why is goto_num_tagged_header() so slow? analysis included...

[jurriaan <thu...@xs...>]

slrn_goto_num_tagged_header calls slrn_goto_header() with the header in
Num_Tag_List.

=====
int slrn_goto_num_tagged_header (int *nump) /*{{{*/
{
   unsigned int num;
   
   num = (unsigned int) *nump;
   num--;
   
   if (num >= Num_Tag_List.len)
     return 0;
   
   if (Num_Tag_List.headers == NULL)
     return 0;
   
   if (-1 == slrn_goto_header (Num_Tag_List.headers[num], 0))
     return 0;
   
   Slrn_Full_Screen_Update = 1;
   return 1;
}
=====

slrn_goto_header starts at the first header, and goes through all of
them, finding the one that matches:


=====
int slrn_goto_header (Slrn_Header_Type *header, int read_flag) /*{{{*/
{
   Slrn_Header_Type *h = Slrn_First_Header;
   
   while ((h != NULL) && (h != header))
     h = h->real_next;

   if (h == NULL) return -1;
   
   Slrn_Current_Header = h;
   if (h->flags & HEADER_HIDDEN) slrn_uncollapse_this_thread (h, 0);
   _art_find_header_line_num ();
   
   if (read_flag) select_article (1);
   return 0;
}
=====

Why does slrn_goto_header go through all of them? Or at least, why
doesn't slrn_goto_num_tagged_header just set Slrn_Current_Header
directly? If it's in Num_Tag_List, it's an existing header and you can
just select it directly, instead of (real life example here) searching
through 1.5 million headers 30,000 times (I read news on a shell at my
providers, then save message-id's if I want to download something. In a
big binary group, these numbers are correct. Also correct is that it
took more than an hour to extract these message-ids).

I hope someone can explain why this searching is good - I'm not getting
it, actually.

If a patch is wanted, I can supply that, of course.

Kind regards,
Jurriaan