Re: Source code as metadata
Jeff Weiss jeffrey.m.we...@gmail.com writes: From browsing git, it looks like the project.clj version hasn't been incremented in 7 months, and the fix for closures came in after that. If you're using serializable.fn from a maven repo, it is out of date, AFAICT. Sorry about that; just pushed a new 1.1.2 version with the latest code from Github. -Phil -- 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
Re: Source code as metadata
On 31 Mar 2012, at 00:00, Cedric Greevey wrote: On Fri, Mar 30, 2012 at 6:17 PM, Lee Spector lspec...@hampshire.edu wrote: On Mar 30, 2012, at 5:11 PM, Cedric Greevey wrote: That opens a giant can of worms. How, for example, do we discover that (partial * 2) and #(* % 2) and (fn [x] (* 2 x)) and #(+ %1 %1) are all equal? Nevermind once we get into situations like #(reduce + (map (constantly 1) %) equals #(loop [n 0 s (seq %)] (if s (recur (inc n) (next s)) n) equals count. In fact that's an uncomputable can of worms. I was pretty sure it smelled like halting problem; thanks for confirming. :) Agreed, but I'm not suggesting determining whether two forms are computationally equivalent, just if two forms are literally equivalent. I know it wouldn't cover all cases, but cases like: (= (partial * 2) (partial * 2)) it would cover, which would seem like an improvement over the situation today. I suppose my bigger point is if code is data when thy isn't the code for functions available for analysis at runtime? -- 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 -- 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
Re: Source code as metadata
I believe the latest code does capture closures properly. I haven't tested all kinds of crazy corner cases, but it does work for all my closures. From browsing git, it looks like the project.clj version hasn't been incremented in 7 months, and the fix for closures came in after that. If you're using serializable.fn from a maven repo, it is out of date, AFAICT. -jeff On Friday, March 30, 2012 3:07:53 PM UTC-4, Phil Hagelberg wrote: Nathan Matthews nathan.r.matth...@gmail.com writes: I wanted to serialise functions and send them over the network. The problem with serializable-fn is that it doesn't capture closures. It's designed to capture closures; if it doesn't that would be an (unsurprising) bug. -Phil -- 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
Re: Source code as metadata
Another idea is to put :indentation metadata on vars, so user-defined macros could be indented properly. Currently I have (define-clojure-indent ...) with a number of forms in my emacs config file, and it seems to be pretty common solution. It'd be nice to replace this hack with an IDE-independent way to control the indentation. Phil, what do you think? Could it be experimentally implemented in clojure-mode? -- 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
Re: Source code as metadata
Vinzent ru.vinz...@gmail.com writes: Phil, what do you think? Could it be experimentally implemented in clojure-mode? No, clojure-mode determines indentation exclusively from static heuristics. There is dynamic indentation support in slime, but I've never looked into it; I'm not sure how I feel about indentation rules changing depending on whether slime is active or not. -Phil -- 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
Re: Source code as metadata
Hi, I wanted to serialise functions and send them over the network. The problem with serializable-fn is that it doesn't capture closures. I wrote some code which re-programmed the fn macro to capture the closures as well as the actual function form, and attach them as meta-data also on the actual function object. I could share this code if people would find this useful. Its non-trivial due to various complexities and bugs in Clojure. Also it bothers me that (= (partial * 2) (partial * 2)) is false. Logically it shouldn't be right? If we captured the function forms, that would enable better equality for functions. On 29 Mar 2012, at 17:44, Phil Hagelberg wrote: On Thu, Mar 29, 2012 at 9:29 AM, Petr Gladkikh petrg...@gmail.com wrote: I am pondering on the idea of having more (or even a lot) of metadata that could be useful for debugging and problem resolution. Since we can store anything in metadata, can we store not only source file path and line number but whole source code that is associated with piece of code? See https://github.com/technomancy/serializable-fn/ for a proof-of-concept of this idea. It's always bothered be that defn puts metadata on the var and not on the function itself. Anyway, supposedly Rich is in favour of having dynamicity knobs according to his Conj 2011 keynote, and this sounds like just the kind of thing that would fall under that. -Phil -- 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 -- 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
