** (0.05 * -1 * corrn_used)
-        n_wdw = np.int(2 * np.floor((sr / 16e3) * 368 / 2))
+        n_wdw = int(2 * np.floor((sr / 16e3) * 368 / 2))
         coch_filter_forward = firwin2(
             n_wdw + 1, hz_used / nyquist, corrn_forward, window=("kaiser", 4)
         )

pyproject.toml

@@ -13,7 +13,6 @@ testpaths = [
     "tests",
 ]
 filterwarnings = [
-    "ignore::DeprecationWarning",
     "ignore::UserWarning"
 ]
 

tests/regression/test_data_HOA_tools_cec2.py

@@ -1,4 +1,5 @@
 import numpy as np
+from numba.typed import List
 from scipy.spatial.transform import Rotation as R
 
 from clarity.data import HOA_tools_cec2 as hoa
@@ -34,7 +35,7 @@ def test_U(regtest):
     theta = 45
     order = 2
     foa = R.from_euler("y", theta, degrees=True).as_matrix()
-    rotmats = [foa, hoa.compute_rotation_matrix(order, foa)]
+    rotmats = List([foa, hoa.compute_rotation_matrix(order, foa)])
     u = hoa.U.py_func(m, n, order, rotmats)
     regtest.write(f"U value {u:0.3f}\n")
 
@@ -44,7 +45,7 @@ def test_V(regtest):
     theta = 45
     order = 2
     foa = R.from_euler("y", theta, degrees=True).as_matrix()
-    rotmats = [foa, hoa.compute_rotation_matrix(order, foa)]
+    rotmats = List([foa, hoa.compute_rotation_matrix(order, foa)])
     for i in range(3):
         v = hoa.V.py_func(i - 1, n, order, rotmats)
         regtest.write(f"V[{i}] value {v:0.3f}\n")
@@ -55,7 +56,7 @@ def test_W(regtest):
     theta = 45
     order = 2
     foa = R.from_euler("y", theta, degrees=True).as_matrix()
-    rotmats = [foa, hoa.compute_rotation_matrix(order, foa)]
+    rotmats = List([foa, hoa.compute_rotation_matrix(order, foa)])
     for i in range(3):
         w = hoa.W.py_func(i - 1, n, order, rotmats)
         regtest.write(f"W[{i}] value {w:0.3f}\n")
@@ -84,10 +85,12 @@ def test_compute_band_rotation(regtest):
 
     sub_matrices = [np.eye(i * 2 + 1) for i in np.arange(n + 1)]
     sub_matrices[1] = foa_rotmat
+    typed_sub_matrices = List()
+    [typed_sub_matrices.append(x) for x in sub_matrices]
 
     if n > 1:
         for i in np.arange(2, n + 1):
-            rot_mat, sub_matrices = hoa.compute_band_rotation.py_func(
-                i, sub_matrices, rot_mat
+            rot_mat, typed_sub_matrices = hoa.compute_band_rotation.py_func(
+                i, typed_sub_matrices, rot_mat
             )
     regtest.write(f"Band rotations {rot_mat}, {sub_matrices}\n")