Use np.random.default_rng instead of legacy functions (python-graph…

…blas#392)

* Use `np.random.default_rng` instead of legacy functions

* bump ruff (and use pathlib)

.pre-commit-config.yaml

@@ -71,7 +71,7 @@ repos:
         additional_dependencies: [tomli]
         files: ^(graphblas|docs)/
   - repo: https://github.com/charliermarsh/ruff-pre-commit
-    rev: v0.0.247
+    rev: v0.0.249
     hooks:
       - id: ruff
   - repo: https://github.com/sphinx-contrib/sphinx-lint

docs/conf.py

@@ -10,10 +10,10 @@
 # add these directories to sys.path here. If the directory is relative to the
 # documentation root, use os.path.abspath to make it absolute, like shown here.
 #
-import os
 import sys
+from pathlib import Path
 
-sys.path.insert(0, os.path.abspath(".."))
+sys.path.insert(0, str(Path("..").resolve()))
 
 
 # -- Project information -----------------------------------------------------

graphblas/__init__.py

@@ -20,14 +20,14 @@ def __reduce__(self):
 
 
 def get_config():
-    import os
+    from pathlib import Path
 
     import donfig
     import yaml
 
-    filename = os.path.join(os.path.dirname(__file__), "graphblas.yaml")
     config = donfig.Config("graphblas")
-    with open(filename) as f:
+    path = Path(__file__).parent / "graphblas.yaml"
+    with path.open() as f:
         defaults = yaml.safe_load(f)
     config.update_defaults(defaults)
     return config

graphblas/core/automethods.py

@@ -339,7 +339,7 @@ def __ixor__(self, other):
 
 # End auto-generated code
 def _main():
-    import os
+    from pathlib import Path
 
     from .utils import _autogenerate_code
 
@@ -460,7 +460,7 @@ def _main():
             f'    raise TypeError(f"{name!r} not supported for {{type(self).__name__}}")\n\n'
         )
 
-    _autogenerate_code(__file__, "\n".join(lines))
+    _autogenerate_code(Path(__file__), "\n".join(lines))
 
     # Copy to scalar.py and infix.py
     lines = []
@@ -481,16 +481,16 @@ def _main():
             continue
         lines.append(f"    {name} = automethods.{name}")
 
-    thisdir = os.path.dirname(__file__)
+    thisdir = Path(__file__).parent
     infix_exclude = {"_get_value"}
 
     def get_name(line):
         return line.strip().split(" ", 1)[0]
 
     text = "\n".join(lines) + "\n    "
-    _autogenerate_code(os.path.join(thisdir, "scalar.py"), text, "Scalar")
+    _autogenerate_code(thisdir / "scalar.py", text, "Scalar")
     text = "\n".join(line for line in lines if get_name(line) not in infix_exclude) + "\n    "
-    _autogenerate_code(os.path.join(thisdir, "infix.py"), text, "Scalar")
+    _autogenerate_code(thisdir / "infix.py", text, "Scalar")
 
     # Copy to vector.py and infix.py
     lines = []
@@ -519,9 +519,9 @@ def get_name(line):
         lines.append(f"    {name} = automethods.{name}")
 
     text = "\n".join(lines) + "\n    "
-    _autogenerate_code(os.path.join(thisdir, "vector.py"), text, "Vector")
+    _autogenerate_code(thisdir / "vector.py", text, "Vector")
     text = "\n".join(line for line in lines if get_name(line) not in infix_exclude) + "\n    "
-    _autogenerate_code(os.path.join(thisdir, "infix.py"), text, "Vector")
+    _autogenerate_code(thisdir / "infix.py", text, "Vector")
 
     # Copy to matrix.py and infix.py
     lines = []
@@ -550,9 +550,9 @@ def get_name(line):
         lines.append(f"    {name} = automethods.{name}")
 
     text = "\n".join(lines) + "\n    "
-    _autogenerate_code(os.path.join(thisdir, "matrix.py"), text, "Matrix")
+    _autogenerate_code(thisdir / "matrix.py", text, "Matrix")
     text = "\n".join(line for line in lines if get_name(line) not in infix_exclude) + "\n    "
-    _autogenerate_code(os.path.join(thisdir, "infix.py"), text, "Matrix")
+    _autogenerate_code(thisdir / "infix.py", text, "Matrix")
 
 
 if __name__ == "__main__":

graphblas/core/infixmethods.py

@@ -426,9 +426,11 @@ def _main():
         "        setattr(VectorIndexExpr, name, val)\n"
         "        setattr(MatrixIndexExpr, name, val)\n"
     )
+    from pathlib import Path
+
     from .utils import _autogenerate_code
 
