Re: Would *default-precision* make sense?

[Stephen C. Gilardi]

On Jun 29, 2009, at 10:11 AM, Nicolas Oury wrote:

I am not sure, but I believe it's due to *warn-on-reflection* being

true).

When I looked, the REPL was called within a macro 'with-bindings repl'
that expands to
(binding [....
          *warn-on-reflection* ....
        ]
        repl)


allowing the execution of REPL to set! *war-on-reflection* and some
other similar vars. (*ns*, *compile-files*, *compile-path*, etc...)

I suppose the compiler does a similar thing, but I have never looked.

bound by the compiler/REPL before evaluating (set! *warn-on- reflection*

That's correct. You can see the vars that are set!-ale in the REPL in:

user=> (doc clojure.main/with-bindings)
-------------------------
clojure.main/with-bindings
([& body])
Macro

  that often need to be set!: *ns* *warn-on-reflection* *print-meta*
  *print-length* *print-level* *compile-path* *command-line-args* *1
  *2 *3 *e
nil



  user=> (binding [*math-context* *math-context*] (clojure.main/repl))
  user=> (set! *math-context* (java.math.MathContext. 4))
  #<MathContext precision=4 roundingMode=HALF_UP>
  user=> (+ 3 1.55555555M)
  4.556M
  user=> (set! *math-context* (java.math.MathContext. 6))
  #<MathContext precision=6 roundingMode=HALF_UP>
  user=> (+ 3 1.55555555M)
  4.55556M
  user=>



Does anyone have any objection to or support for that idea?

--Steve

Executes body in the context of thread-local bindings for several vars In your own REPL, you could make *math-context* set!-able by launching a new repl wrapped in code that binds *math-context*: It seems to me that clojure.main/with-bindings should be enhanced provide a root binding for *math-context* so it's set!-able at the repl.

smime.p7s
Description: S/MIME cryptographic signature