yet another gui framework for Erlang

Mats Cronqvist mats.cronqvist@REDACTED
Wed Aug 17 16:05:51 CEST 2005

   as usual, there's much talk about GUI stuff for Erlang. here's my attempt to 
add to the confusion.

   i've found that it is difficult to get people to use non-GUI tools,
no matter how good they might be. also, it is immensely boring and
time-consuming to create GUIs in code (not to mention creating your
own widgets).
   i tried all the Erlang GUI frameworks i could find, and the one i
liked the best was erlgtk. chiefly because one could use glade, GTK's
eminent GUI builder tool. also, GTK looks good and is well
documented. however, erlgtk is no longer maintained.
   so maybe i could write my own with reasonable effort. i wanted these

* GUI separated from application by message passing
* distributed (application and GUI can run on different machines)
* centered around a GUI builder
* small volume of (hand-written) code
* pre-documented

   the result, gtkNode, is a c-node that expects messages in the form;
   where F is the name of a C function, and Args is an argument
list. (the point of making this a list is to allow for bulking of
messages). if there is a function F linked inte the c-node, it will be
called with a pointer to Args. F must unpack Args (which is in the
external format) and verify the number and types of the
arguments. after finishing, F writes a return value in the external
format to the return message from the c-node. the marshalling and
type-checking is done by calls to library functions using ei (Erlangs
external interface library) and GLib (a general-purpose libarary used
be e.g. GTK). the return message looks like this;
   the c-node can also send signals;

   the point of all this is that was quite simple to auto-generate
wrapper functions for most GTK functions (from the header files). it
was also simple to get the c-node to instantite a GUI from a glade
   so one can design the interface i glade; ione can easily create
a very complicated one in a few hours. all the widgets are named, the
names are available on the Erlang side as atoms. one also specifies
which events should result in signals.
   after the c-node is started and the glade file is loaded, it's just
a matter of writing an Erlang program that receives signals and send

   for example, if the user selects a menu item named 'connect', the
Erlang program receives this;

   if the Erlang program wants to grey out that menu item, it sends this;

   code should be available from jungerl as of today (although it hasn't
appeared in the CVS web interface yet).


More information about the erlang-questions mailing list