-    _autogenerate_code(__file__, "\n".join(lines))
+    _autogenerate_code(Path(__file__), "\n".join(lines))
 
 
 if __name__ == "__main__":

graphblas/core/utils.py

@@ -352,14 +352,14 @@ def __init__(self, matrices, exc_arg=None, *, name):
 
 
 def _autogenerate_code(
-    filename,
+    filepath,
     text,
     specializer=None,
     begin="# Begin auto-generated code",
     end="# End auto-generated code",
 ):
     """Super low-tech auto-code generation used by automethods.py and infixmethods.py."""
-    with open(filename) as f:  # pragma: no branch (flaky)
+    with filepath.open() as f:  # pragma: no branch (flaky)
         orig_text = f.read()
     if specializer:
         begin = f"{begin}: {specializer}"
@@ -379,11 +379,11 @@ def _autogenerate_code(
     new_text = orig_text
     for start, stop in reversed(boundaries):
         new_text = f"{new_text[:start]}{begin}{text}{new_text[stop:]}"
-    with open(filename, "w") as f:  # pragma: no branch (flaky)
+    with filepath.open("w") as f:  # pragma: no branch (flaky)
         f.write(new_text)
     import subprocess
 
     try:
-        subprocess.check_call(["black", filename])
+        subprocess.check_call(["black", filepath])
     except FileNotFoundError:  # pragma: no cover (safety)
         pass  # It's okay if `black` isn't installed; pre-commit hooks will do linting

graphblas/tests/conftest.py

@@ -1,6 +1,7 @@
 import atexit
 import functools
 import itertools
+from pathlib import Path
 
 import numpy as np
 import pytest
@@ -12,26 +13,27 @@
 
 
 def pytest_configure(config):
+    rng = np.random.default_rng()
     randomly = config.getoption("--randomly", False)
     backend = config.getoption("--backend", None)
     if backend is None:
         if randomly:
-            backend = "suitesparse" if np.random.rand() < 0.5 else "suitesparse-vanilla"
+            backend = "suitesparse" if rng.random() < 0.5 else "suitesparse-vanilla"
         else:
             backend = "suitesparse"
     blocking = config.getoption("--blocking", True)
     if blocking is None:  # pragma: no branch
-        blocking = np.random.rand() < 0.5 if randomly else True
+        blocking = rng.random() < 0.5 if randomly else True
     record = config.getoption("--record", False)
     if record is None:  # pragma: no branch
-        record = np.random.rand() < 0.5 if randomly else False
+        record = rng.random() < 0.5 if randomly else False
     mapnumpy = config.getoption("--mapnumpy", False)
     if mapnumpy is None:
-        mapnumpy = np.random.rand() < 0.5 if randomly else False
+        mapnumpy = rng.random() < 0.5 if randomly else False
     runslow = config.getoption("--runslow", False)
     if runslow is None:
         # Add a small amount of randomization to be safer
-        runslow = np.random.rand() < 0.05 if randomly else False
+        runslow = rng.random() < 0.05 if randomly else False
     config.runslow = runslow
 
     gb.config.set(autocompute=False, mapnumpy=mapnumpy)
@@ -46,7 +48,7 @@ def pytest_configure(config):
         rec.start()
 
         def save_records():
-            with open("record.txt", "w") as f:  # pragma: no cover
+            with Path("record.txt").open("w") as f:  # pragma: no cover
                 f.write("\n".join(rec.data))
 
         # I'm sure there's a `pytest` way to do this...
@@ -83,7 +85,7 @@ def pytest_runtest_setup(item):
         pytest.skip("need --runslow option to run")
 
 
-@pytest.fixture(autouse=True, scope="function")
+@pytest.fixture(autouse=True)
 def _reset_name_counters():
     """Reset automatic names for each test for easier comparison of record.txt."""
     gb.Matrix._name_counter = itertools.count()

graphblas/tests/test_dtype.py

@@ -193,7 +193,8 @@ def test_bad_register():
 
 
 def test_auto_register():
-    n = np.random.randint(10, 64)
+    rng = np.random.default_rng()
+    n = rng.integers(10, 64, endpoint=True)
     np_type = np.dtype(f"({n},)int16")
     assert lookup_dtype(np_type).np_type == np_type
 

