Merge pull request #731 from StingraySoftware/simplify_numba_mock

Simplify numba mocking functions and improve docs

docs/changes/731.trivial.rst

@@ -0,0 +1 @@
+Simplify numba mocking code, and possibly improve code coverage estimate

docs/install.rst

@@ -1,3 +1,53 @@
+Dependencies
+============
+A **minimal installation** of Stingray requires the following dependencies:
+
++ astropy>=4.0
++ numpy>=1.17.0
++ scipy>=1.1.0
++ matplotlib>=3.0,!=3.4.0
+
+In **typical** uses, requiring input/output, caching of results, and faster processing, we **recommend the following dependencies**:
+
++ numba (**highly** recommended)
++ tbb (needed by numba)
++ tqdm (for progress bars, always useful)
++ pyfftw (for the fastest FFT in the West)
++ h5py (for input/output)
++ pyyaml (for input/output)
++ emcee (for MCMC analysis, e.g. for PSD fitting)
++ corner (for the plotting of MCMC results)
++ statsmodels (for some statistical analysis)
+
+For **pulsar searches and timing**, we recommend installing
+
++ pint-pulsar
+
+Some of the dependencies are available in ``conda``, the others via ``pip``.
+To install all required and recommended dependencies in a recent installation, you should be good running the following command:
+
+    $ pip install astropy scipy matplotlib numpy h5py tqdm numba pint-pulsar emcee corner statsmodels pyfftw tbb
+
+For development work, you will need the following extra libraries:
+
++ pytest
++ pytest-astropy
++ tox
++ jinja2<=3.0.0
++ docutils
++ sphinx-astropy
++ nbsphinx>=0.8.3,!=0.8.8
++ pandoc
++ ipython
++ jupyter
++ notebook
++ towncrier<22.12.0
++ black
+
+Which can be installed with the following command:
+
+    $ pip install pytest pytest-astropy jinja2<=3.0.0 docutils sphinx-astropy nbsphinx pandoc ipython jupyter notebook towncrier<22.12.0 tox black
+
 Downloading and Installing Stingray
 ===================================
 
@@ -37,12 +87,7 @@ source (it *will* have bugs; you've been warned!), first clone
 
     $ git clone --recursive https://github.com/StingraySoftware/stingray.git
 
-Now ``cd`` into the newly created ``stingray`` directory and install the necessary
-dependencies: ::
-
-    $ cd stingray
-    $ pip install astropy scipy matplotlib numpy pytest pytest-astropy h5py tqdm
-
+Now ``cd`` into the newly created ``stingray`` directory.
 Finally, install ``stingray`` itself: ::
 
     $ pip install -e "."

setup.cfg

@@ -18,7 +18,6 @@ python_requires = >=3.8
 setup_requires = setuptools_scm
 install_requires =
     astropy>=4.0
-    six
     numpy>=1.17.0
     scipy>=1.1.0
 ; Matplotlib 3.4.0 is incompatible with Astropy

stingray/tests/test_utils.py

@@ -406,3 +406,24 @@ def test_is_sorted_no_array():
     input_array = 1.0
 
     assert utils.is_sorted(input_array)
+
+
+@utils.njit()
+def bla_1():
+    return
+
+
+@utils.njit
+def bla_2():
+    return
+
+
+@utils.vectorize(["float64(float64)"])
+def bla_3(j):
+    return j
+
+
+def test_numba_compiled_or_mocked():
+    bla_1()
+    bla_2()
+    bla_3(1.0)

stingray/utils.py

@@ -40,7 +40,6 @@
 
 # If numba is installed, import jit. Otherwise, define an empty decorator with
 # the same name.
-HAS_NUMBA = False
 try:
     from numba import jit
 
@@ -49,36 +48,25 @@
     from numba.core.errors import NumbaValueError
 except ImportError:
     warnings.warn("Numba not installed. Faking it")
+    HAS_NUMBA = False
     NumbaValueError = Exception
 
-    class jit(object):
-        def __init__(self, *args, **kwargs):
-            pass
+    def njit(f=None, *args, **kwargs):
+        def decorator(func, *a, **kw):
+            return func
 
-        def __call__(self, func):
-            def wrapped_f(*args, **kwargs):
-                return func(*args, **kwargs)
-
-            return wrapped_f
-
-    class njit(object):
-        def __init__(self, *args, **kwargs):
-            pass
-
-        def __call__(self, func):
-            def wrapped_f(*args, **kwargs):
-                return func(*args, **kwargs)
-
-            return wrapped_f
+        if callable(f):
+            return f
+        else:
+            return decorator
 
-    class vectorize(object):
-        def __init__(self, *args, **kwargs):
-            pass
+    jit = njit
 
-        def __call__(self, func):
-            wrapped_f = np.vectorize(func)
+    def vectorize(*args, **kwargs):
+        def decorator(func, *a, **kw):
+            return np.vectorize(func)
 
-            return wrapped_f
+        return decorator
 
     def generic(x, y=None):
         return None