Remove code that supported optional numba installation

quantecon/lss.py

@@ -9,10 +9,9 @@
 import numpy as np
 from numpy.random import multivariate_normal
 from scipy.linalg import solve
+from numba import jit
 
-#-Check if Numba is Available-#
-from .util import numba_installed, jit
-
+@jit
 def simulate_linear_model(A, x0, v, ts_length):
     """
     This is a separate function for simulating a vector linear system of 
@@ -55,8 +54,6 @@ def simulate_linear_model(A, x0, v, ts_length):
                 x[i, t+1] += A[i, j] * x[j, t]   #Dot Product
     return x
 
-if numba_installed:
-    simulate_linear_model = jit(simulate_linear_model)
 
 class LinearStateSpace(object):
     """

quantecon/markov/gth_solve.py

@@ -10,15 +10,7 @@
 import numpy as np
 from numba import jit
 
-from ..util import numba_installed, jit
-if not numba_installed:
-    try:
-        xrange
-    except:  # python3
-        xrange = range
-
-
-def gth_solve(A, overwrite=False):
+def gth_solve(A, overwrite=False, use_jit=True):
     r"""
     This routine computes the stationary distribution of an irreducible
     Markov transition matrix (stochastic matrix) or transition rate
@@ -74,13 +66,13 @@ def gth_solve(A, overwrite=False):
     n = A1.shape[0]
     x = np.zeros(n)
 
-    if numba_installed:
+    if use_jit:
         _gth_solve_jit(A1, x)
         return x
 
-    # if not numba_installed
+    # if not using jit
     # === Reduction === #
-    for k in xrange(n-1):
+    for k in range(n-1):
         scale = np.sum(A1[k, k+1:n])
         if scale <= 0:
             # There is one (and only one) recurrent class contained in
@@ -94,7 +86,7 @@ def gth_solve(A, overwrite=False):
 
     # === Backward substitution === #
     x[n-1] = 1
-    for k in xrange(n-2, -1, -1):
+    for k in range(n-2, -1, -1):
         x[k] = np.dot(x[k+1:n], A1[k+1:n, k])
 
     # === Normalization === #
@@ -103,46 +95,45 @@ def gth_solve(A, overwrite=False):
     return x
 
 
-if numba_installed:
-    @jit(nopython=True)
-    def _gth_solve_jit(A, out):
-        """
-        JIT complied version of the main routine of gth_solve.
-
-        Parameters
-        ----------
-        A : numpy.ndarray(float, ndim=2)
-            Stochastic matrix or generator matrix. Must be of shape n x n.
-            Data will be overwritten.
-
-        out : numpy.ndarray(float, ndim=1)
-            Output array in which to place the stationary distribution of A.
-
-        """
-        n = A.shape[0]
-
-        # === Reduction === #
-        for k in range(n-1):
-            scale = np.sum(A[k, k+1:n])
-            if scale <= 0:
-                # There is one (and only one) recurrent class contained in
-                # {0, ..., k};
-                # compute the solution associated with that recurrent class.
-                n = k+1
-                break
-            for i in range(k+1, n):
-                A[i, k] /= scale
-
-                for j in range(k+1, n):
-                    A[i, j] += A[i, k] * A[k, j]
-
-        # === Backward substitution === #
-        out[n-1] = 1
-        for k in range(n-2, -1, -1):
-            for i in range(k+1, n):
-                out[k] += out[i] * A[i, k]
-
-        # === Normalization === #
-        norm = np.sum(out)
-        for k in range(n):
-            out[k] /= norm
+@jit(nopython=True)
+def _gth_solve_jit(A, out):
+    """
+    JIT complied version of the main routine of gth_solve.
+
+    Parameters
+    ----------
+    A : numpy.ndarray(float, ndim=2)
+        Stochastic matrix or generator matrix. Must be of shape n x n.
+        Data will be overwritten.
+
+    out : numpy.ndarray(float, ndim=1)
+        Output array in which to place the stationary distribution of A.
+
+    """
+    n = A.shape[0]
+
+    # === Reduction === #
+    for k in range(n-1):
+        scale = np.sum(A[k, k+1:n])
+        if scale <= 0:
+            # There is one (and only one) recurrent class contained in
+            # {0, ..., k};
+            # compute the solution associated with that recurrent class.
+            n = k+1
+            break
+        for i in range(k+1, n):
+            A[i, k] /= scale
+
+            for j in range(k+1, n):
+                A[i, j] += A[i, k] * A[k, j]
+
+    # === Backward substitution === #
+    out[n-1] = 1
+    for k in range(n-2, -1, -1):
+        for i in range(k+1, n):
+            out[k] += out[i] * A[i, k]
+
+    # === Normalization === #
+    norm = np.sum(out)
+    for k in range(n):
+        out[k] /= norm

