X-Git-Url: https://git.donarmstrong.com/?a=blobdiff_plain;f=sam_header.c;h=a1b518101ad20aefb297c240a68756da8eebc371;hb=607d7a0ab79152be4153fa09861cfa889488f2c7;hp=a119c02b9c4cfaa0e6c4251454fd86cd12db3743;hpb=87d9ea7a580d0d1a58b8f0dca3ba3528ca118e46;p=samtools.git

diff --git a/sam_header.c b/sam_header.c
index a119c02..a1b5181 100644
--- a/sam_header.c
+++ b/sam_header.c
@@ -10,6 +10,7 @@ KHASH_MAP_INIT_STR(str, const char *)
 
 struct _HeaderList
 {
+    struct _HeaderList *last;   // Hack: Used and maintained only by list_append_to_end. Maintained in the root node only.
     struct _HeaderList *next;
     void *data;
 };
@@ -37,7 +38,7 @@ const char *o_sq_tags[] = {"AS","M5","UR","SP",NULL};
 const char *r_sq_tags[] = {"SN","LN",NULL};
 const char *u_sq_tags[] = {"SN",NULL};
 
-const char *o_rg_tags[] = {"LB","DS","PU","PI","CN","DT","PL",NULL};
+const char *o_rg_tags[] = {"CN","DS","DT","FO","KS","LB","PG","PI","PL","PU","SM",NULL};
 const char *r_rg_tags[] = {"ID",NULL};
 const char *u_rg_tags[] = {"ID",NULL};
 
@@ -58,6 +59,34 @@ static void debug(const char *format, ...)
     va_end(ap);
 }
 
+#if 0
+// Replaced by list_append_to_end
+static list_t *list_prepend(list_t *root, void *data)
+{
+    list_t *l = malloc(sizeof(list_t));
+    l->next = root;
+    l->data = data;
+    return l;
+}
+#endif
+
+// Relies on the root->last being correct. Do not use with the other list_*
+//  routines unless they are fixed to modify root->last as well.
+static list_t *list_append_to_end(list_t *root, void *data)
+{
+    list_t *l = malloc(sizeof(list_t));
+    l->last = l;
+    l->next = NULL;
+    l->data = data;
+
+    if ( !root )
+        return l;
+
+    root->last->next = l;
+    root->last = l;
+    return root;
+}
+
 static list_t *list_append(list_t *root, void *data)
 {
     list_t *l = root;
@@ -322,7 +351,7 @@ static HeaderLine *sam_header_line_parse(const char *headerLine)
 
     while (*to && *to!='\t') to++;
     if ( to-from != 2 ) {
-		debug("[sam_header_line_parse] expected '@XY', got [%s]\n", headerLine);
+		debug("[sam_header_line_parse] expected '@XY', got [%s]\nHint: The header tags must be tab-separated.\n", headerLine);
 		return 0;
 	}
     
@@ -345,7 +374,11 @@ static HeaderLine *sam_header_line_parse(const char *headerLine)
         while (*to && *to!='\t') to++;
 
         if ( !required_tags[itype] && !optional_tags[itype] )
+        {
+            // CO is a special case, it can contain anything, including tabs
+            if ( *to ) { to++; continue; }
             tag = new_tag("  ",from,to-1);
+        }
         else
             tag = new_tag(from,from+3,to-1);
 
@@ -401,8 +434,14 @@ static int sam_header_line_validate(HeaderLine *hline)
         tag = tags->data;
         if ( !tag_exists(tag->key,required_tags[itype]) && !tag_exists(tag->key,optional_tags[itype]) )
         {
-            debug("Unknown tag [%c%c] for [%c%c].\n", tag->key[0],tag->key[1], hline->type[0],hline->type[1]);
-            return 0;
+            // Lower case tags are user-defined values.
+            if( !(islower(tag->key[0]) || islower(tag->key[1])) )
+            {
+                // Neither is lower case, but tag was not recognized.
+                debug("Unknown tag [%c%c] for [%c%c].\n", tag->key[0],tag->key[1], hline->type[0],hline->type[1]);
+                // return 0; // Even unknown tags are allowed - for forward compatibility with new attributes
+            }
+            // else - allow user defined tag
         }
         tags = tags->next;
     }
@@ -530,6 +569,7 @@ void *sam_header_parse2(const char *headerText)
     const char *text;
     char *buf=NULL;
     size_t nbuf = 0;
+	int tovalidate = 0;
 
     if ( !headerText )
 		return 0;
@@ -538,8 +578,9 @@ void *sam_header_parse2(const char *headerText)
     while ( (text=nextline(&buf, &nbuf, text)) )
     {
         hline = sam_header_line_parse(buf);
-        if ( hline && sam_header_line_validate(hline) )
-            hlines = list_append(hlines, hline);
+        if ( hline && (!tovalidate || sam_header_line_validate(hline)) )
+            // With too many (~250,000) reference sequences the header parsing was too slow with list_append.
+            hlines = list_append_to_end(hlines, hline);
         else
         {
 			if (hline) sam_header_line_free(hline);
@@ -628,6 +669,36 @@ char **sam_header2list(const void *_dict, char type[2], char key_tag[2], int *_n
     return ret;
 }
 
+void *sam_header2key_val(void *iter, const char type[2], const char key_tag[2], const char value_tag[2], const char **_key, const char **_value)
+{
+    list_t *l = iter;
+    if ( !l ) return NULL;
+
+    while (l)
+    {
+        HeaderLine *hline = l->data;
+        if ( hline->type[0]!=type[0] || hline->type[1]!=type[1] )
+        {
+            l = l->next;
+            continue;
+        }
+
+        HeaderTag *key, *value;
+        key   = header_line_has_tag(hline,key_tag);
+        value = header_line_has_tag(hline,value_tag);
+        if ( !key && !value ) 
+        {
+            l = l->next;
+            continue;
+        }
+
+        *_key = key->value;
+        *_value = value->value;
+        return l->next;
+    }
+    return l;
+}
+
 const char *sam_tbl_get(void *h, const char *key)
 {
 	khash_t(str) *tbl = (khash_t(str)*)h;