<div class="gmail_quote">On Thu, Jun 18, 2009 at 11:43 AM, Michael Lucas-Smith <span dir="ltr"><<a href="mailto:mlucas-smith@cincom.com">mlucas-smith@cincom.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;">
<div class="im">Julian Fitzell wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
Why not? Isn't that a bug? (I'm only half joking :) )<br>
<br>
According to the ANSI spec, though, #allSubclasses need not necessarily return on OrderedCollection (in fact, why should it be ordered?). Maybe that's too theoretical an issue to worry about but if you're going to force a conversion of one of the collections, it's a bit odd not to ensure the type of the other one.<br>
<br>
If this is coming up now, why don't we at least define what kind of collection #potentialParents should return (I don't think it needs to be ordered?), write a unit test for it, and make sure it always returns that. It might be better to do "allSubclasses asSet"?<br>
<br>
</blockquote></div>
I didn't really think about it this way. You can't "concatenate" two sets together - you can merge them, but it pretty much means that #, doesn't exist on Set in VW. I published a new version that makes no attempt to settle the dispute of OrderedCollection vs Set one way or another. Instead, it uses #addAll: which should be safe across Smalltalks.<br>
</blockquote><div><br></div><div>Ah yes, good point. I never really thought about #, being "concatenate" rather than just "combine". #addAll: should be fine unless, of course, #allSubclasses returns an Array on some platform :) we can probably cross that bridge if we come to it... ;)</div>
<div><br></div><div>Julian</div></div>