quantecon/markov/random.py

@@ -8,10 +8,11 @@
 """
 import numpy as np
 import scipy.sparse
+from numba import jit
 
 from .core import MarkovChain
 from .ddp import DiscreteDP
-from ..util import check_random_state, numba_installed, jit
+from ..util import check_random_state
 from ..random import probvec, sample_without_replacement
 
 
@@ -218,7 +219,7 @@ def random_discrete_dp(num_states, num_actions, beta=None,
     ddp = DiscreteDP(R, Q, beta, s_indices, a_indices)
     return ddp
 
-
+@jit
 def _sa_indices(num_states, num_actions):
     L = num_states * num_actions
     s_indices = np.empty(L, dtype=int)
@@ -233,5 +234,3 @@ def _sa_indices(num_states, num_actions):
 
     return s_indices, a_indices
 
-if numba_installed:
-    _sa_indices = jit(_sa_indices)

quantecon/random/utilities.py

@@ -3,7 +3,9 @@
 """
 
 import numpy as np
-from ..util import check_random_state, numba_installed, jit
+from numba import jit
+
+from ..util import check_random_state
 
 #-Generating Arrays and Vectors-#
 
@@ -74,7 +76,7 @@ def _diff(r, out):
             out[i, j] = r[i, j] - r[i, j-1]
         out[i, n] = 1 - r[i, n-1]
 
-
+@jit
 def sample_without_replacement(n, k, num_trials=None, random_state=None):
     """
     Randomly choose k integers without replacement from 0, ..., n-1.
@@ -141,8 +143,3 @@ def sample_without_replacement(n, k, num_trials=None, random_state=None):
         return result[0]
     else:
         return result
-
-if numba_installed:
-    docs = sample_without_replacement.__doc__
-    sample_without_replacement = jit(sample_without_replacement)
-    sample_without_replacement.__doc__ = docs

quantecon/util/__init__.py

@@ -3,7 +3,6 @@
 """
 
 from .array import searchsorted
-from .external import jit, numba_installed
 from .notebooks import fetch_nb_dependencies
 from .random import check_random_state
 from .timing import tic, tac, toc

quantecon/util/array.py

@@ -9,14 +9,14 @@
 """
 
 import numpy as np
-from .external import numba_installed, jit
+from numba import jit
 
 # ----------------- #
 # -ARRAY UTILITIES- #
 # ----------------- #
 
-
-def _searchsorted(a, v):
+@jit(nopython=True)
+def searchsorted(a, v):
     """
     Custom version of np.searchsorted. Return the largest index `i` such
     that `a[i-1] <= v < a[i]` (for `i = 0`, `v < a[0]`); if `v[n-1] <=
@@ -61,11 +61,3 @@ def _searchsorted(a, v):
         else:
             lo = m
     return hi
-
-if numba_installed:
-    searchsorted = jit(nopython=True)(_searchsorted)
-else:
-    def searchsorted(a, v):
-        return np.searchsorted(a, v, side='right')
-
-searchsorted.__doc__ = _searchsorted.__doc__