Ah yes. We agree. I presume that Zack is concerned with the normal circumstance of running a racket program from the command line, but I recognize that you could do lots of creative things that will behave properly. <div><br> </div><div>Jay<span></span></div><div><br>On Thursday, February 14, 2013, Robby Findler wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr">That is also what I meant by code (well, I was speaking generally about all bytecode). <div> <br></div><div>The point I'm making is that this is not a limitation of our runtime system, it is just because you are imagining there are still references to that bytecode.</div> <div><br></div><div>In the specific instance you mention below, the namespace holding onto that module could become garbage. This happens in DrRacket a lot, for example.</div></div><div><br><br><div> On Thu, Feb 14, 2013 at 7:45 AM, Jay McCarthy <span dir="ltr"><<a>jay.mccarthy@gmail.com</a>></span> wrote:<br><blockquote style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"> When I say "code", I mean module top-level binding byte-code. There's<br> no way for Racket to remove<br> <br> (define (f x) (lambda (y) (printf "Gigantic string: ~a" (+ x y))))<br> <br> at the top-level. The byte-code will always contain the gigantic<br> string. If you ever call f and get back something, that closure could<br> be collected, but not the source that generated it.<br> <span><font color="#888888"><br> Jay<br> </font></span><div><div><br> On Thu, Feb 14, 2013 at 6:37 AM, Robby Findler<br> <<a>robby@eecs.northwestern.edu</a>> wrote:<br> > I'm pretty sure code is collected in (regular) Racket. Small numbers aren't<br> > collected, true, but they also don't really take any space. Big numbers are<br> > collected, too. (I'm not sure about interned symbols.)<br> ><br> > Robby<br> ><br> ><br> > On Thu, Feb 14, 2013 at 7:26 AM, Jay McCarthy <<a>jay.mccarthy@gmail.com</a>><br> > wrote:<br> >><br> >> In the Web server language, all anonymous functions are turned into<br> >> top-level structures masquerading as functions where the fields are<br> >> their free variables. For instance:<br> >><br> >> (define (f x) (lambda (y) (+ x y))<br> >><br> >> becomes, roughly,<br> >><br> >> (struct a-lambda42 (x)<br> >> #:property prop:procedure<br> >> (lambda (the-structure y)<br> >> (define x (a-lambda42-x the-structure))<br> >> (+ x y)))<br> >> (define (f x) (make-a-lambda42 x))<br> >><br> >> In your program, the expression<br> >><br> >> (web-cell-shadow wc-anonymous (λ (x) 'more-distinctive-text))<br> >><br> >> becomes<br> >><br> >> (struct a-lambda43 () #:property prop:procedure (lambda (s x)<br> >> 'more-distinctive-text))<br> >> (web-cell-shadow wc-anonymous (make-a-lambda43))<br> >><br> >> Notice that the anonymous function has no fields because there are no<br> >> free variables. The distinctive text is not shown anywhere in the URL<br> >> because it is part of the code, not part of the data, of the closure.<br> >><br> >> If you change the program slightly to<br> >><br> >> (web-cell-shadow wc-anonymous<br> >> (let () (define y 'more-distinctive-text)<br> >> (λ (x) y)))<br> >><br> >> then y is a free variable, so when you look at the URL:<br> >><br> >><br> >> <a href="http://localhost:8000/;((%22c%22%20.%20%220((3)%204%20(((lib%20%5C%22web-server%2Flang%2Fabort-resume.rkt%5C%22)%20.%20%5C%22lifted.6%5C%22)%20((lib%20%5C%22web-server%2Flang%2Fweb-cells.rkt%5C%22)%20.%20deserialize-info:frame-v0)%20(%23%5C%22%2Ftmp%2Ft.rkt%5C%22%20.%20%5C%22lifted.1683%5C%22)%20(2%20.%20%5C%22lifted.1991%5C%22))%200%20()%20()%20(0%20(1%20(h%20-%20()%20(lifted.1485-0%20.%20this-is-a-song-about-Japan)%20(lifted.1521-1%202%20(b!%20.%20more-distinctive-text))))%20(c%20(v!%20(3)%20%23f%20%23f))))%22))" target="_blank">http://localhost:8000/;((%22c%22%20.%20%220((3)%204%20(((lib%20%5C%22web-server%2Flang%2Fabort-resume.rkt%5C%22)%20.%20%5C%22lifted.6%5C%22)%20((lib%20%5C%22web-server%2Flang%2Fweb-cells.rkt%5C%22)%20.%20deserialize-info:frame-v0)%20(%23%5C%22%2Ftmp%2Ft.rkt%5C%22%20.%20%5C%22lifted.1683%5C%22)%20(2%20.%20%5C%22lifted.1991%5C%22))%200%20()%20()%20(0%20(1%20(h%20-%20()%20(lifted.1485-0%20.%20this-is-a-song-about-Japan)%20(lifted.1521-1%202%20(b!%20.%20more-distinctive-text))))%20(c%20(v!%20(3)%20%23f%20%23f))))%22))</a><br> >><br> >> You see the distinctive text.<br> >><br> >> As far as GC goes, code is not generally collected in Racket anyways.<br> >> Any data that is an intrinsic part of the program text (such as<br> >> literal symbols and numbers) will always be around in your program<br> >> whether you use #lang web-server or not.<br> >><br> >> Jay<br> >><br> >><br> >> On Wed, Feb 13, 2013 at 8:34 PM, Galler <<a>lzgaller@optonline.net</a>> wrote:<br> >> > Good evening,<br> >> ><br> >> > I've been investigating lang web-server, performance and garbage<br> >> > collection<br> >> ><br> >> > I have a question </div></div></blockquote></div></div></blockquote></div><br><br>-- <br>Jay McCarthy <<a href="mailto:jay@cs.byu.edu" target="_blank">jay@cs.byu.edu</a>><br>Assistant Professor / Brigham Young University<br> <a href="http://faculty.cs.byu.edu/~jay" target="_blank">http://faculty.cs.byu.edu/~jay</a><br><br>"The glory of God is Intelligence" - D&C 93<br>