Clojure compile-time escape mechanism - debugging

The language Forth offers a "compile-time" escape mechanism where code can be executed immediately, while the compiler is running (not at run-time). You can include print statements, for example, to debug tricky syntax or type errors).
Does Clojure have anything similar? I am getting a compile-time IllegalArgumentException in one of my function calls and would like to add a compile-time print statement to determine the argument type ((.getClass)).
Thanks.
UPDATE: Here is the complete defn that is failing compilation:
(ns my.ns.name
(:gen-class
:main true)
(:use
[clojure.contrib.str-utils2 :only (join)])
(:import
[java.io PrintWriter]
[java.net URL]
[java.util.concurrent Executors]
[java.util.jar Manifest]
[org.apache.commons.cli CommandLine HelpFormatter Options Option ParseException PosixParser]))
(defn set-version
"Set the version variable to the build number."
[]
(def version
(-> (str "jar:" (.. my.ns.name (getProtectionDomain)
(getCodeSource)
(getLocation))
"!/META-INF/MANIFEST.MF")
(URL.)
(.openStream)
(Manifest.)
(.. getMainAttributes)
(.getValue "Build-number"))))
This defn works:
(defn set-version
"Set the version variable to the build number."
[]
(println (str (.getClass my.ns.name)))
(def version
(-> (str "jar:" (-> my.ns.name (.getProtectionDomain)
(.getCodeSource)
(.getLocation))
"!/META-INF/MANIFEST.MF")
(URL.)
(.openStream)
(Manifest.)
(.. getMainAttributes)
(.getValue "Build-number"))))

Printing the class of things during compilation is pretty restricted to special cases. You will mostly get Symbols and Seqs. Only literals have a meaningful type during compilation. You can execute arbitrary code during compilation via macros.
(defmacro debug-type
[x]
(println (type x))
x)
However as I said: this will normally not be very helpful. And no: in general you cannot wrap x in an eval, eg. if x is a symbol refering to a let-local.
EDIT: Update for updated question.
(def version
(-> (str "jar:" (-> *ns* (.getProtectionDomain)
(.getCodeSource)
(.getLocation))
"!/META-INF/MANIFEST.MF")
(URL.)
(.openStream)
(Manifest.)
(.getMainAttributes)
(.getValue "Build-number")))
Try this. There is no need for a function. def inside defn should ring alarm bells.

Related

How to show error message from eval in Scheme?

I'm trying to create code that evaluates expression and return error as string for error:
(cond-expand
(gambit)
(gauche)
(kawa)
(guile
(import (rnrs base)
(rnrs exceptions)
(rnrs conditions))
(define (error-object-message cond)
(condition-message cond))))
(define (evaluate expr env)
(call-with-current-continuation
(lambda (exit)
(with-exception-handler
(lambda (e)
(exit (error-object-message e)))
(lambda ()
(eval expr env))))))
;; trigger error
(display (evaluate 'xxx (interaction-environment)))
(newline)
I've got
Guile message Unbound variable: ~S how to get actual error message and not a template?
Kawa exception: Argument #1 'unbound location: xxx' to 'error-object-message' has wrong type (gnu.mapping.UnboundLocationException) (gnu.mapping.UnboundLocationException cannot be cast to kawa.lang.NamedException)
Gauche core dump
Gambit freezes
NOTE: this is part of REPL that I'm testing in all Scheme implementations that I have on my system. It almost work, it can run itself, but I would like to show proper error message when exception happen, instead of exiting the REPL.
The reason you get an infinite loop with Gambit is that the variable xxx is unbound so the exception handler (lambda (e) (exit (error-object-message e))) is called with an unbound-global-exception object and this causes error-object-message to be called but the parameter is not an error-object (which is specific to the exceptions raised by a call to the error procedure) so this raises a type-exception object that causes the same exception handler to be called, and so on forever.
If you want an exception handling that "pops" the current exception handler, use with-exception-catcher instead of with-exception-handler. This will avoid the infinite loop.
Converting an exception object to a string can be done in Gambit this way:
(define (exception->string exc)
(with-output-to-string
(lambda ()
(display-exception exc))))
It works for error-objects and other kinds of exceptions, and also any non-exception object.
This is a more portable solution (superficially tested):
(import (scheme base)
(scheme r5rs))
(cond-expand
((or lips kawa gauche)
(define (exception->string exc)
(error-object-message exc)))
(gambit
(define (exception->string exc)
(with-output-to-string
(lambda ()
(display-exception exc)))))
(guile
(import (rnrs base)
(rnrs exceptions)
(rnrs conditions))
(define (exception->string exc)
(condition-message exc))))
(define (evaluate expr env)
(call-with-current-continuation
(lambda (exit)
(with-exception-handler
(lambda (e)
(exit (exception->string e)))
(lambda ()
(eval expr env))))))
(display (evaluate 'xxx (interaction-environment)))
(newline)
Gauche core dump
Oops. It's not ideal, for sure, but it can be explained.
By default, Gauche's with-exception-handler is SRFI-18's, not R7RS's, because of a historical reason. That means the exception handler is called in the dynamic environment where the exception occurs, including exception handler settings. If an exception occurs within the exception handler, the same exception handler is invoked, causing unbounded recursion. Apparently Gauche's runtime eats up C stack or something.
error-object-message is not defined in Gauche's default namespace. So that triggers an exception in the first place.
Add
(import (scheme base) (scheme write) (scheme r5rs))
at the beginning of code makes the program run in R7RS bindings. Then you'll get:
unbound variable: xxx
Actually, your code is not a valid R7RS program (which should begin with at least one import declaration), so anything can happen, depending on the default interpretation of the non-conforming code in the implementation.
[Edit] IMHO, with-exception-handler should be considered as the lowest level construct on which easy-to-use utilities are built, and thus should be used with extra care. In general use case, guard provides a good abstraction.
For Kawa:
(define (exception->string exc)
(call-with-port (open-output-string)
(lambda (port)
(display exc port)
(get-output-string port)))))
This will convert exception to string and get error message