Re: Source code as metadata
Nathan Matthews nathan.r.matth...@gmail.com writes: I wanted to serialise functions and send them over the network. The problem with serializable-fn is that it doesn't capture closures. It's designed to capture closures; if it doesn't that would be an (unsurprising) bug. -Phil -- 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
Re: Source code as metadata
I'm not sure how I feel about indentation rules changing depending on whether slime is active or not. What I was thinking, is that there'd be some function which would collect and save indentation metadata, so it can be used later. Thus, active slime connection required only the first time one uses a library with non-standart indentation rules. -- 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
Re: Source code as metadata
Nathan Matthews nathan.r.matth...@gmail.com writes: I wrote some code which re-programmed the fn macro to capture the closures as well as the actual function form, and attach them as meta-data also on the actual function object. Could you submit it as a patch to serializable-fn? It would be nice to have everything in one place. -Phil -- 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
Re: Source code as metadata
Vinzent ru.vinz...@gmail.com writes: I'm not sure how I feel about indentation rules changing depending on whether slime is active or not. What I was thinking, is that there'd be some function which would collect and save indentation metadata, so it can be used later. Thus, active slime connection required only the first time one uses a library with non-standart indentation rules. Currently clojure-mode operates on each file in isolation. As soon as you start remembering macro indentation, you're back in a non-deterministic state where indentation depends on the history of which files you've visited in the current Emacs session. You could make it work just for the macros in a single file, but that doesn't seem very valuable. -Phil -- 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
Re: Source code as metadata
On Fri, Mar 30, 2012 at 1:26 PM, Vinzent ru.vinz...@gmail.com wrote: Another idea is to put :indentation metadata on vars, so user-defined macros could be indented properly. Currently I have (define-clojure-indent ...) with a number of forms in my emacs config file, and it seems to be pretty common solution. It'd be nice to replace this hack with an IDE-independent way to control the indentation. Phil, what do you think? Could it be experimentally implemented in clojure-mode? My thought on this? Just look for the macro to have a rest arg. Function argument indentation should be used for non-rest macro args, and body indentation for macro rest args. -- 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
Re: Source code as metadata
On Fri, Mar 30, 2012 at 2:48 PM, Nathan Matthews nathan.r.matth...@gmail.com wrote: Also it bothers me that (= (partial * 2) (partial * 2)) is false. Logically it shouldn't be right? If we captured the function forms, that would enable better equality for functions. That opens a giant can of worms. How, for example, do we discover that (partial * 2) and #(* % 2) and (fn [x] (* 2 x)) and #(+ %1 %1) are all equal? Nevermind once we get into situations like #(reduce + (map (constantly 1) %) equals #(loop [n 0 s (seq %)] (if s (recur (inc n) (next s)) n) equals count. -- 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
Re: Source code as metadata
Probably you slightly misunderstood what I mean. Consider this scenario:
I've set up a project which uses a new library with non-standart indent.
I've connected to swank and compiled it. Then I'm calling some
clojure-mode-update-indent function, which walks through all loaded
namespaces and collects :indent metadata (on the clojure side) and saves
appropriate define-clojure-indent form somewhere in config file (on the
elisp side). After that I have all macros indented properly.
It's better than manually writing define-clojure-indent form because it's
tool-agnostic; say, CCW can use the same data, so library maintainers only
have to add {:indent :defn} to their defmacros.
суббота, 31 марта 2012 г. 2:32:35 UTC+6 пользователь Phil Hagelberg написал:
Vinzent ru.vinz...@gmail.com writes:
I'm not sure how I feel about indentation rules
changing depending on whether slime is active or not.
What I was thinking, is that there'd be some function which would
collect and save indentation metadata, so it can be used later. Thus,
active slime connection required only the first time one uses a
library with non-standart indentation rules.
Currently clojure-mode operates on each file in isolation. As soon as
you start remembering macro indentation, you're back in a
non-deterministic state where indentation depends on the history of
which files you've visited in the current Emacs session.
