Fix up merge issues and disable test.

Author: jeromekelleher

lib/tsinfer.h

@@ -1,5 +1,5 @@
 /*
-** Copyright (C) 2018 University of Oxford
+** Copyright (C) 2018-2020 University of Oxford
 **
 ** This file is part of tsinfer.
 **
@@ -27,6 +27,9 @@
 #include "object_heap.h"
 #include "avl.h"
 
+/* TODO remove this when we update tskit version. */
+#define TSK_MISSING_DATA (-1)
+
 /* NULL_LIKELIHOOD represents a compressed path and NONZERO_ROOT_LIKELIHOOD
  * marks a node that is not in the current tree. */
 #define NULL_LIKELIHOOD (-1)

tests/test_formats.py

@@ -1183,7 +1183,6 @@ def test_missing_data(self):
             self.assertEqual(i, 1)
 
 
-
 class TestAncestorData(unittest.TestCase, DataContainerMixin):
     """
     Test cases for the sample data file format.

tests/test_inference.py

@@ -820,10 +820,10 @@ def test_all_missing_at_adjacent_site(self):
         self.assertTrue(len(site_0_anc) == 1)
         site_0_anc = site_0_anc[0]
         # Sites 0 and 2 should share the same ancestor
-        assert np.all(adc.ancestors_focal_sites[:][site_0_anc] == [0, 2])
+        self.assertTrue(np.all(adc.ancestors_focal_sites[:][site_0_anc] == [0, 2]))
         focal_site_0_haplotype = adc.ancestors_haplotype[:][site_0_anc]
         # Sites with all missing data should default to 0
-        assert np.all(focal_site_0_haplotype == expected_hap_focal_site_0)
+        self.assertTrue(np.all(focal_site_0_haplotype == expected_hap_focal_site_0))
 
     def test_with_recombination_long_threads(self):
         ts = msprime.simulate(
@@ -2358,6 +2358,7 @@ def test_recombination_with_dist_noninference_intermediate(self):
         self.verify_single_recombination_position([0.0, 0.9, 2.0], G, 1)
 
 
+@unittest.skip("Missing data not yet implemented")
 class TestMissingDataImputed(unittest.TestCase):
     """
     Test that sites with tskit.MISSING_DATA are imputed, using both the PY and C engines

tsinfer/formats.py

@@ -1268,24 +1268,18 @@ def add_site(
 
         if alleles is None:
             alleles = ["0", "1"]
-        else:
-            # Don't modify the input parameter
-            alleles = list(alleles)
         n_alleles = len(alleles)
+        non_missing = genotypes != tskit.MISSING_DATA
         if len(set(alleles)) != n_alleles:
             raise ValueError("Alleles must be distinct")
         if np.any(genotypes == tskit.MISSING_DATA) and alleles[-1] is not None:
-            alleles.append(None)
-        if np.any(genotypes >= n_alleles):
-            raise ValueError(
-                "Non-missing genotype values must be between 0 and len(alleles) - 1")
+            # Don't modify the input parameter
+            alleles = list(alleles) + [None]
         if np.any(np.logical_and(genotypes < 0, genotypes != tskit.MISSING_DATA)):
             raise ValueError("Non-missing values for genotypes cannot be negative")
         if genotypes.shape != (self.num_samples,):
             raise ValueError(
                 "Must have {} (num_samples) genotypes.".format(self.num_samples))
-        if np.any(genotypes[non_missing] < 0):
-            raise ValueError("Non-missing values for genotypes cannot be negative")
         if np.any(genotypes[non_missing] >= n_alleles):
             raise ValueError("Non-missing values for genotypes must be < num alleles")
         if position < 0: