Ben, I can't comment on the workings of Dr. Geo, but since you ask about #do:, nil, least surprise etc., I can tell you that errors are your friends. Note that there is a difference between sending #do: to nil (would you really expect "nothing" to know how to iterate its contents?) and sending #do: to a collection that contains elements, some of which happen to be nil? In the latter case, would it really be non-surprising to have nil elements treated differently from their peers? Would you have them treated differently only if nil encounters a problem doing what is asked of it? In that case, #do: would have to trap and suppress errors, at least for nil entries.
All of that sounds pretty surprising to me. Eventually, it will to you too. Happy Smalltalking! Bill ________________________________________ From: [email protected] [[email protected]] On Behalf Of Ben Coman [[email protected]] Sent: Saturday, October 22, 2011 2:36 PM To: [email protected] Subject: [Pharo-project] [BUG]UndefinedObject(Object)>>doesNotUnderstand: #update I was using DrGeo for the first time from the bundled release 11.08 on Windows 7. About 30 minutes in I drew three equally intersecting circles, added a reflection line with each circle reflected. I added a free point with a middle (I think to a circle) and was trying to add a locus. When moving any of the points defining the circles I am getting a dozen or more cascading MNUs down the screen. Moving the reflection line still works okay. The error occurs in following method... DrGeo>>updateDirtyCostumes self presenter tool costumes do: [ :each | each update ]. self area changed In the debugger inspecting "self presenter tool costumes" returns nil. While I can't judge whether returning nil is valid here, as a Smalltalk newbie what I am really interested in is: 1. The semantics of sending a "do:" message to nil. By the principle of least surprise I would expect that it would do ...nothing, whereas here it tries later sending the "update" method to it. 2. Why in lines 0027 and 0641 is the "do:" method received by OrderedCollection and not UndefinedObject? Actually I can't see "do:" in the hierarchy of UndefinedObject, so why does UndefinedObject not raise a MNU for "do:" ? The image with the error is available at... http://blog.openinworld.com/miscfiles/DrGeo-updateDirtyCostumes-MNU-2011-10-22-btc.zip The stack trace generated by the <Mail out bug report> menu item is below. I wasn't sure how much of it to cut or leave, so I've left it all. ------------------8<------------------------- 00000001 22 October 2011 10:27:03 pm 00000002 00000003 VM: Win32 - IX86 - NT - Squeak3.10.2 of 11 February 2010 [latest update: #9314]
