ISSUE closed: [cgmo-webcgm] Empty string or null

[Benoit Bezaire <benoit@itedo.com>]

I haven't heard any feedback on this, so I'm going to make a proposal
and unless someone objects, I suggest we adopt it.

The good news is that I think I've found a solution (and I think its
the correct solution)... the bad news is that it's work (I think
someone needs to write an ECMAScript binding).

The main issue was: when should an implementation return null instead
of empty string?

And the answer is: it depends on the scripting language.

We've agreed that for this version of the specification, we're only
interested in an ECMAScript binding, but I don't think the spec
wording should be specific to ECMAScript (that's why we have the IDL
snippet). We may in the future want to have different bindings (VB,
C#, Java etc...)

As you know, I've followed the DOM level 2/3 specifications carefully
when making proposals for the WebCGM DOM... the DOM level 2/3
specifications often refer to "empty string" and "null" without
defining what they really mean; but the answer can be found in the
specific bindings.

When should an implementation be allowed to return empty string? Ans:
in the case of ECMAScript, when the method or attribute returns an
object of type "String".

When should an implementation be allowed to return null? Ans:
in the case of ECMAScript, when the method or attribute returns an
object that implements a specific interface.

I've research the ECMAScript language:
http://www.ecma-international.org/publications/files/ECMA-ST/Ecma-262.pdf

And found that a String and null, cannot be compared (see page 55/56)
because they are of different types. An ECMAScript interpreter will
never return 'true' when comparing a String and null. A String has to be
compared with "" (empty string).

So we need a document similar to this, but for our DOM:
http://www.w3.org/TR/2004/REC-DOM-Level-3-Core-20040407/ecma-script-binding.html

We need the following actions to be taken.
- Everyone who wrote tests need to verify that null and "" are used
when they are suppose to be used and update their tests accordingly.

- We need someone to generate an ECMAScript binding. This will likely
result in finding a tool to convert IDL to ECMAScript or writing such
a tool themselves. Any takers?

- Also, if people have interest in other bindings they should state so
now...

Regards,

-- 
 Benoit   mailto:benoit@itedo.com

 
Wednesday, May 18, 2005, 12:47:03 PM, Benoit wrote:

BB> Hi Ulrich,

BB>   In the test you have submitted, we noticed that
BB>   node-aps.htm, line 22: if( apsNode.nodeValue == "" )
BB>   and node-pict.htm, line 21: if( cgmPic.nodeValue == null )

BB>   In one case you are comparing with empty string and in the other
BB>   with null. That seems a bit odd.

BB>   I'm a bit puzzle by the null keyword in the DOM specification and
BB>   what it implies in ECMAScript, (i.e., what does it map to?), what's
BB>   your opinion? (Others can comment as well)... For interoperability
BB>   concerns, it would be wise to have our implementation have the same
BB>   mapping. Lofton, I recall you had some comments on this a while back
BB>   (something to do with undefined)...

BB>   Assuming your implementation passes both those tests... is your
BB>   implementation returning BSTRs for both calls or is one of them
BB>   returning some other data type (say a VARIANT)?

BB>   Cheers,