Check indexes out of order

Author: hyanwong

tests/test_inference.py

@@ -1933,18 +1933,20 @@ def test_partial_samples(self):
                 ts2 = tsinfer.match_samples(sd, anc_ts, indexes=samples).simplify()
                 assert ts1.simplify(samples).equals(ts2, ignore_provenance=True)
 
-    def test_partial_bad_indexes(self):
-        sd = tsinfer.SampleData.from_tree_sequence(
-            msprime.simulate(
-                10, mutation_rate=2, recombination_rate=2, random_seed=233
-            ),
-            use_sites_time=False,
-        )
-        ancestors = tsinfer.generate_ancestors(sd)
-        a_ts = tsinfer.match_ancestors(sd, ancestors)
-        for bad_samples in [[], [-1, 0], [0, 10]]:
-            with pytest.raises(ValueError):
-                tsinfer.match_samples(sd, a_ts, indexes=bad_samples)
+    @pytest.mark.parametrize(
+        "bad_indexes, match",
+        [
+            ([], "at least one"),
+            ([-1, 0], "bounds"),
+            ([0, 1000], "bounds"),
+            ([1, 0], "increasing"),
+        ],
+    )
+    def test_partial_bad_indexes(self, small_sd_fixture, bad_indexes, match):
+        ancestors = tsinfer.generate_ancestors(small_sd_fixture)
+        a_ts = tsinfer.match_ancestors(small_sd_fixture, ancestors)
+        with pytest.raises(ValueError, match=match):
+            tsinfer.match_samples(small_sd_fixture, a_ts, indexes=bad_indexes)
 
     def test_time_units_default_uncalibrated(self):
         with tsinfer.SampleData(1.0) as sample_data:

tsinfer/inference.py

@@ -207,6 +207,8 @@ def check_sample_indexes(sample_data, indexes):
         raise ValueError("Must supply at least one sample to match")
     if np.any(indexes < 0) or np.any(indexes >= sample_data.num_samples):
         raise ValueError("Sample index out of bounds")
+    if np.any(indexes[:-1] >= indexes[1:]):
+        raise ValueError("Sample indexes must be in increasing order")
     return indexes
 
 
@@ -491,7 +493,7 @@ def augment_ancestors(
         :class:`tskit.TreeSequence` instance representing the inferred
         history among ancestral ancestral haplotypes.
     :param array indexes: The sample indexes to insert into the ancestors
-        tree sequence.
+        tree sequence, in increasing order.
     :param recombination_rate: Either a floating point value giving a constant rate
         :math:`\\rho` per unit length of genome, or an :class:`msprime.RateMap`
         object. This is used to calculate the probability of recombination between
@@ -591,7 +593,7 @@ def match_samples(
         ``filter_individuals`` set to False and ``keep_unary`` set to True
         (default = ``True``).
     :param array_like indexes: An array of indexes into the sample_data file of
-        the samples to match, or None for all samples.
+        the samples to match (in increasing order) or None for all samples.
     :param bool force_sample_times: After matching, should an attempt be made to
         adjust the time of "historical samples" (those associated with an individual
         having a non-zero time) such that the sample nodes in the tree sequence