This is the mail archive of the kawa@sourceware.org mailing list for the Kawa project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: strange int->boolean behavior


On 12/18/2014 10:19 AM, Jamison Hope wrote:
Sometimes, a primitive 0 is false:

#|kawa:1|# (->boolean (->int 3))
Exception in thread "main" java.lang.Error: unsupported CodeAttr.emitConvert
	at gnu.bytecode.CodeAttr.emitConvert(CodeAttr.java:2006)
	[...]

#|kawa:1|# (if (->int 3) "YES" "NO")
YES
#|kawa:2|# (if (->int 0) "YES" "NO")
YES
#|kawa:3|# (define (f x::int) ::string (if x "YES" "NO"))
#|kawa:4|# (f 3)
YES
#|kawa:5|# (f 0)
NO
#|kawa:6|# (disassemble f)
In class atInteractiveLevel$3

Method name:"f" public static Signature: (int)java.lang.CharSequence
Attribute "Code", length:70, max_stack:1, max_locals:1, code_length:14
   0: iload_0
   1: ifeq 10
   4: getstatic <Field atInteractiveLevel$3.Lit0 java.lang.String>
   7: goto 13
  10: getstatic <Field atInteractiveLevel$3.Lit1 java.lang.String>
  13: areturn

I get similar results with the other primitive types.  If I'm following things correctly,
it's ConditionalTarget#compileFromStack which is treating zero as false?

Yes. I checked in a fix for this.  The change in ConditionalTarget depends on the language;
specifically whether language.isTrue(Integer.valueOf(0)).  If so, we an just
emitGoto the true label.

I forgot about or overlooking this message, but found it while searching for
another discussion (which turn out to be "proposed changes to handling of false
and end-of-list").

(I'm looking into doing at least part of that previous proposal - specifically
treating #!null as false.)
--
	--Per Bothner
per@bothner.com   http://per.bothner.com/


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]