Elisp (void-variable\]) when trying to evaluate a function in property list

In my Spacemacs configuration, I configure my org layer to scale any latex it generates like so
(org :variables
org-format-latex-options '(:foreground "#90ee90" :background default :scale 2.0
:html-foreground default
:html-background "Transparent" :html-scale 1 :matchers
("begin" "$1" "$" "$$" "\\(" "\\["))
I use this configuration on multiple machines, and I like different scaling's for different displays, so I wrote a little function
(defun switch-scale ()
(cond
((equal (system-name) "WMachine") 5.0)
(t 2.0) ;; default
)
)
And rewrote the above code to call the function in the :scale property, like so
org-format-latex-options '(:foreground "#90ee90" :background default :scale (switch-scale)
...
When I test switch-scale in a scratch buffer, it works fine (returns 5.0), but when I add it to my configuration it triggers the following error when trying to generate latex in org mode
Debugger entered--Lisp error: (void-variable \])
eval(\] nil)
elisp--eval-last-sexp(nil)
eval-last-sexp(nil)
funcall-interactively(eval-last-sexp nil)
call-interactively(eval-last-sexp nil nil)
command-execute(eval-last-sexp)
I'm stuck as to what's happening, it seems like (switch-scale) just isn't being evaluated???
Answer was I don't understand elisps evaluation system ... so because it was in a quoted list everything inside that list was literal, so I actually needed to use a backquote to quote the list and then use a comma to evaluate the (switch-scale) function, as explained here https://www.gnu.org/software/emacs/manual/html_node/elisp/Backquote.html

edebug-trace ceased to exist

In Emacs 25.2, suddenly the variable edebug-trace ceased to exist. When I set it with setq, it has no effect (the trace buffer does not appear). What could have happened and how can I fix it?
In the meanwhile, is there another way to know which function gets called when I click on an org-mode link
You could use trace.el to trace all org functions like so (I suggest not evaluating this until you're ready to click the link).
(mapatoms
(lambda (sym)
(and (fboundp sym)
(string-prefix-p "org-" (symbol-name sym))
(trace-function-foreground sym))))
Afterwards, you can remove the traces with:
M-x untrace-all RET
Edit: We could also convert that into a command ala elp-instrument-package:
(defun my-trace-package (prefix)
"Trace all functions which start with PREFIX.
For example, to trace all ELP functions, do the following:
\\[my-trace-package] RET elp- RET"
(interactive ;; derived from `elp-instrument-package'.
(list (completing-read "Prefix of package to trace: "
obarray 'my-traceable-p)))
(if (zerop (length prefix))
(error "Tracing all Emacs functions would render Emacs unusable"))
(mapc (lambda (name)
(trace-function-foreground (intern name)))
(all-completions prefix obarray 'my-traceable-p))
(message "Use %s to cease tracing."
(substitute-command-keys "\\[untrace-all]")))
(defun my-traceable-p (fun)
"Predicate for `my-trace-package'."
(or (functionp fun) (macrop fun)))

js/console.log in ClojureScript

I want to implement a function with ClojureScript to simplify js/console.log like this:
(defn log [& args]
(apply js/console.log args))
Calling it : (log "foo" "bar")
throws: TypeError: Illegal invocation
but this works : (js/console.log "foo" "bar")
What is the problem ?
js/something is for accessing a js object, but you shouldnt nest dots after that, since it is not clojure compatible syntax and it was going to be removed. In older versions of the compiler (2138) your code works, but it is possible that it has been deprecated in newer versions of the compiler. Which version are you using?
The correct way would be using straightforward js interop like this: [Caveat: see comment below from David Nolen, ClojureScript lead developer]
(defn log [& args] (apply (.-log js/console) args))
And even shorter, since console.log is already variadic (just make an alias):
(def log (.-log js/console))
You can also just use println if you first put this at top of your file: (enable-console-print!).
And pprint has been ported:
:require [cljs.pprint :refer [pprint]]
I found the actual answer
(.apply (.-log js/console) js/console (clj->js args))
Here is a working code for your function (tested with [org.clojure/clojurescript "1.7.228"]):
; working
(defn log [& args]
(.apply js/console.log js/console (to-array args)))
; not working
; (defn log [& args] (apply (.-log js/console) args))
; (defn log [& args] (apply js/console.log args))
; (def log js/console.log)
Here is an article that describes why (apply...) is not playing well with JS functions.
http://clojurescriptmadeeasy.com/blog/how-to-apply-with-the-console-api.html
With console.log makes sense to use a macro instead of a function. If you implement log as a function all the messages will be logged with the line number of where your log function is defined.
A macro solves this problem by generating inline code during compilation, it's important to understand that macros run at compile time.
Define this macro in macros.cljc:
(ns test.macros)
(defmacro log
[& msgs]
`(.log js/console ~#msgs))
`: is like ' or quote but:
Symbols are auto-resolved to include their namespace, preventing ambiguous symbols at the time of evaluation.
Evaluated forms can be inserted using ~ or unquote, as i did for msgs adding # to unpack multiple arguments: ~#msgs. More info about syntax quote.
Then call it from core.cljs:
(ns test.core
(:require-macros [test.macros :refer [log]]))
(log "foo" "bar")

Elisp/texi2dvi: How to call texi2dvi from Emacs?

I try to write a function based on the code from: Latex, Emacs: automatically open *TeX Help* buffer on error and close it after correction of the error?
I would like to replace latexmk by texi2dvi, but TeX-master-file does not contain the file ending .tex (which seems to be required for texi2dvi). I found out that one can add .tex by using TeX-master-file t. However, I can't make it work (I'm not an elisp programmer). Here is what I tried:
;; texi2dvi
(defun run-texi2dvi ()
(interactive)
(let ((TeX-save-query nil)
(TeX-process-asynchronous nil)
(master-file (expand-file-name (TeX-master-file t)))); append `.tex`
(TeX-save-document "")
(TeX-run-TeX "texi2dvi"
(TeX-command-expand "PDFLATEX='pdflatex -synctex=1' texi2dvi -p %s" 'TeX-master-file)
master-file)
(if (plist-get TeX-error-report-switches (intern master-file))
(TeX-next-error t)
(progn
(demolish-tex-help)
(minibuffer-message "texi2dvi: done.")))))
No clue if there's a better way to do it, but this version should work. Basically, TeX-command-expand was given the function TeX-master-file as a symbol which was called internally, and there it was called without the I-want-the-extension argument. The replacing lambda forces that.
(defun run-texi2dvi ()
(interactive)
(let ((TeX-save-query nil)
(TeX-process-asynchronous nil)
(master-file (expand-file-name (TeX-master-file t)))); append `.tex`
(TeX-save-document "")
(TeX-run-TeX "texi2dvi"
(TeX-command-expand
"PDFLATEX='pdflatex -synctex=1' texi2dvi -p %s"
(lambda (ext-ignored nondir)
(TeX-master-file t nondir)))
master-file)
(if (plist-get TeX-error-report-switches (intern master-file))
(TeX-next-error t)
(progn
(demolish-tex-help)
(minibuffer-message "texi2dvi: done.")))))
See here for a more detailed description of the issue and a simple workaround: https://tex.stackexchange.com/questions/67244/how-to-set-up-texi2dvi-with-synctex-and-error-handling/67384#67384

Resources