You could generate a file and call 'load-file'. Then if it didn't work the way you expect you have an artifact you can debug with. I think eval is really making your life a bit more difficult and I am not following your eval'd code very well.
Chris On Jun 23, 4:12 am, Nicolas Oury <nicolas.o...@gmail.com> wrote: > Hello, > > Thank you very much for your answers. > I know generating code at runtime (evaling) is not something very > common, but that's something I need to do. > > I am writing a simulation software for a rule-based modelisation > language. > Typically, this kind of thing does millions of time the same loop, where > it applies some rules chosen by the user at runtime. > In my experiments in Common LISP (even in ABCL), there is a huge pay-off > to generate and compile some code once you know what the rules of the > simulation are. And then to run the compiled code. > (It's better to compile and run than to interpret, isn't it? Especially > in a tight loop executed millions of time.) > > So I need, to compile some expressions before I run the loop that keeps > evaluating these expressions. So there are a few solutions: > > - do it at compile-time, using macros. But that means that the end-user > needs to be more knowledgeable that he will probably be. And he needs to > have a full environment. This solution is not user friendly. And you > cannot add/remove a rule at run time. > > - use the ASM lib shipped with clojure. I could do that, but it's really > more difficult than generating clojure code. > > - Generating clojure code at runtime and have it compiled. This is what > I try to do. As I need to generate a few functions (and even multi > methods), It would be cleaner to generate a namespace and write > a program in it. Have it compiled and called from the main program. > This is what the REPL does, so it must be possible. > I agree macros help a lot for this kind of things because it simplifies > a lot the code I generate at runtime. But without some way of compiling > at runtime, macros alone can't do what I want. > > > > > I see the repl in main.clj does something similar to you in its > > with-bindings, but the reason is to make those vars able to be set!, > > not to establish the external namespace for the eval. In your eval > > string, the ns is the default 'user ns. To use a different ns, you > > would need to establish that inside the string. > > That was what I was trying to mimic. Allow *ns* to be settable in my > eval, in order to be able to create a new namespace. > > With the help of Chouser's answer, I managed to write an example that > actually compiles something at runtime: > > (defn -main [] > (binding [*ns* *ns*] > (eval '(ns Blop)) > (println *ns*) > (eval '(defn f [x] (+ x x))) > (eval '(println (f 5))) > (eval '(println (Blop/f 5))) > )) > I will try to think how encapsulate this usage pattern in a higher level > way. > > Generating a .class for the namespace would be useful too, even if less > critical. Is there a way to trick an eval into believing, it is a > compiler? (I know the trick must involve *compile-files* and > *compile-path* in some way, but I can't manage to make this work.) > > Are there other people on the mailing list compiling (calling eval) at > runtime that might be interested? > > Thank you very much for your help, > > Nicolas. --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "Clojure" group. To post to this group, send email to clojure@googlegroups.com Note that posts from new members are moderated - please be patient with your first post. To unsubscribe from this group, send email to clojure+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/clojure?hl=en -~----------~----~----~----~------~----~------~--~---
