W3C home > Mailing lists > Public > public-webapps@w3.org > April to June 2009

Re: [WebIDL] double

From: Cameron McCormack <cam@mcc.id.au>
Date: Mon, 15 Jun 2009 10:41:24 +1000
To: Ian Hickson <ian@hixie.ch>
Cc: public-webapps@w3.org, public-geolocation@w3.org
Message-ID: <20090615004124.GA14034@arc.mcc.id.au>
Ian Hickson:
> Oh. Wouldn't it make more sense for JS compatiblility to have float be 
> 64-bit? It seems bad to have a type that is 32-bit but then have its most 
> common implementation not throw an exception or anything if setting a 
> number that is 64-bit into it.

When converting Number values to IDL short or long, for example, the
ToInt16 and ToInt32 operators do things other than throw exceptions, and
instead coerce the number into the right range.

> Or do UAs treat 'float's as 32-bit and trim the numbers when storing
> them?)

Seems they do store them as 32-bit, just with one quick test:

  <svg xmlns='http://www.w3.org/2000/svg'>
    <script>
      var p = document.documentElement.createSVGPoint();
      p.x = 1.23456789012;
      var y = p.x;
      alert(y);
    </script>
  </svg>

alerts 1.2345678806304932 in Opera 10 beta, a recent Firefox nightly and
a recent WebKit in Midori.

-- 
Cameron McCormack ≝ http://mcc.id.au/
Received on Monday, 15 June 2009 00:42:17 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 18:49:31 GMT