tag:blogger.com,1999:blog-2080885971644496896.post1513383412261723218..comments2009-05-24T21:20:02.916-04:00Jens Axel Søgaardhttp://www.blogger.com/profile/15211030864341077735noreply@blogger.comBlogger4125tag:blogger.com,1999:blog-2080885971644496896.post-82639490407616085662009-05-24T21:20:02.916-04:002009-05-24T21:20:02.916-04:00derSlom: Yes, it is a demonstration of how to do ER-style for people who like that for the listed reason, and as a demonstration that syntax values don't require black magic, it's not a complete library. Expanding it into a library will require adding a hash table to (1) unwrap each syntax value once and get the cost loss back, and (2) to associate each unwrapped list with the lost syntactic information (which, I think, will also solve the `#%app' problem). BTW, unlike your code, I'm talking about a persistent hash, allowing you to add information to existing syntax values.<br /><br />Sam: that sounds right, but I explicitly avoided that too. I do think, though, that it's possible to express ER using something like the above -- even with #%app.Eli Barzilayhttp://www.blogger.com/profile/08165322672792335619noreply@blogger.comtag:blogger.com,1999:blog-2080885971644496896.post-19286645249946495292009-05-24T20:20:37.804-04:002009-05-24T20:20:37.804-04:00Although I don't know of any formal proofs of this, I don't believe that either ER or syntactic closures can express syntax case, nor can syntax case (in the original formulation) express ER. <br /><br />Also, ER really doesn't have a clear specification, so it's hard to make sure that an ER implementation that matches the original note works with actual code.Sam THhttp://www.blogger.com/profile/12049111326816937908noreply@blogger.comtag:blogger.com,1999:blog-2080885971644496896.post-75511064431468205172009-05-24T15:29:45.968-04:002009-05-24T15:29:45.968-04:00to also add useful something besides nitpicking, I have been down that road as well: http://paste.lisp.org/display/41754derSlomhttp://www.blogger.com/profile/12160110480520538395noreply@blogger.comtag:blogger.com,1999:blog-2080885971644496896.post-14449101056175685322009-05-24T14:51:10.373-04:002009-05-24T14:51:10.373-04:00turing tarpit? next corner right and there you are ;)<br /><br />As you are quite aware, your "solution" changes the cost of macro-expansion to quadric. This is in constrast to linear cost for native syntax-case AND native ER-macros. As it is quite possible to implement syntax-case in terms of ER-transformers (or syntactic-closures) without loosing the linear complexity, many people claim that these two are more basic than syntax-case ... and therefore should be in the actual standard.<br /><br />Whats your opinion eli?derSlomhttp://www.blogger.com/profile/12160110480520538395noreply@blogger.com