fastremap 1.14.2__tar.gz → 1.15.0__tar.gz

{fastremap-1.14.2 → fastremap-1.15.0}/.github/workflows/build_wheel.yml

@@ -33,6 +33,7 @@ jobs:
         env:
           CIBW_ARCHS_LINUX: ${{matrix.arch}}
           CIBW_BEFORE_BUILD: pip install numpy setuptools wheel cython 
+          CIBW_ARCHS_MACOS: "x86_64 arm64"
 
       - uses: actions/upload-artifact@v2
         with:

{fastremap-1.14.2 → fastremap-1.15.0}/AUTHORS

@@ -1,3 +1,4 @@
 Andreas Schwab <schwab@linux-m68k.org>
 Max <maxme1@users.noreply.github.com>
+William Silversmith <william.silvermsith@gmail.com>
 William Silversmith <william.silversmith@gmail.com>

{fastremap-1.14.2 → fastremap-1.15.0}/ChangeLog

@@ -1,6 +1,16 @@
 CHANGES
 =======
 
+1.15.0
+------
+
+* docs: remove travis badge
+* refactor: clean up repo and put source files in "src" dir
+* ci: add arm64 and x86\_64 to macos build
+* perf: add return\_inverse to fr.unique
+* perf(unique/array): avoid two operations for return\_index
+* feat: point\_cloud supports 2d images
+
 1.14.2
 ------
 

fastremap-1.15.0/MANIFEST.in

@@ -0,0 +1,3 @@
+recursive-include src *
+exclude src/fastremap.cpp
+include LICENSE

{fastremap-1.14.2 → fastremap-1.15.0}/PKG-INFO

@@ -1,12 +1,12 @@
 Metadata-Version: 2.1
 Name: fastremap
-Version: 1.14.2
+Version: 1.15.0
 Summary: Remap, mask, renumber, unique, and in-place transposition of 3D labeled images. Point cloud too.
 Home-page: https://github.com/seung-lab/fastremap/
 Author: William Silversmith
 Author-email: ws9@princeton.edu
 License: UNKNOWN
-Description: [![Build Status](https://travis-ci.org/seung-lab/fastremap.svg?branch=master)](https://travis-ci.org/seung-lab/fastremap) [![PyPI version](https://badge.fury.io/py/fastremap.svg)](https://badge.fury.io/py/fastremap)  
+Description: [![PyPI version](https://badge.fury.io/py/fastremap.svg)](https://badge.fury.io/py/fastremap)  
         
         # fastremap
         

{fastremap-1.14.2 → fastremap-1.15.0}/README.md

@@ -1,4 +1,4 @@
-[![Build Status](https://travis-ci.org/seung-lab/fastremap.svg?branch=master)](https://travis-ci.org/seung-lab/fastremap) [![PyPI version](https://badge.fury.io/py/fastremap.svg)](https://badge.fury.io/py/fastremap)  
+[![PyPI version](https://badge.fury.io/py/fastremap.svg)](https://badge.fury.io/py/fastremap)  
 
 # fastremap
 

{fastremap-1.14.2 → fastremap-1.15.0}/automated_test.py

@@ -408,72 +408,81 @@ def test_unique(order):
 
   # array_unique
   labels = reorder(np.random.randint(0, 500, size=(128,128,128)))
-  uniq_np, idx_np, cts_np = np.unique(labels, return_counts=True, return_index=True)
-  uniq_fr, idx_fr, cts_fr = fastremap.unique(labels, return_counts=True, return_index=True)
+  uniq_np, idx_np, inv_np, cts_np = np.unique(labels, return_counts=True, return_index=True, return_inverse=True)
+  uniq_fr, idx_fr, inv_fr, cts_fr = fastremap.unique(labels, return_counts=True, return_index=True, return_inverse=True)
   assert np.all(uniq_np == uniq_fr)
+  assert np.all(inv_np == inv_fr)
   assert np.all(cts_np == cts_fr)
   assert np.all(labels.flatten()[idx_np] == labels.flatten()[idx_fr])
 
   labels = reorder(np.random.randint(0, 500, size=(128,128,128)))
-  uniq_np, idx_np, cts_np = np.unique(labels, return_counts=True,return_index=True)
-  uniq_fr, idx_fr, cts_fr = fastremap.unique_via_array(labels.flatten(), np.max(labels), return_index=True)
+  uniq_np, idx_np, inv_np, cts_np = np.unique(labels, return_counts=True, return_index=True, return_inverse=True)
+  uniq_fr, idx_fr, cts_fr, inv_fr = fastremap.unique_via_array(labels.flatten(), np.max(labels), return_index=True, return_inverse=True)
   assert np.all(uniq_np == uniq_fr)
+  assert np.all(inv_np.flatten() == inv_fr)
   assert np.all(cts_np == cts_fr)
   assert np.all(labels.flatten()[idx_np] == labels.flatten()[idx_fr])
 
   # array_unique + shift
   labels = reorder(np.random.randint(-500, 500, size=(128,128,128)))
-  uniq_np, idx_np, cts_np = np.unique(labels, return_counts=True, return_index=True)
-  uniq_fr, idx_fr, cts_fr = fastremap.unique(labels, return_counts=True, return_index=True)
+  uniq_np, idx_np, inv_np, cts_np = np.unique(labels, return_counts=True, return_index=True, return_inverse=True)
+  uniq_fr, idx_fr, inv_fr, cts_fr = fastremap.unique(labels, return_counts=True, return_index=True, return_inverse=True)
   assert np.all(uniq_np == uniq_fr)
+  assert np.all(inv_np == inv_fr)
   assert np.all(cts_np == cts_fr)
   assert np.all(labels.flatten()[idx_np] == labels.flatten()[idx_fr])
 
   labels = reorder(np.random.randint(-500, 500, size=(128,128,128)))
-  uniq_np, idx_np, cts_np = np.unique(labels, return_counts=True, return_index=True)
-  uniq_fr, idx_fr, cts_fr = fastremap.unique_via_shifted_array(labels.flatten(), return_index=True)
+  uniq_np, idx_np, inv_np, cts_np = np.unique(labels, return_counts=True, return_index=True, return_inverse=True)
+  uniq_fr, idx_fr, cts_fr, inv_fr = fastremap.unique_via_shifted_array(labels.flatten(), return_index=True, return_inverse=True)
   assert np.all(uniq_np == uniq_fr)
+  assert np.all(inv_np.flatten() == inv_fr)
   assert np.all(cts_np == cts_fr)
   assert np.all(labels.flatten()[idx_np] == labels.flatten()[idx_fr])
 
   # array_unique + shift
   labels = reorder(np.random.randint(128**3 - 500, 128**3 + 500, size=(128,128,128)))
-  uniq_np, idx_np, cts_np = np.unique(labels, return_counts=True, return_index=True)
-  uniq_fr, idx_fr, cts_fr = fastremap.unique(labels, return_counts=True, return_index=True)
+  uniq_np, idx_np, inv_np, cts_np = np.unique(labels, return_counts=True, return_index=True, return_inverse=True)
+  uniq_fr, idx_fr, inv_fr, cts_fr = fastremap.unique(labels, return_counts=True, return_index=True, return_inverse=True)
   assert np.all(uniq_np == uniq_fr)
+  assert np.all(inv_np == inv_fr)
   assert np.all(cts_np == cts_fr)
   assert np.all(labels.flatten()[idx_np] == labels.flatten()[idx_fr])
 
   # array_unique + shift
   labels = reorder(np.random.randint(128**3 - 500, 128**3 + 500, size=(128,128,128)))
-  uniq_np, idx_np, cts_np = np.unique(labels, return_counts=True, return_index=True)
-  uniq_fr, idx_fr, cts_fr = fastremap.unique_via_shifted_array(labels.flatten(), return_index=True)
+  uniq_np, idx_np, inv_np, cts_np = np.unique(labels, return_counts=True, return_index=True, return_inverse=True)
+  uniq_fr, idx_fr, cts_fr, inv_fr = fastremap.unique_via_shifted_array(labels.flatten(), return_index=True, return_inverse=True)
   assert np.all(uniq_np == uniq_fr)
+  assert np.all(inv_np.flatten() == inv_fr)
   assert np.all(cts_np == cts_fr)
   assert np.all(labels.flatten()[idx_np] == labels.flatten()[idx_fr])
 
   # renumber + array_unique
   labels = reorder(np.random.randint(0, 1, size=(128,128,128)))
   labels[0,0,0] = 128**3 + 10
-  uniq_np, idx_np, cts_np = np.unique(labels, return_counts=True, return_index=True)
-  uniq_fr, idx_fr, cts_fr = fastremap.unique(labels, return_counts=True, return_index=True)
+  uniq_np, idx_np, inv_np, cts_np = np.unique(labels, return_counts=True, return_index=True, return_inverse=True)
+  uniq_fr, idx_fr, inv_fr, cts_fr = fastremap.unique(labels, return_counts=True, return_index=True, return_inverse=True)
   assert np.all(uniq_np == uniq_fr)
+  assert np.all(inv_np == inv_fr)
   assert np.all(cts_np == cts_fr)  
   assert np.all(labels.flatten()[idx_np] == labels.flatten()[idx_fr])
 
   labels = reorder(np.random.randint(0, 1, size=(128,128,128)))
   labels[0,0,0] = 128**3 + 10
-  uniq_np, idx_np, cts_np = np.unique(labels, return_counts=True, return_index=True)
-  uniq_fr, idx_fr, cts_fr = fastremap.unique_via_renumber(labels.flatten(), return_index=True)
+  uniq_np, idx_np, inv_np, cts_np = np.unique(labels, return_counts=True, return_index=True, return_inverse=True)
+  uniq_fr, idx_fr, cts_fr, inv_fr = fastremap.unique_via_renumber(labels.flatten(), return_index=True, return_inverse=True)
   assert np.all(uniq_np == uniq_fr)
+  assert np.all(inv_np.flatten() == inv_fr)
   assert np.all(cts_np == cts_fr)  
   assert np.all(labels.flatten()[idx_np] == labels.flatten()[idx_fr])
 
   # sort
   labels = reorder(np.random.randint(-1000, 128**3, size=(100,100,100)))
-  uniq_np, idx_np, cts_np = np.unique(labels, return_counts=True, return_index=True)
-  uniq_fr, idx_fr, cts_fr = fastremap.unique(labels, return_counts=True, return_index=True)
+  uniq_np, idx_np, inv_np, cts_np = np.unique(labels, return_counts=True, return_index=True, return_inverse=True)
+  uniq_fr, idx_fr, inv_fr, cts_fr = fastremap.unique(labels, return_counts=True, return_index=True, return_inverse=True)
   assert np.all(uniq_np == uniq_fr)
+  assert np.all(inv_np == inv_fr)
   assert np.all(cts_np == cts_fr)  
   assert np.all(labels.flatten()[idx_np] == labels.flatten()[idx_fr])
 

{fastremap-1.14.2 → fastremap-1.15.0}/fastremap.egg-info/PKG-INFO

@@ -1,12 +1,12 @@
 Metadata-Version: 2.1
 Name: fastremap
-Version: 1.14.2
+Version: 1.15.0
 Summary: Remap, mask, renumber, unique, and in-place transposition of 3D labeled images. Point cloud too.
 Home-page: https://github.com/seung-lab/fastremap/
 Author: William Silversmith
 Author-email: ws9@princeton.edu
 License: UNKNOWN
-Description: [![Build Status](https://travis-ci.org/seung-lab/fastremap.svg?branch=master)](https://travis-ci.org/seung-lab/fastremap) [![PyPI version](https://badge.fury.io/py/fastremap.svg)](https://badge.fury.io/py/fastremap)  
+Description: [![PyPI version](https://badge.fury.io/py/fastremap.svg)](https://badge.fury.io/py/fastremap)  
         
         # fastremap
         

{fastremap-1.14.2 → fastremap-1.15.0}/fastremap.egg-info/SOURCES.txt

@@ -2,20 +2,18 @@
 AUTHORS
 ChangeLog
 LICENSE
+MANIFEST.in
 README.md
 automated_test.py
 build_linux.sh
-fastremap.pxd
-fastremap.pyx
-ipt.hpp
 manylinux1.Dockerfile
 manylinux2010.Dockerfile
 manylinux2014.Dockerfile
+pyproject.toml
 requirements.txt
 requirements_dev.txt
 setup.cfg
 setup.py
-ska_flat_hash_map.hpp
 test.cpp
 test.py
 tox.ini
@@ -26,4 +24,8 @@ fastremap.egg-info/dependency_links.txt
 fastremap.egg-info/not-zip-safe
 fastremap.egg-info/pbr.json
 fastremap.egg-info/requires.txt
-fastremap.egg-info/top_level.txt
+fastremap.egg-info/top_level.txt
+src/fastremap.pxd
+src/fastremap.pyx
+src/ipt.hpp
+src/ska_flat_hash_map.hpp

fastremap-1.15.0/fastremap.egg-info/pbr.json

@@ -0,0 +1 @@
+{"git_version": "80ce9fc", "is_release": true}

fastremap-1.15.0/pyproject.toml

@@ -0,0 +1,7 @@
+[build-system]
+requires = [
+    "setuptools",
+    "wheel",
+    "numpy",
+    "cython",
+]

{fastremap-1.14.2 → fastremap-1.15.0}/setup.py

@@ -28,11 +28,11 @@ setuptools.setup(
   ext_modules=[
     setuptools.Extension(
       'fastremap',
-      sources=['fastremap.pyx'],
+      sources=['src/fastremap.pyx'],
       depends=[],
       language='c++',
       language_level=3,
-      include_dirs=[str(NumpyImport())],
+      include_dirs=["src", str(NumpyImport())],
       extra_compile_args=extra_compile_args,
     )
   ],

{fastremap-1.14.2 → fastremap-1.15.0/src}/fastremap.pyx

@@ -753,13 +753,26 @@ def unique(labels, return_index=False, return_inverse=False, return_counts=False
 
   return_index: also return the index of the first detected occurance 
     of each label.
+  return_inverse: If True, also return the indices of the unique array 
+    (for the specified axis, if provided) that can be used to reconstruct
+    the input array.
   return_counts: also return the unique label frequency as an array.
 
   Returns: 
-    if return_counts or return_index:
-      return (unique_labels, unique_index, unique_counts)
-    else:
-      return unique_labels
+    unique ndarray
+        The sorted unique values.
+    
+    unique_indices ndarray, optional
+        The indices of the first occurrences of the unique values in the original array. 
+        Only provided if return_index is True.
+    
+    unique_inverse ndarray, optional
+        The indices to reconstruct the original array from the unique array. 
+        Only provided if return_inverse is True.
+    
+    unique_counts ndarray, optional
+        The number of times each of the unique values comes up in the original array. 
+        Only provided if return_counts is True.
   """
   if not isinstance(labels, np.ndarray):
     labels = np.array(labels)
@@ -769,7 +782,7 @@ def unique(labels, return_index=False, return_inverse=False, return_counts=False
 
   # These flags are currently unsupported so call uncle and
   # use the standard implementation instead.
-  if return_inverse or (axis is not None):
+  if axis is not None:
     return np.unique(
       labels, 
       return_index=return_index, 
@@ -781,7 +794,8 @@ def unique(labels, return_index=False, return_inverse=False, return_counts=False
   cdef size_t voxels = labels.size
 
   shape = labels.shape
-  fortran_order = labels.flags['F_CONTIGUOUS']
+  fortran_order = labels.flags.f_contiguous
+  order = "F" if fortran_order else "C"
   labels_orig = labels
   labels = reshape(labels, (voxels,))
 
@@ -792,33 +806,38 @@ def unique(labels, return_index=False, return_inverse=False, return_counts=False
   else:
     min_label, max_label = (0, 0)
 
-  def c_order_index(index):
+  def c_order_index(arr):
     if len(shape) > 1 and fortran_order:
       return np.ravel_multi_index(
-        np.unravel_index(index, shape, order='F'), 
+        np.unravel_index(arr, shape, order='F'), 
         shape, order='C'
       )
-    return index
+    return arr
 
   if voxels == 0:
     uniq = np.array([], dtype=labels.dtype)
     counts = np.array([], dtype=np.uint32)
     index = np.array([], dtype=np.uint64)
+    inverse = np.array([], dtype=np.uintp)
   elif min_label >= 0 and max_label < <int64_t>voxels:
-    uniq, index, counts = unique_via_array(labels, max_label, return_index=return_index)
+    uniq, index, counts, inverse = unique_via_array(labels, max_label, return_index=return_index, return_inverse=return_inverse)
   elif (max_label - min_label) <= <int64_t>voxels:
-    uniq, index, counts = unique_via_shifted_array(labels, min_label, max_label, return_index=return_index)
+    uniq, index, counts, inverse = unique_via_shifted_array(labels, min_label, max_label, return_index=return_index, return_inverse=return_inverse)
   elif float(pixel_pairs(labels)) / float(voxels) > 0.66:
-    uniq, index, counts = unique_via_renumber(labels, return_index=return_index)
-  elif return_index:
-    return np.unique(labels_orig, return_index=return_index, return_counts=return_counts)
+    uniq, index, counts, inverse = unique_via_renumber(labels, return_index=return_index, return_inverse=return_inverse)
+  elif return_index or return_inverse:
+    return np.unique(labels_orig, return_index=return_index, return_counts=return_counts, return_inverse=return_inverse)
   else:
     uniq, counts = unique_via_sort(labels)
+    index = None
+    inverse = None
 
   results = [ uniq ]
   if return_index:
     # This is required to match numpy's behavior
     results.append(c_order_index(index))
+  if return_inverse:
+    results.append(reshape(inverse, shape, order=order))
   if return_counts:
     results.append(counts)
 
@@ -826,23 +845,23 @@ def unique(labels, return_index=False, return_inverse=False, return_counts=False
     return tuple(results)
   return uniq
 
-def unique_via_shifted_array(labels, min_label=None, max_label=None, return_index=False):
+def unique_via_shifted_array(labels, min_label=None, max_label=None, return_index=False, return_inverse=False):
   if min_label is None or max_label is None:
     min_label, max_label = minmax(labels)
 
   labels -= min_label
-  uniq, idx, counts = unique_via_array(labels, max_label - min_label + 1, return_index)
+  uniq, idx, counts, inverse = unique_via_array(labels, max_label - min_label + 1, return_index, return_inverse)
   labels += min_label
   uniq += min_label
-  return uniq, idx, counts
+  return uniq, idx, counts, inverse
 
-def unique_via_renumber(labels, return_index=False):
+def unique_via_renumber(labels, return_index=False, return_inverse=False):
   dtype = labels.dtype
   labels, remap = renumber(labels)
   remap = { v:k for k,v in remap.items() }
-  uniq, idx, counts = unique_via_array(labels, max(remap.keys()), return_index)
+  uniq, idx, counts, inverse = unique_via_array(labels, max(remap.keys()), return_index, return_inverse)
   uniq = np.array([ remap[segid] for segid in uniq ], dtype=dtype)
-  return uniq, idx, counts
+  return uniq, idx, counts, inverse
 
 @cython.boundscheck(False)
 @cython.wraparound(False)  # turn off negative index wrapping for entire function
@@ -886,7 +905,7 @@ def unique_via_sort(cnp.ndarray[ALLINT, ndim=1] labels):
 def unique_via_array(
   cnp.ndarray[ALLINT, ndim=1] labels, 
   size_t max_label, 
-  return_index
+  return_index, return_inverse,
 ):
   cdef cnp.ndarray[uint64_t, ndim=1] counts = np.zeros( 
     (max_label+1,), dtype=np.uint64
@@ -895,9 +914,9 @@ def unique_via_array(
   
   cdef uintptr_t sentinel = np.iinfo(np.uintp).max
   if return_index:
-    index = np.zeros( 
-      (max_label+1,), dtype=np.uintp
-    ) + sentinel
+    index = np.full( 
+      (max_label+1,), sentinel, dtype=np.uintp
+    )
 
   cdef size_t voxels = labels.shape[0]
   cdef size_t i = 0
@@ -935,12 +954,26 @@ def unique_via_array(
     for i in range(max_label + 1):
       if counts[i] > 0:
         idx[j] = index[i]
-        j += 1   
+        j += 1
+  
+  cdef cnp.ndarray[uintptr_t, ndim=1] mapping
+
+  if return_inverse:
+    if segids.size:
+      mapping = np.zeros([segids[segids.size - 1] + 1], dtype=np.uintp)
+      for i in range(real_size):
+        mapping[segids[i]] = i
+      inverse_idx = mapping[labels]
+    else:
+      inverse_idx = np.zeros([0], dtype=np.uintp)
 
+  ret = [ segids, None, cts, None ]
   if return_index:
-    return segids, idx, cts
-  else:
-    return segids, None, cts
+    ret[1] = idx
+  if return_inverse:
+    ret[3] = inverse_idx
+
+  return ret
 
 def transpose(arr):
   """
@@ -1225,20 +1258,77 @@ def _foreground(cnp.ndarray[ALLINT, ndim=1] arr):
     n_foreground += <size_t>(arr[i] != 0)
   return n_foreground
 
-@cython.boundscheck(False)
-@cython.wraparound(False)  # turn off negative index wrapping for entire function
-@cython.nonecheck(False)
-def point_cloud(cnp.ndarray[ALLINT, ndim=3] arr):
+def point_cloud(arr):
   """
   point_cloud(arr)
 
-  Given a 3D integer image, return a mapping from
+  Given a 2D or 3D integer image, return a mapping from
   labels to their (x,y,z) position in the image.
 
   Zero is considered a background label.
 
-  Returns: ndarray(N, 3, dtype=uint16)
+  Returns: ndarray(N, 2 or 3, dtype=uint16)
   """
+  if arr.dtype == bool:
+    arr = arr.view(np.uint8)
+
+  if arr.ndim == 2:
+    return _point_cloud_2d(arr)
+  else:
+    return _point_cloud_3d(arr)
+
+@cython.boundscheck(False)
+@cython.wraparound(False)  # turn off negative index wrapping for entire function
+@cython.nonecheck(False)
+def _point_cloud_2d(cnp.ndarray[ALLINT, ndim=2] arr):
+  cdef size_t n_foreground = foreground(arr)
+
+  cdef size_t sx = arr.shape[0]
+  cdef size_t sy = arr.shape[1]
+
+  if n_foreground == 0:
+    return {}
+
+  cdef cnp.ndarray[ALLINT, ndim=1] ptlabel = np.zeros((n_foreground,), dtype=arr.dtype)
+  cdef cnp.ndarray[uint16_t, ndim=2] ptcloud = np.zeros((n_foreground, 2), dtype=np.uint16)
+
+  cdef size_t i = 0
+  cdef size_t j = 0
+  
+  cdef size_t idx = 0
+  for i in range(sx):
+    for j in range(sy):
+        if arr[i,j] != 0:
+          ptlabel[idx] = arr[i,j]
+          ptcloud[idx,0] = i
+          ptcloud[idx,1] = j
+          idx += 1
+
+  sortidx = ptlabel.argsort()
+  ptlabel = ptlabel[sortidx]
+  ptcloud = ptcloud[sortidx]
+  del sortidx
+
+  ptcloud_by_label = {}
+  if n_foreground == 1:
+    ptcloud_by_label[ptlabel[0]] = ptcloud
+    return ptcloud_by_label
+
+  cdef size_t start = 0
+  cdef size_t end = 0
+  for end in range(1, n_foreground):
+    if ptlabel[end] != ptlabel[end - 1]:
+      ptcloud_by_label[ptlabel[end - 1]] = ptcloud[start:end,:]
+      start = end
+
+  ptcloud_by_label[ptlabel[end]] = ptcloud[start:,:]
+
+  return ptcloud_by_label
+
+@cython.boundscheck(False)
+@cython.wraparound(False)  # turn off negative index wrapping for entire function
+@cython.nonecheck(False)
+def _point_cloud_3d(cnp.ndarray[ALLINT, ndim=3] arr):
   cdef size_t n_foreground = foreground(arr)
 
   cdef size_t sx = arr.shape[0]

{fastremap-1.14.2 → fastremap-1.15.0}/test.py

@@ -1,11 +1,14 @@
 import fastremap
 import numpy as np 
 
-x = np.ones((1700,1700,1700), dtype=np.uint8)
-uniq, cts = fastremap.unique(x, return_counts=True)
-print(uniq, cts)
+x = np.ones((512,512,512), dtype=np.float32)
+x = fastremap.asfortranarray(x)
+print(x)
+print(x.flags)
+print(x.strides)
+
+print(x.dtype)
 
-print(cts[0] / 2**32)
 
 
 # @profile

fastremap-1.14.2/fastremap.egg-info/pbr.json

@@ -1 +0,0 @@
-{"git_version": "9b0e91f", "is_release": true}