[gt-users] gff3 parser

[Sascha Steinbiss]

Sascha Steinbiss wrote:
> Gordon Gremme wrote:
>>> ok. i think i've resolved this:
>>> http://gist.github.com/62770
> This patch introduced some memleaks

Some word about the memory leaks:

+static void make_id_unique(GtGFF3Visitor *gff3_visitor, GtStr *id)
+{
+
+  GtCstrTable *used_ids = gff3_visitor->gt_used_ids;
+  unsigned long i = 0;
+  GtStr *id_clone;
+
+  if (gt_cstr_table_get(used_ids, gt_str_get(id))) {
+      id_clone = gt_str_clone(id);
+      while ( gt_cstr_table_get(used_ids, gt_str_get(id_clone) )) {
+        /* TODO: add warning */
+        id_clone = gt_str_clone(id);
+        make_unique_id_string(id_clone, ++i);
+      }
+      /* update table with the new id */
+      gt_str_set(id, gt_str_get(id_clone));
+  }
+  gt_cstr_table_add(used_ids, gt_str_get(id));
+
+}

Note that gt_str_clone() creates a new copy of the string which needs to 
be properly freed using gt_str_delete(). This is not easy here, as the 
only reference to the new clone is overwritten in each iteration, 
leading to leaking memory. Sometimes it is better to use a single GtStr 
as a buffer for temporary strings:

static bool id_string_is_unique(GtStr *id, GtStr *buf, GtCstrTable *tab,
                                 unsigned long i)
{
   gt_str_reset(buf);
   gt_str_append_str(buf, id);
   make_unique_id_string(buf, i);
   return (gt_cstr_table_get(tab, gt_str_get(buf)) == NULL);
}
static void make_id_unique(GtGFF3Visitor *gff3_visitor, GtStr *id)
{
   unsigned long i = 1;

   if (gt_cstr_table_get(gff3_visitor->used_ids, gt_str_get(id)))
   {
     GtStr *buf = gt_str_new();
     gt_warning("feature ID '%s' is not unique", gt_str_get(id));
     while (!id_string_is_unique(id, buf, gff3_visitor->used_ids, i))
     {
       i++;
     }
     gt_warning("make feature ID unique: %s -> %s", gt_str_get(id),
                                                    gt_str_get(buf));
     /* update table with the new id */
     gt_str_set(id, gt_str_get(buf));
     gt_str_delete(buf);
   }
   gt_cstr_table_add(gff3_visitor->used_ids, gt_str_get(id));
}

Also, you need to take care of freeing the GtCStrTable created in the 
visitor constructor.

As for the tests, the error messages of the failing tests can be found 
in the corresponding <GT_HOME>/testsuite/stest_testsuite/test* directories.

Have fun coding,
Sascha

-- 
Sascha Steinbiss
Center for Bioinformatics
University of Hamburg
Bundesstr. 43
20146 Hamburg
Germany

Email:  steinbiss at zbh.uni-hamburg.de
URL:    http://www.zbh.uni-hamburg.de/steinbiss
Phone:  +49 (40) 42838 7322
FAX:    +49 (40) 42838 7312