freenode/#sicl - IRC Chatlog
Search
7:41:40
beach
We still need conclusions. Also some preparation for the bootstrapping procedure (creation of E0), and a few figures.
7:42:40
beach
If you have the time, you should tell me whether you think this description is understandable.
7:46:38
beach
While it was not a goal in itself, it just so happened that 3 of the current 5 references are to My own papers. There will be more references of course, like for Common Lisp, CLOS, SBCL, etc.
7:47:38
beach
The bibliography is a breeze these days. Go to dl.acm.org. Find the paper entry, click on "BibTeX" entry, download, copy to paper. Done.
7:49:03
beach
And we have two months until the ELS deadline. Should be doable, especially since getting this far only took three days or so. :)
7:56:37
makomo
beach: that's great, hopefully we'll see it at ELS! :-) so i guess you've made this paper your priority now (i remember you mentioned you had a few other ideas like method combination and something else)?
7:57:16
beach
Oh the method combination stuff has been done and it works. I am planning a joint paper with Didier Verna about that.
7:59:23
beach
I decided to work on this paper because I finished phase 4 of the bootstrapping procedure, and everything seems to work as I hoped.
8:01:07
beach
I had an idea though. The generic functions in environment E4 need to be satiated. It is unfortunate to do that with a slot writer, because it does not specialize on the class of the object being written to the slot, only on the class that defines the slot.
8:01:31
beach
So satiation will stick in every class in the entire system in the call history for the unspecialized parameter.
8:02:26
beach
What if the :TYPE option of a slot that mentions a class could result in a slot writer method that specialized on both its parameters?
8:03:48
beach
Then for things like (SETF GENERIC-FUNCTION-METHODS) there would be a specialization to the class LIST, and I would avoid having entries for FLOAT, FIXNUM, COMPLEX, SYMBOL, STREAM, you name it, in the call history.
8:10:24
trittweiler
"p.3 end of section 4.1: "Thus, in this sense the word generic function is not a generic function in the sens of the host Common Lisp implementation. " typo 'sens' (also I am not sure if a comma is required after "in this sense")
8:12:00
trittweiler
I don't think I am able to follow that paragraph and the preceding one (about /class/), at least in my first reading. (FWIW)
8:13:57
beach
There are too many "sense" in there anyway. I'll think about better ways of expressing this idea.
14:36:16
beach
Why does SICL have a (SETF ACCESSOR-METHOD-SLOT-DEFINITION) function? The MOP doesn't mention any slot writer as far as I can tell.
14:39:00
beach
OK, I think I figured out how to avoid satiation altogether. Therefore, this idea of having the :TYPE slot option influence the method specializer is probably no longer interesting, at least not for the initial reason I thought about it.
14:42:12
beach
There seems to be around 80 or so functions (generic or not) in the MOP. SICL defines a few more, of course, and there are the ones from the Common Lisp HyperSpec, like make-instance, initialize-instance, shared-initialize, etc. etc.
14:46:15
beach
I think I need to go thorough each and every one of those to see in which environment it ends up, which other functions call it, and which other functions are called by it.
14:58:08
shka_
so you could build queries like: please give me all X where X is a class that happens to be superclass of Y where Y is something that beach created