Added option --ref-seq to goalign mask #6

cmd/mask.go

@@ -12,6 +12,7 @@ var maskout string = "stdout"
 var maskstart int
 var masklength int
 var maskunique bool
+var maskrefseq string
 
 // subseqCmd represents the subseq command
 var maskCmd = &cobra.Command{
@@ -32,6 +33,10 @@ goalign mask -p -i al.phy -s 9 -l 10
 
 This will replace 10 positions with N|X from the 10th position.
 
+If --ref-seq is specified, then coordinates are considered on the given reference sequence
+without considering gaps. So far, all insertions compared to the reference sequence are also masked).
+
+
 If --unique is specified, 'goalign mask --unique' will replace characters that
 are unique in their column (except GAPS) with N or X.
 In this case, --length and --start are ignored.
@@ -51,6 +56,8 @@ The output format is the same than input format.
 			return
 		}
 
+		refseq := cmd.Flags().Changed("ref-seq")
+
 		for al := range aligns.Achan {
 			if aligns.Err != nil {
 				err = aligns.Err
@@ -62,7 +69,12 @@ The output format is the same than input format.
 					return
 				}
 			} else {
-				if err = al.Mask(maskstart, masklength); err != nil {
+				start := maskstart
+				length := masklength
+				if refseq {
+					start, length, err = al.RefCoordinates(maskrefseq, start, length)
+				}
+				if err = al.Mask(start, length); err != nil {
 					io.LogError(err)
 					return
 				}
@@ -80,5 +92,6 @@ func init() {
 	maskCmd.PersistentFlags().StringVarP(&maskout, "output", "o", "stdout", "Alignment output file")
 	maskCmd.PersistentFlags().IntVarP(&maskstart, "start", "s", 0, "Start position (0-based inclusive)")
 	maskCmd.PersistentFlags().IntVarP(&masklength, "length", "l", 10, "Length of the sub alignment")
+	maskCmd.PersistentFlags().StringVar(&maskrefseq, "ref-seq", "none", "Coordinates are considered wrt. to the given reference sequence (no effect with --unique)")
 	maskCmd.PersistentFlags().BoolVar(&maskunique, "unique", false, "If given, then masks characters that are unique in their columns (start and length are ignored)")
 }

docs/commands/mask.md

@@ -21,6 +21,9 @@ This will replace 10 positions with N|X from the 10th position.
 If --unique is specified, 'goalign mask --unique' will replace characters that
 are unique in their column with N or X. In this case, --length and --start are ignored.
 
+If --ref-seq is specified, then coordinates are considered on the given reference sequence
+without considering gaps. So far, all insertions compared to the reference sequence are also masked).
+
 The output format is the same than input format.
 
 #### Usage

test.sh

@@ -3388,6 +3388,39 @@ ${GOALIGN} mask -i input -o result -s 0 -l 2 -p
 diff -q -b expected result
 rm -f input expected result
 
+echo "->goalign mask / nucl --ref-seq"
+cat > input <<EOF
+   10   20
+Seq0000  GATTAATTTG CCGTAGGCCA
+Seq0001  GAATCTGAAG ATCGAACACT
+Seq0002  TTAAGT--TC ACTTC--ATG
+Seq0003  GAGAGGACTA GTTCATACTT
+Seq0004  TTTAAACACT TTTACATCGA
+Seq0005  TGTCGGACCT AAGTATTGAG
+Seq0006  TACAACGGTG TATTCCAGCG
+Seq0007  GTGGAGAGGT CTATTTTTCC
+Seq0008  GGTTGAAGGA CTCTAGAGCT
+Seq0009  GTAAAGGGTA TGGCCATGTG
+EOF
+
+cat > expected <<EOF
+   10   20
+Seq0000  GATTAATTNN NNNNNNNNCA
+Seq0001  GAATCTGANN NNNNNNNNCT
+Seq0002  TTAAGT--NN NNNNNNNNTG
+Seq0003  GAGAGGACNN NNNNNNNNTT
+Seq0004  TTTAAACANN NNNNNNNNGA
+Seq0005  TGTCGGACNN NNNNNNNNAG
+Seq0006  TACAACGGNN NNNNNNNNCG
+Seq0007  GTGGAGAGNN NNNNNNNNCC
+Seq0008  GGTTGAAGNN NNNNNNNNCT
+Seq0009  GTAAAGGGNN NNNNNNNNTG
+EOF
+
+${GOALIGN} mask -i input -o result -s 6 -l 8 --ref-seq Seq0002 -p
+diff -q -b expected result
+rm -f input expected result
+
 echo "->goalign mask --unique"
 cat > input <<EOF
 >A