Merge pull request #291 from choderalab/no_string_raises

No string raises

pymbar/confidenceintervals.py

@@ -223,7 +223,7 @@ def generateConfidenceIntervals(replicates, K):
                     print(replicate['error'])
                     print("destimated")
                     print(replicate['destimated'])
-                    raise "isnan"
+                    raise ArithmaticError("Encountered isnan in computation")
                 else:
                     if abs(replicate['error']) <= alpha * replicate['destimated']:
                         a += 1.0
@@ -238,7 +238,7 @@ def generateConfidenceIntervals(replicates, K):
                         print(replicate['error'])
                         print("destimated")
                         print(replicate['destimated'])
-                        raise "isnan"
+                        raise ArithmaticError("Encountered isnan in computation")
                     else:
                         if abs(replicate['error'][i]) <= alpha * replicate['destimated'][i]:
                             a += 1.0
@@ -254,7 +254,7 @@ def generateConfidenceIntervals(replicates, K):
                             print(replicate['error'])
                             print("ij_estimated")
                             print(replicate['destimated'])
-                            raise "isnan"
+                            raise ArithmaticError("Encountered isnan in computation")
                         else:
                             if abs(replicate['error'][i, j]) <= alpha * replicate['destimated'][i, j]:
                                 a += 1.0

pymbar/mbar.py

@@ -41,7 +41,7 @@
 import numpy as np
 import numpy.linalg as linalg
 from pymbar import mbar_solvers
-from pymbar.utils import kln_to_kn, kn_to_n, ParameterError, logsumexp, check_w_normalized
+from pymbar.utils import kln_to_kn, kn_to_n, ParameterError, DataError, logsumexp, check_w_normalized
 
 DEFAULT_SOLVER_PROTOCOL = mbar_solvers.DEFAULT_SOLVER_PROTOCOL
 DEFAULT_SUBSAMPLING_PROTOCOL = mbar_solvers.DEFAULT_SUBSAMPLING_PROTOCOL
@@ -1129,7 +1129,7 @@ def computePerturbedFreeEnergies(self, u_ln, compute_uncertainty=True, uncertain
 
         # Check dimensions.
         if (N < self.N):
-            raise "There seems to be too few samples in u_kn. You must evaluate at the new potential with all of the samples used originally."
+            raise DataError("There seems to be too few samples in u_kn. You must evaluate at the new potential with all of the samples used originally.")
 
         state_list = np.arange(L)   # need to get it into the correct shape
         A_in = np.array([0])
@@ -1357,7 +1357,7 @@ def computePMF(self, u_n, bin_n, nbins, uncertainties='from-lowest', pmf_referen
 
             # Sanity check.
             if (len(indices) == 0):
-                raise "WARNING: bin %d has no samples -- all bins must have at least one sample." % i
+                raise DataError("WARNING: bin %d has no samples -- all bins must have at least one sample." % i)
 
             # Compute dimensionless free energy of occupying state i.
             f_i[i] = - logsumexp(log_w_n[indices])
@@ -1441,7 +1441,7 @@ def computePMF(self, u_n, bin_n, nbins, uncertainties='from-lowest', pmf_referen
             return (f_i, df_i)
 
         else:
-            raise "Uncertainty method '%s' not recognized." % uncertainties
+            raise ParameterError("Uncertainty method '%s' not recognized." % uncertainties)
 
 
     #=========================================================================

pymbar/old_mbar.py

@@ -42,7 +42,7 @@
 import math
 import numpy as np
 import numpy.linalg as linalg
-from pymbar.utils import _logsum, kln_to_kn, kn_to_n, ParameterError
+from pymbar.utils import _logsum, kln_to_kn, kn_to_n, ParameterError, DataError
 
 #=========================================================================
 # MBAR class definition
@@ -1147,7 +1147,7 @@ def computePerturbedFreeEnergies(self, u_ln, compute_uncertainty=True, uncertain
 
         # Check dimensions.
         if (N < self.N):
-            raise "There seems to be too few samples in u_kn. You must evaluate at the new potential with all of the samples used originally."
+            raise DataError("There seems to be too few samples in u_kn. You must evaluate at the new potential with all of the samples used originally.")
 
         # Retrieve N and K for convenience.
         N = self.N
@@ -1543,7 +1543,7 @@ def computePMF(self, u_n, bin_n, nbins, uncertainties='from-lowest', pmf_referen
             return (f_i, df_i)
 
         else:
-            raise "Uncertainty method '%s' not recognized." % uncertainties
+            raise ParameterError("Uncertainty method '%s' not recognized." % uncertainties)
 
         return
 
@@ -1604,7 +1604,7 @@ def computePMF_states(self, u_n, bin_n, nbins):
 
             # Sanity check.
             if (len(indices) == 0):
-                raise "WARNING: bin %d has no samples -- all bins must have at least one sample." % i
+                raise DataError("WARNING: bin %d has no samples -- all bins must have at least one sample." % i)
 
             # Compute dimensionless free energy of occupying state i.
             f_i[i] = - _logsum(log_w_n[indices])
@@ -1754,8 +1754,8 @@ def _pseudoinverse(self, A, tol=1.0e-10):
         # Get size
         [M, N] = A.shape
         if N != M:
-            raise "pseudoinverse can only be computed for square matrices: dimensions were %d x %d" % (
-                M, N)
+            raise DataError("pseudoinverse can only be computed for square matrices: dimensions were %d x %d" % (
+                M, N))
 
         # Make sure A contains no nan.
         if(np.any(np.isnan(A))):

pymbar/utils.py

@@ -400,3 +400,12 @@ class BoundsError(Exception):
 
     """
     pass
+
+
+class DataError(Exception):
+
+    """
+    Data is inconsistent.
+
+    """
+    pass