try to fix cython int type error; add github CI for quick checks

closes issue #134
(hopefully)

.circleci/config.yml

@@ -17,7 +17,8 @@ jobs:
 
       - run: sudo ln -s /usr/lib/python3/dist-packages/numpy/core/include/numpy/ /usr/include/numpy
 
-      - run: sudo python3 -m pip install -r pip-requirements.txt pytest-html coveralls pyyaml mpi4py
+      - run: sudo python3 -m pip install -r pip-requirements.txt pytest-html coveralls pyyaml mpi4py pydocstyle pycodestyle flake8
+
       - run: mkdir -p test-reports
 
       - run: python3 -m pip install -e .
@@ -35,6 +36,10 @@ jobs:
       - run: coverage3 report --include="$PWD/*" --omit="$PWD/.eggs/*"
       - run: coverage3 html --include="$PWD/*" --omit="$PWD/.eggs/*" && mv htmlcov test-reports
 
+      - run: flake8 $(ls ultranest/*.py | grep -Ev '^ultranest/(flatnuts|dychmc|dyhmc).py')
+      - run: pycodestyle $(ls ultranest/*.py | grep -Ev '^ultranest/(flatnuts|dychmc|dyhmc).py')
+      - run: pydocstyle $(ls ultranest/*.py | grep -Ev '^ultranest/(flatnuts|dychmc|dyhmc).py')
+
       - run: coveralls
 
       - store_test_results:

.github/workflows/tests.yml

@@ -0,0 +1,44 @@
+# This workflow will install Python dependencies, run tests and lint with a variety of Python versions
+# For more information see: https://help.github.com/actions/language-and-framework-guides/using-python-with-github-actions
+
+name: build
+
+on:
+  push:
+  pull_request:
+  schedule:
+    - cron: '42 4 5,20 * *'
+
+jobs:
+  build:
+
+    runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        python-version: [3.7, 3.8, 3.9, "3.10", 3.11, 3.12]
+
+    steps:
+    - uses: actions/checkout@v2
+    - name: Set up Python ${{ matrix.python-version }}
+      uses: actions/setup-python@v2
+      with:
+        python-version: ${{ matrix.python-version }}
+
+    - name: Install dependencies
+      run: python -m pip install -r pip-requirements.txt
+
+    - name: Lint with flake8
+      run: flake8 $(ls ultranest/*.py | grep -Ev '^ultranest/(flatnuts|dychmc|dyhmc).py')
+
+    - name: Check code style
+      run: pycodestyle $(ls ultranest/*.py | grep -Ev '^ultranest/(flatnuts|dychmc|dyhmc).py')
+
+    - name: Check doc style
+      run: pydocstyle $(ls ultranest/*.py | grep -Ev '^ultranest/(flatnuts|dychmc|dyhmc).py')
+
+    - name: Install package
+      run: python -m pip install -e .
+
+    - name: Test with pytest
+      run: pytest -n -x

ultranest/integrator.py

@@ -37,6 +37,8 @@
 
 __all__ = ['ReactiveNestedSampler', 'NestedSampler', 'read_file', 'warmstart_from_similar_file']
 
+int_t = int
+
 
 def _get_cumsum_range(pi, dp):
     """Compute quantile indices from probabilities.
@@ -2015,8 +2017,8 @@ def _update_region(
 
             # instead, track the clusters from before by matching manually
             oldt = self.transformLayer.transform(oldu)
-            clusterids = np.zeros(len(active_u), dtype=np.int_)
-            nnearby = np.empty(len(self.region.unormed), dtype=np.int_)
+            clusterids = np.zeros(len(active_u), dtype=int_t)
+            nnearby = np.empty(len(self.region.unormed), dtype=int_t)
             for ci in np.unique(self.transformLayer.clusterids):
                 if ci == 0:
                     continue

ultranest/mlfriends.pyx

@@ -27,7 +27,7 @@ cdef count_nearby(
     np.ndarray[np.float_t, ndim=2] apts,
     np.ndarray[np.float_t, ndim=2] bpts,
     np.float_t radiussq,
-    np.ndarray[np.int_, ndim=1] nnearby
+    np.ndarray[np.int_t, ndim=1] nnearby
 ):
     """Count the number of points in ``apts`` within square radius ``radiussq`` for each point ``b`` in `bpts``.
 
