refactor: replace bottleneck with numpy (bn->np)

src/agora/io/signal.py

@@ -6,7 +6,7 @@
 from functools import cached_property, lru_cache
 from pathlib import Path
 
-import bottleneck as bn
+import numpy as np
 import h5py
 import numpy as np
 import pandas as pd
@@ -109,7 +109,7 @@ def get_retained(df, cutoff):
         Cells must be present for at least cutoff fraction of the total number
         of time points.
         """
-        return df.loc[bn.nansum(df.notna(), axis=1) > df.shape[1] * cutoff]
+        return df.loc[np.nansum(df.notna(), axis=1) > df.shape[1] * cutoff]
 
     @property
     def channels(self) -> t.Collection[str]:

src/agora/utils/kymograph.py

@@ -143,12 +143,12 @@ def filt_cluster(
 
 
 def cluster_kymograph(kymograph: pd.DataFrame, n: int = 2):
-    import bottleneck as bn
+    import numpy as np
     from sklearn.cluster import KMeans
 
     # Normalise according to mean value of signal
     X = (
-        kymograph.divide(bn.nanmean(kymograph, axis=1), axis=0)
+        kymograph.divide(np.nanmean(kymograph, axis=1), axis=0)
         .dropna(axis=1)
         .values
     )

src/extraction/core/extractor.py

@@ -5,7 +5,6 @@
 from pathlib import Path
 
 import aliby.global_settings as global_settings
-import bottleneck as bn
 import h5py
 import numpy as np
 import pandas as pd
@@ -613,7 +612,7 @@ def get_imgs_background_subtract(self, tree, tiles, bgs):
                         # move Z to last column to allow subtraction
                         lambda img, bgs: np.moveaxis(img, 0, -1)
                         # median of background over all pixels for each Z section
-                        - bn.median(img[:, bgs], axis=1),
+                        - np.median(img[:, bgs], axis=1),
                         img[ch],
                         bgs,
                     )

src/extraction/core/functions/cell.py

@@ -15,7 +15,6 @@
 import math
 import typing as t
 
-import bottleneck as bn
 import numpy as np
 from scipy import ndimage
 
@@ -98,7 +97,7 @@ def median(cell_mask, trap_image) -> int:
          Segmentation mask for the cell.
     trap_image: 2d array
     """
-    return bn.median(trap_image[cell_mask])
+    return np.median(trap_image[cell_mask])
 
 
 def max2p5pc(cell_mask, trap_image) -> float:
@@ -116,7 +115,7 @@ def max2p5pc(cell_mask, trap_image) -> float:
     n_top = int(np.ceil(npixels * 0.025))
     # sort pixels in cell and find highest 2.5%
     pixels = trap_image[cell_mask]
-    top_values = bn.partition(pixels, len(pixels) - n_top)[-n_top:]
+    top_values = np.partition(pixels, len(pixels) - n_top)[-n_top:]
     # find mean of these highest pixels
     return np.mean(top_values)
 
@@ -137,7 +136,7 @@ def max5px_median(cell_mask, trap_image) -> float:
     # sort pixels in cell
     pixels = trap_image[cell_mask]
     if len(pixels) > 5:
-        top_values = bn.partition(pixels, len(pixels) - 5)[-5:]
+        top_values = np.partition(pixels, len(pixels) - 5)[-5:]
         # find mean of five brightest pixels
         max5px = np.mean(top_values)
         med = np.median(pixels)

src/extraction/core/functions/distributors.py

@@ -1,6 +1,5 @@
 import typing as t
 
-import bottleneck as bn
 import numpy as np
 
 
@@ -40,7 +39,7 @@ def reduce_z(trap_image: np.ndarray, fun: t.Callable, axis: int = 0):
     if (
         hasattr(fun, "__module__") and fun.__module__[:10] == "bottleneck"
     ):  # Bottleneck type
-        return getattr(bn.reduce, fun.__name__)(trap_image, axis=axis)
+        return getattr(np.reduce, fun.__name__)(trap_image, axis=axis)
     elif isinstance(fun, np.ufunc):
         # optimise the reduction function if possible
         return fun.reduce(trap_image, axis=axis)

src/extraction/core/functions/loaders.py

@@ -5,7 +5,6 @@
 import numpy as np
 from cp_measure.bulk import get_all_measurements
 from skimage.measure import regionprops_table
-import bottleneck as bn
 
 from extraction.core.functions import cell, trap
 from extraction.core.functions.custom import localisation
@@ -152,11 +151,11 @@ def load_redfuns(
         Functions to perform the reduction.
     """
     RED_FUNS = {
-        "max": bn.nanmax,
-        "mean": bn.nanmean,
-        "median": bn.nanmedian,
+        "max": np.nanmax,
+        "mean": np.nanmean,
+        "median": np.nanmedian,
         "div0": div0,
-        "add": bn.nansum,
+        "add": np.nansum,
         "None": None,
     }
     if additional_reducers is not None: