Re: Super Nested Threads

[Peter Bloomfield <peterbloomfield bellsouth net>]

On 11/22/2004 07:00:36 AM, Miquel van Smoorenburg wrote:
[ snip ]
> What some clients and web forums do is that after the first
> level of indenting, a message doesn't indent if it has only
> one reply.
>
> Like
>
>   1 message
>   2   reply to 1
>   3     reply to 2
>   4       reply to 3
>   5       reply to 4
>   6     reply to 2
>   7       reply to 6
>   8       reply to 7
>   9     reply to 2
>
> Mike.

Hi Mike!

Thanks for pointing that out--I have a version working, and it  
controls thread-explosion pretty well[1].  I'm attaching a patch  
for anyone who'd like to try it.

We'd want to leave JWZ threading as is, for the purists (well, it  
isn't really JWZ's proposal, but it's close), so I guess we'd  
need to add a threading option--perhaps "JWZ (flatter)"?

While we're on the topic--how about "JWZ (references only)" for a  
version that leaves out the lunatic "gathering by subject"?  I'm  
for ever losing messages with subjects like "Re: Today's meeting"  
because they get threaded under decades-old messages with the  
parent subject!!!

Regards,

Peter

[1] Of course, nothing will control a really convoluted thread  
except a hard limit!
Index: libbalsa/mailbox_local.c
===================================================================
RCS file: /cvs/gnome/balsa/libbalsa/mailbox_local.c,v
retrieving revision 1.120
diff -u -r1.120 mailbox_local.c
--- libbalsa/mailbox_local.c	26 Oct 2004 10:48:36 -0000	1.120
+++ libbalsa/mailbox_local.c	23 Nov 2004 00:21:27 -0000
@@ -626,6 +626,7 @@
 static gboolean lbml_prune(GNode * node, ThreadingInfo * ti);
 static void lbml_subject_gather(GNode * node, ThreadingInfo * ti);
 static void lbml_subject_merge(GNode * node, ThreadingInfo * ti);
+static gboolean lbml_flatten(GNode * node, ThreadingInfo * ti);
 static const gchar *lbml_chop_re(const gchar * str);
 #ifdef MAKE_EMPTY_CONTAINER_FOR_MISSING_PARENT
 static void lbml_clear_empty(GNode * msg_tree);
@@ -667,6 +668,9 @@
     g_node_children_foreach(msg_tree, G_TRAVERSE_ALL,
 			    (GNodeForeachFunc) lbml_subject_merge, &ti);
 
+    g_node_traverse(msg_tree, G_PRE_ORDER, G_TRAVERSE_ALL, -1,
+		    (GNodeTraverseFunc) lbml_flatten, &ti);
+
 #ifdef MAKE_EMPTY_CONTAINER_FOR_MISSING_PARENT
     lbml_clear_empty(msg_tree);
 #endif				/* MAKE_EMPTY_CONTAINER_FOR_MISSING_PARENT */
@@ -1091,6 +1095,27 @@
 	libbalsa_mailbox_unlink_and_prepend(ti->mailbox, new_node, node2);
 #endif				/* MAKE_EMPTY_CONTAINER_FOR_MISSING_PARENT */
     }
+}
+
+static gboolean
+lbml_flatten(GNode * node, ThreadingInfo * ti)
+{
+    GNode *parent = node->parent;
+    GNode *child;
+
+    if (!parent || G_NODE_IS_ROOT(parent) || node->prev || node->next)
+        return FALSE;
+
+    for (child = node; child->children && !child->children->next;
+         child = child->children)
+        /* Nothing */ ;
+    while (child != node) {
+        GNode *next = child->parent;
+        libbalsa_mailbox_unlink_and_prepend(ti->mailbox, child, parent);
+        child = next;
+    }
+
+    return FALSE;
 }
 
 /* The more heuristics should be added. */