@@ -139,7 +139,7 @@ def find_nearby(
     np.ndarray[np.float_t, ndim=2] apts,
     np.ndarray[np.float_t, ndim=2] bpts,
     np.float_t radiussq,
-    np.ndarray[np.int_, ndim=1] nnearby
+    np.ndarray[np.int_t, ndim=1] nnearby
 ):
     """Gets the index of a point in `a` within square radius `radiussq`, for each point `b` in `bpts`.
 
@@ -223,7 +223,7 @@ cdef float compute_maxradiussq(np.ndarray[np.float_t, ndim=2] apts, np.ndarray[n
 @cython.wraparound(False)
 def compute_mean_pair_distance(
     np.ndarray[np.float_t, ndim=2] pts,
-    np.ndarray[np.int_, ndim=1] clusterids
+    np.ndarray[np.int_t, ndim=1] clusterids
 ):
     """Compute the average distance between pairs of points.
     Pairs from different clusters are excluded in the computation.
@@ -271,7 +271,7 @@ cdef _update_clusters(
     np.ndarray[np.float_t, ndim=2] upoints,
     np.ndarray[np.float_t, ndim=2] tpoints,
     np.float_t maxradiussq,
-    np.ndarray[np.int_, ndim=1] clusterids,
+    np.ndarray[np.int_t, ndim=1] clusterids,
 ):
     """same signature as ``update_clusters()``, see there."""
     assert upoints.shape[0] == tpoints.shape[0], ('different number of points', upoints.shape[0], tpoints.shape[0])
@@ -845,7 +845,7 @@ class LocalAffineLayer(AffineLayer):
         return s
 
 
-def vol_prefactor(np.int_ n):
+def vol_prefactor(np.int_t n):
     """Volume constant for an ``n``-dimensional sphere.
 
     for ``n`` even:  $$    (2pi)^(n    /2) / (2 * 4 * ... * n)$$

ultranest/popstepsampler.py

@@ -17,6 +17,8 @@
 from ultranest.stepfuncs import generate_differential_direction, generate_mixture_random_direction
 import scipy.stats
 
+int_t = int
+
 
 def unitcube_line_intersection(ray_origin, ray_direction):
     r"""Compute intersection of a line (ray) and a unit box (0:1 in all axes).
@@ -425,7 +427,7 @@ def _setup(self, ndim):
         self.allL = np.zeros((self.popsize, self.nsteps + 1)) + np.nan
         self.currentt = np.zeros(self.popsize) + np.nan
         self.currentv = np.zeros((self.popsize, ndim)) + np.nan
-        self.generation = np.zeros(self.popsize, dtype=np.int_) - 1
+        self.generation = np.zeros(self.popsize, dtype=int_t) - 1
         self.current_left = np.zeros(self.popsize)
         self.current_right = np.zeros(self.popsize)
         self.searching_left = np.zeros(self.popsize, dtype=bool)
@@ -923,9 +925,9 @@ def __next__(
                 # Slice bounds for each points
                 tleft, tright = self.slice_limit(tleft_unitcube,tright_unitcube)
                 # Index of the workers working concurrently
-                worker_running = np.arange(0, self.popsize, 1, dtype=np.int_)
+                worker_running = np.arange(0, self.popsize, 1, dtype=int_t)
                 # Status indicating if a points has already find its next position
-                status = np.zeros(self.popsize, dtype=np.int_)  # one for success, zero for running
+                status = np.zeros(self.popsize, dtype=int_t)  # one for success, zero for running
 
                 # Loop until each points has found its next position or we reached 100 iterations
                 for it in range(self.max_it):

ultranest/stepfuncs.pyx

@@ -331,7 +331,7 @@ def step_back(Lmin, allL, generation, currentt, log=False):
 
 cdef _fill_directions(
     np.ndarray[np.float_t, ndim=2] v,
-    np.ndarray[np.int_, ndim=1] indices,
+    np.ndarray[np.int_t, ndim=1] indices,
     float scale
 ):
     cdef size_t nsamples = v.shape[0]
@@ -533,7 +533,7 @@ cpdef tuple update_vectorised_slice_sampler(\
                           np.ndarray[np.float_t, ndim=1] t, np.ndarray[np.float_t, ndim=1] tleft,\
                           np.ndarray[np.float_t, ndim=1] tright, np.ndarray[np.float_t, ndim=1] proposed_L,\
                           np.ndarray[np.float_t, ndim=2] proposed_u, np.ndarray[np.float_t, ndim=2] proposed_p,\
-                          np.ndarray[np.int_, ndim=1] worker_running, np.ndarray[np.int_, ndim=1] status,\
+                          np.ndarray[np.int_t, ndim=1] worker_running, np.ndarray[np.int_t, ndim=1] status,\
                           np.float_t Likelihood_threshold,np.float_t shrink_factor, np.ndarray[np.float_t, ndim=2] allu,\
                           np.ndarray[np.float_t, ndim=1] allL, np.ndarray[np.float_t, ndim=2] allp, int popsize):
 

ultranest/stepsampler.py

@@ -466,33 +466,36 @@ def select_random_livepoint(us, Ls, Lmin):
 
 
 class IslandPopulationRandomLivepointSelector(object):
+    """Mutually isolated live point subsets.
+
+    To replace dead points, chains are only started from the same
+    island as the dead point. Island refers to chunks of
+    live point indices (0,1,2,3 as stored, not sorted).
+    Each chunk has size ´island_size´.
+
+    If ´island_size´ is large, for example, the total number of live points,
+    then clumping can occur more easily. This is the observed behaviour
+    that a limited random walk is run from one live point, giving
+    two similar points, then the next dead point replacement is
+    likely run again from these, giving more and more similar live points.
+    This gives a run-away process leading to clumps of similar,
+    highly correlated points.
+
+    If ´island_size´ is small, for example, 1, then each dead point
+    is replaced by a chain started from it. This is a problem because
+    modes can never die out. Nested sampling can then not complete.
+
+    In a multi-modal run, within a given number of live points,
+    the number of live points per mode is proportional to the mode's
+    prior volume, but can fluctuate. If the number of live points
+    is small, a fluctuation can lead to mode die-out, which cannot
+    be reversed. Therefore, the number of island members should be
+    large enough to represent each mode.
+    """
+
     def __init__(self, island_size, exchange_probability=0):
         """Set up multiple isolated islands.
 
-        To replace dead points, chains are only started from the same
-        island as the dead point. Island refers to chunks of
-        live point indices (0,1,2,3 as stored, not sorted).
-        Each chunk has size ´island_size´.
-
-        If ´island_size´ is large, for example, the total number of live points,
-        then clumping can occur more easily. This is the observed behaviour
-        that a limited random walk is run from one live point, giving
-        two similar points, then the next dead point replacement is
-        likely run again from these, giving more and more similar live points.
-        This gives a run-away process leading to clumps of similar,
-        highly correlated points.
-
-        If ´island_size´ is small, for example, 1, then each dead point
-        is replaced by a chain started from it. This is a problem because
-        modes can never die out. Nested sampling can then not complete.
-
-        In a multi-modal run, within a given number of live points,
-        the number of live points per mode is proportional to the mode's
-        prior volume, but can fluctuate. If the number of live points
-        is small, a fluctuation can lead to mode die-out, which cannot
-        be reversed. Therefore, the number of island members should be
-        large enough to represent each mode.
-
         Parameters
         -----------
         island_size: int