You could make it work just for the macros in a single file, but that
doesn't seem very valuable.
-Phil
--
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
Re: Source code as metadata
Counter-example: one could write if-authenticated macro, which will take fixed number of args, but should be indented as normal if. суббота, 31 марта 2012 г. 3:07:23 UTC+6 пользователь Cedric Greevey написал: On Fri, Mar 30, 2012 at 1:26 PM, Vinzent ru.vinz...@gmail.com wrote: Another idea is to put :indentation metadata on vars, so user-defined macros could be indented properly. Currently I have (define-clojure-indent ...) with a number of forms in my emacs config file, and it seems to be pretty common solution. It'd be nice to replace this hack with an IDE-independent way to control the indentation. Phil, what do you think? Could it be experimentally implemented in clojure-mode? My thought on this? Just look for the macro to have a rest arg. Function argument indentation should be used for non-rest macro args, and body indentation for macro rest args. -- 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
Re: Source code as metadata
Vinzent ru.vinz...@gmail.com writes: Probably you slightly misunderstood what I mean. Consider this scenario: I've set up a project which uses a new library with non-standart indent. I've connected to swank and compiled it. Then I'm calling some clojure-mode-update-indent function, which walks through all loaded namespaces and collects :indent metadata (on the clojure side) and saves appropriate define-clojure-indent form somewhere in config file (on the elisp side). After that I have all macros indented properly. I see. Yes, that would be better. I don't think it belongs in clojure-mode, but you could make a tool that could collect that data and export it into .dir-locals.el files or something like that. -Phil -- 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
Re: Source code as metadata
On Mar 30, 2012, at 5:11 PM, Cedric Greevey wrote: That opens a giant can of worms. How, for example, do we discover that (partial * 2) and #(* % 2) and (fn [x] (* 2 x)) and #(+ %1 %1) are all equal? Nevermind once we get into situations like #(reduce + (map (constantly 1) %) equals #(loop [n 0 s (seq %)] (if s (recur (inc n) (next s)) n) equals count. In fact that's an uncomputable can of worms. But detecting identity of functions that were defined using exactly the same code is computable, whether or not it's advisable or useful (which I don't know). -Lee -- 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
Re: Source code as metadata
2012/3/30 Vinzent ru.vinz...@gmail.com: Counter-example: one could write if-authenticated macro, which will take fixed number of args, but should be indented as normal if. OK, check the macro structure to see if any args are incorporated as invokable forms -- so, in arguments in special forms and macros that are bodies. (Yes, this is obviously sort of recursive.) The first such arg is treated as the start of a body. Some static dataflow analysis may be needed in case the macro takes the arg, processes it in some way, and then inserts it into its output rather than inserting it directly. -- 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
Re: Source code as metadata
On Fri, Mar 30, 2012 at 6:17 PM, Lee Spector lspec...@hampshire.edu wrote: On Mar 30, 2012, at 5:11 PM, Cedric Greevey wrote: That opens a giant can of worms. How, for example, do we discover that (partial * 2) and #(* % 2) and (fn [x] (* 2 x)) and #(+ %1 %1) are all equal? Nevermind once we get into situations like #(reduce + (map (constantly 1) %) equals #(loop [n 0 s (seq %)] (if s (recur (inc n) (next s)) n) equals count. In fact that's an uncomputable can of worms. I was pretty sure it smelled like halting problem; thanks for confirming. :) -- 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
Re: Source code as metadata
On Thu, Mar 29, 2012 at 9:29 AM, Petr Gladkikh petrg...@gmail.com wrote: I am pondering on the idea of having more (or even a lot) of metadata that could be useful for debugging and problem resolution. Since we can store anything in metadata, can we store not only source file path and line number but whole source code that is associated with piece of code? See https://github.com/technomancy/serializable-fn/ for a proof-of-concept of this idea. It's always bothered be that defn puts metadata on the var and not on the function itself. Anyway, supposedly Rich is in favour of having dynamicity knobs according to his Conj 2011 keynote, and this sounds like just the kind of thing that would fall under that. -Phil -- 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
