Fix has_overlap/get_overlap for flipped (reverse-strand) coordinates (2.0.6)

CHANGELOG.md

@@ -6,6 +6,10 @@
 * planned new feature: during import of long reads, (optionally) correct for short exon alignment issues.
 * separate new read import and classification of isoforms.
 
+## [2.0.6]
+
+* fixed: `has_overlap`/`get_overlap` gave wrong results for reverse-strand features, where genomic coordinates are sometimes passed as `(end, start)` instead of `(start, end)`; this caused `add_domains_to_table` to silently miss overlapping domains on the reverse strand (#23)
+
 ## [2.0.5]
 
 * fixed: gtf/gff3 import no longer reads the file twice for the progress bar (was pre-counting all lines); progress is now tracked by bytes read, roughly halving import time with `progress_bar=True` (#37)

VERSION.txt

@@ -1 +1 @@
-2.0.5
+2.0.6

src/isotools/_utils.py

@@ -276,14 +276,19 @@ def find_orfs(sequence, start_codons=None, stop_codons=None, ref_cds=None):
 
 def has_overlap(r1, r2):
     "check the overlap of two intervals"
-    # assuming start < end
-    return r1[1] > r2[0] and r2[1] > r1[0]
+    # coordinates on reverse-strand features are sometimes passed as (end, start);
+    # normalize rather than assume start < end. Only the first two elements
+    # are used -- r1/r2 may be e.g. an intervaltree.Interval with extra data
+    r1_lo, r1_hi = min(r1[0], r1[1]), max(r1[0], r1[1])
+    r2_lo, r2_hi = min(r2[0], r2[1]), max(r2[0], r2[1])
+    return r1_hi > r2_lo and r2_hi > r1_lo
 
 
 def get_overlap(r1, r2):
     "check the overlap of two intervals"
-    # assuming start < end
-    return max(0, min(r1[1], r2[1]) - max(r1[0], r2[0]))
+    r1_lo, r1_hi = min(r1[0], r1[1]), max(r1[0], r1[1])
+    r2_lo, r2_hi = min(r2[0], r2[1]), max(r2[0], r2[1])
+    return max(0, min(r1_hi, r2_hi) - max(r1_lo, r2_lo))
 
 
 def get_intersects(tr1, tr2):

tests/README.md

@@ -27,3 +27,13 @@ Two different test styles exist in this directory:
 
 **When adding a test for a specific bug or feature, use style 1.** Reserve
 the pipeline chain for genuine end-to-end integration concerns.
+
+## File organization
+
+One test file per feature area, named after what it covers -- not
+necessarily 1:1 with a source file, but close (`domain_test.py` for
+`domains.py`, `splice_graph_test.py` for `splice_graph.py`, `orf_test.py`
+for ORF-prediction logic in `gene.py`, `data_import_test.py` for import
+functions in `_transcriptome_io.py`, `utils_test.py` for the general
+primitives in `_utils.py`, etc.). Before creating a new file, check whether
+an existing one already covers that area.

tests/utils_test.py

@@ -0,0 +1,18 @@
+from isotools._utils import has_overlap, get_overlap
+
+
+def test_has_overlap_flipped_coordinates():
+    # regression test for #23: coordinates on reverse-strand features are
+    # sometimes passed as (end, start) rather than (start, end) -- both
+    # orderings must give the same result
+    assert has_overlap((90, 118), (100, 200)) is True
+    assert has_overlap((118, 90), (100, 200)) is True  # previously wrongly False
+    assert has_overlap((10, 20), (100, 200)) is False
+    assert has_overlap((20, 10), (100, 200)) is False
+
+
+def test_get_overlap_flipped_coordinates():
+    assert get_overlap((90, 118), (100, 200)) == 18
+    assert get_overlap((118, 90), (100, 200)) == 18  # previously wrong result
+    assert get_overlap((10, 20), (100, 200)) == 0
+    assert get_overlap((20, 10), (100, 200)) == 0