Subj : Re: spidermonkey docu
To   : Jens Thiele <karme@unforgettable.com>
From : Brendan Eich <brendan@meer.net>
Date : Wed Apr 07 2004 02:44 pm

Jens Thiele wrote:
> the JS_ValueToInt32 docu is wrong:
> http://mozilla.org/js/spidermonkey/apidoc/gen/complete.html#JS_ValueToInt32
> 
> Description
> JS_ValueToInt32 converts a specified JS value, v, to a JS double, and
> then to a 32-bit integer, if it fits. If the double is out of range,
> JS_ValueToInt32 reports an error and conversion fails. The double
> rounded to the nearest integer value.

Agh, the doc is both too detailed (as if based on some misreading of 
code), and not complete (it doesn't define "out of range" fully).


> 
> the corresponding source:
> js_ValueToInt32(JSContext *cx, jsval v, int32 *ip)
> {
>     jsdouble d;
>     JSString *str;
> 
>     if (JSVAL_IS_INT(v)) {
>         *ip = JSVAL_TO_INT(v);
>         return JS_TRUE;
>     }
> [...]
> 
> => js_ValueToInt32 does IMHO the right thing ;-)
> if it is a JS int value convert to C int32
> if not try to convert it to an C int32

But the doc does not lie.  Converting to double and then to int loses no 
information, it's just less efficient than what's actually done in the 
implementation.

So I'm not sure why you say the doc is wrong.  It is not in error, it's 
just (as usual, thanks to whoever wrote it for Netscape, ages ago) a bit 
misleading about implementation, a bit too detailed in that respect, and 
oddly not detailed enough in other ways.

/be

.