ported pr248 for vcf4.4 - Allow colon in chromosome names

EBIvariation · Jul 8, 2024 · d535ac0 · d535ac0
1 parent 274f7c0
commit d535ac0
Show file tree

Hide file tree

Showing 8 changed files with 574 additions and 410 deletions.
diff --git a/inc/vcf/validator_detail_v44.hpp b/inc/vcf/validator_detail_v44.hpp
diff --git a/src/vcf/vcf_v44.ragel b/src/vcf/vcf_v44.ragel
@@ -63,8 +63,8 @@
 
 
     ## A contig must be a sequence name allowed by the SAM format ( regex [!-)+-<>-~][!-~]* ) excluding the characters <>[]:*
-    ## A chromosome must be a string with no white-spaces or colons, and may be surronded by < > symbols (for contigs)
-    meta_contig_char= alnum | ( punct - (':' | '<' | '>' | '[' | ']' | '*' | '=' | ',' ) ) ;
+    ## A chromosome must be a string with no white-spaces, and may be surrounded by < > symbols (for contigs)
+    meta_contig_char= alnum | ( punct - ( '<' | '>' | '[' | ']' | '*' | '=' | ',' ) ) ;
     chrom_basic     = (meta_contig_char - '#') (meta_contig_char)* ;
     chrom_contig    = '<' chrom_basic '>' ;
     chromosome      = chrom_basic | chrom_contig ;

diff --git a/test/input_files/v4.4/failed/failed_body_chrom_001.vcf b/test/input_files/v4.4/failed/failed_body_chrom_001.vcf
@@ -1,4 +1,4 @@
 ##fileformat=VCFv4.4
-##CauseOfFailure=contig ID contains a colon
+##CauseOfFailure=contig ID contains a whitespace
 #CHROM	POS	ID	REF	ALT	QUAL	FILTER	INFO	FORMAT	HG00096	HG00097
-chr:1	13302	rs180734498	C	T	100	PASS	AN=2184;AC=249;AF=0.11	GT:DS:GL	0|0:0.050:-0.13,-0.58,-3.62	0|1:1.000:-2.45,-0.00,-5.00
+chr 1	13302	rs180734498	C	T	100	PASS	AN=2184;AC=249;AF=0.11	GT:DS:GL	0|0:0.050:-0.13,-0.58,-3.62	0|1:1.000:-2.45,-0.00,-5.00
diff --git a/test/input_files/v4.4/failed/failed_body_chrom_002.vcf b/test/input_files/v4.4/failed/failed_body_chrom_002.vcf
@@ -1,4 +1,4 @@
 ##fileformat=VCFv4.4
-##CauseOfFailure=contig ID contains a whitespace
+##CauseOfFailure=contig ID contains a comma
 #CHROM	POS	ID	REF	ALT	QUAL	FILTER	INFO	FORMAT	HG00096	HG00097
-chr 1	13302	rs180734498	C	T	100	PASS	AN=2184;AC=249;AF=0.11	GT:DS:GL	0|0:0.050:-0.13,-0.58,-3.62	0|1:1.000:-2.45,-0.00,-5.00
+chr,1	13302	rs180734498	C	T	100	PASS	AN=2184;AC=249;AF=0.11	GT:DS:GL	0|0:0.050:-0.13,-0.58,-3.62	0|1:1.000:-2.45,-0.00,-5.00
diff --git a/test/input_files/v4.4/failed/failed_body_chrom_003.vcf b/test/input_files/v4.4/failed/failed_body_chrom_003.vcf
@@ -1,4 +1,4 @@
 ##fileformat=VCFv4.4
-##CauseOfFailure=contig ID contains a comma
+##CauseOfFailure=contig ID contains illegal character
 #CHROM	POS	ID	REF	ALT	QUAL	FILTER	INFO	FORMAT	HG00096	HG00097
-chr,1	13302	rs180734498	C	T	100	PASS	AN=2184;AC=249;AF=0.11	GT:DS:GL	0|0:0.050:-0.13,-0.58,-3.62	0|1:1.000:-2.45,-0.00,-5.00
+chr*1	13302	rs180734498	C	T	100	PASS	AN=2184;AC=249;AF=0.11	GT:DS:GL	0|0:0.050:-0.13,-0.58,-3.62	0|1:1.000:-2.45,-0.00,-5.00
diff --git a/test/input_files/v4.4/failed/failed_body_chrom_004.vcf b/test/input_files/v4.4/failed/failed_body_chrom_004.vcf
diff --git a/test/input_files/v4.4/passed/passed_body_chrom.vcf b/test/input_files/v4.4/passed/passed_body_chrom.vcf
@@ -5,3 +5,4 @@
 1ABC	13302	rs180734498	C	T	100	PASS	AN=2184;AC=249;AF=0.11	GT:DS:GL	0|0:0.050:-0.13,-0.58,-3.62	0|1:1.000:-2.45,-0.00,-5.00
 1.A	13302	rs180734498	C	T	100	PASS	AN=2184;AC=249;AF=0.11	GT:DS:GL	0|0:0.050:-0.13,-0.58,-3.62	0|1:1.000:-2.45,-0.00,-5.00
 <1A>	13302	rs180734498	C	T	100	PASS	AN=2184;AC=249;AF=0.11	GT:DS:GL	0|0:0.050:-0.13,-0.58,-3.62	0|1:1.000:-2.45,-0.00,-5.00
+1ABC:123	13302	rs180734498	C	T	100	PASS	AN=2184;AC=249;AF=0.11	GT:DS:GL	0|0:0.050:-0.13,-0.58,-3.62	0|1:1.000:-2.45,-0.00,-5.00
diff --git a/test/input_files/v4.4/passed/passed_meta_contig.vcf b/test/input_files/v4.4/passed/passed_meta_contig.vcf
@@ -3,6 +3,7 @@
 ##contig=<ID=1,length=123456>
 ##contig=<ID=1AC,length=123456>
 ##contig=<ID=ABcd123,length=123456>
+##contig=<ID=ABcd:123,length=123456>
 ##contig=<ID=contig_url,length=123456,URL=ftp://somewhere.org/assembly.fa>
 ##contig=<ID=contig_accession,species="Homo sapiens",accession=GCA_000001405.1>
 #CHROM	POS	ID	REF	ALT	QUAL	FILTER	INFO