graphblas/tests/test_matrix.py

@@ -3794,13 +3794,14 @@ def get_expected(row_offset, col_offset, nrows, ncols, is_transposed):
 def test_to_coo_sort():
     # How can we get a matrix to a jumbled state in SS so that export won't be sorted?
     N = 1000000
-    r = np.unique(np.random.randint(N, size=100))
-    c = np.unique(np.random.randint(N, size=100))
+    rng = np.random.default_rng()
+    r = np.unique(rng.integers(N, size=100))
+    c = np.unique(rng.integers(N, size=100))
     r = r[: c.size].copy()  # make sure same length
     c = c[: r.size].copy()
     expected_rows = r.copy()
-    np.random.shuffle(r)
-    np.random.shuffle(c)
+    rng.shuffle(r)
+    rng.shuffle(c)
     A = Matrix.from_coo(r, c, r, nrows=N, ncols=N)
     rows, cols, values = A.to_coo(sort=False)
     A = Matrix.from_coo(r, c, r, nrows=N, ncols=N)

graphblas/tests/test_pickle.py

@@ -1,5 +1,5 @@
-import os
 import pickle
+from pathlib import Path
 
 import numpy as np
 import pytest
@@ -38,12 +38,12 @@ def extra():
 
 @pytest.mark.slow
 def test_deserialize(extra):
-    thisdir = os.path.dirname(__file__)
-    with open(os.path.join(thisdir, f"pickle1{extra}.pkl"), "rb") as f:
+    path = Path(__file__).parent / f"pickle1{extra}.pkl"
+    with path.open("rb") as f:
         d = pickle.load(f)
     check_values(d)
     # Again!
-    with open(os.path.join(thisdir, f"pickle1{extra}.pkl"), "rb") as f:
+    with path.open("rb") as f:
         d = pickle.load(f)
     check_values(d)
 
@@ -287,11 +287,11 @@ def test_serialize_parameterized():
 
 @pytest.mark.slow
 def test_deserialize_parameterized(extra):
-    thisdir = os.path.dirname(__file__)
-    with open(os.path.join(thisdir, f"pickle2{extra}.pkl"), "rb") as f:
+    path = Path(__file__).parent / f"pickle2{extra}.pkl"
+    with path.open("rb") as f:
         pickle.load(f)  # TODO: check results
     # Again!
-    with open(os.path.join(thisdir, f"pickle2{extra}.pkl"), "rb") as f:
+    with path.open("rb") as f:
         pickle.load(f)  # TODO: check results
 
 
@@ -306,8 +306,8 @@ def test_udt(extra):
     assert udt2._is_anonymous
     assert pickle.loads(pickle.dumps(udt2)).np_type == udt2.np_type
 
-    thisdir = os.path.dirname(__file__)
-    with open(os.path.join(thisdir, f"pickle3{extra}.pkl"), "rb") as f:
+    path = Path(__file__).parent / f"pickle3{extra}.pkl"
+    with path.open("rb") as f:
         d = pickle.load(f)
     udt3 = d["PickledUDT"]
     v = d["v"]

graphblas/tests/test_prefix_scan.py

@@ -14,7 +14,8 @@
 @pytest.mark.parametrize("do_random", [False, True])
 def test_scan_matrix(method, length, do_random):
     if do_random:
-        A = np.random.randint(10, size=2 * length).reshape((2, length))
+        rng = np.random.default_rng()
+        A = rng.integers(10, size=2 * length).reshape((2, length))
         mask = (A % 2).astype(bool)
         A[~mask] = 0
         M = Matrix.ss.import_bitmapr(values=A, bitmap=mask, name="A")
@@ -44,7 +45,8 @@ def test_scan_matrix(method, length, do_random):
 @pytest.mark.parametrize("do_random", [False, True])
 def test_scan_vector(length, do_random):
     if do_random:
-        a = np.random.randint(10, size=length)
+        rng = np.random.default_rng()
+        a = rng.integers(10, size=length)
         mask = (a % 2).astype(bool)
         a[~mask] = 0
         v = Vector.ss.import_bitmap(values=a, bitmap=mask)

graphblas/tests/test_vector.py

@@ -2347,9 +2347,10 @@ def get_expected(offset, size):
 def test_to_coo_sort():
     # How can we get a vector to a jumbled state in SS so that export won't be sorted?
     N = 1000000
-    a = np.unique(np.random.randint(N, size=100))
+    rng = np.random.default_rng()
+    a = np.unique(rng.integers(N, size=100))
     expected = a.copy()
-    np.random.shuffle(a)
+    rng.shuffle(a)
     v = Vector.from_coo(a, a, size=N)
     indices, values = v.to_coo(sort=False)
     v = Vector.from_coo(a, a, size=N)