Merge pull request #274 from kdm9/kdm9-paf-fix

fix: invalid paf produced for some patch alignments
waveygang · Sep 24, 2024 · 73c1331 · 73c1331
2 parents 1f485f3 + dd0bbc3
commit 73c1331
Show file tree

Hide file tree

Showing 2 changed files with 28 additions and 22 deletions.
diff --git a/src/common/wflign/src/wflign_patch.cpp b/src/common/wflign/src/wflign_patch.cpp
@@ -2032,26 +2032,29 @@ query_start : query_end)
         // write how many reverse complement alignments were found
         //std::cerr << "got " << rev_patch_alns.size() << " rev patch alns" << std::endl;
         for (auto& patch_aln : multi_patch_alns) {
-            write_alignment_paf(
-                out,
-                patch_aln,
-                query_name,
-                query_total_length,
-                query_offset,
-                query_length,
-                query_is_rev,
-                target_name,
-                target_total_length,
-                target_offset,
-                target_length,
-                min_identity,
-                mashmap_estimated_identity,
-                false,  // Don't add an endline after each alignment
-                true);  // This is a reverse complement alignment
-            // write tag indicating that this is a multipatch alignment
-            out << "\t" << "pt:Z:true" << "\t"
-                // and if the patch is inverted as well
-                << "\t" << "iv:Z:" << (patch_aln.is_rev ? "true" : "false") << "\n";
+            // write_alignment_paf only writes anything if aln.ok. We need to guard the manual tag writing below with the same conditional to avoid writing an invalid PAF.
+            bool wrote = write_alignment_paf(
+                    out,
+                    patch_aln,
+                    query_name,
+                    query_total_length,
+                    query_offset,
+                    query_length,
+                    query_is_rev,
+                    target_name,
+                    target_total_length,
+                    target_offset,
+                    target_length,
+                    min_identity,
+                    mashmap_estimated_identity,
+                    false,  // Don't add an endline after each alignment
+                    true);  // This is a reverse complement alignment
+            if (wrote) {
+                // write tag indicating that this is a multipatch alignment
+                out << "\t" << "pt:Z:true" << "\t"
+                    // and if the patch is inverted as well
+                    << "iv:Z:" << (patch_aln.is_rev ? "true" : "false") << "\n";
+            }
         }
     }
     out << std::flush;
@@ -2221,7 +2224,7 @@ void write_alignment_sam(
     free(patch_cigar);
 }
 
-void write_alignment_paf(
+bool write_alignment_paf(
         std::ostream& out,
         const alignment_t& aln,
         const std::string& query_name,
@@ -2237,6 +2240,7 @@ void write_alignment_paf(
         const float& mashmap_estimated_identity,
         const bool& with_endline,
         const bool& is_rev_patch) {
+    bool ret = false;  // return true if we wrote the alignment
 
     if (aln.ok) {
         uint64_t matches = 0;
@@ -2291,9 +2295,11 @@ void write_alignment_paf(
             if (with_endline) {
                 out << std::endl;
             }
+            ret = true;
         }
         free(cigar);
     }
+    return ret;
 }
 
 double float2phred(const double& prob) {

diff --git a/src/common/wflign/src/wflign_patch.hpp b/src/common/wflign/src/wflign_patch.hpp
@@ -144,7 +144,7 @@ namespace wflign {
             const char* query,
             const char* target,
             const int64_t& target_pointer_shift);
-        void write_alignment_paf(
+        bool write_alignment_paf(
                 std::ostream& out,
                 const alignment_t& aln,
                 const std::string& query_name,