Testing wheather string conversion is possible

[Ronald Fischer]

Assume the following JavaScript function:

function bracketize(s)
{
return '['+s+']';
}

This function which doesn't assume anything about its argument
except that it must be convertible to a string.

I would now like to enhance this function, so that it does not
complain an error ("cant't convert s to string") for those cases
where s happens to be not convertable to a string:

function bracketize(s)
{
return
(`s can be converted to string`)
? '['+s+']'
: null
;
}

How can I write the condition `s can be converted to string` in JavaScript?

Ronald

[Michael Winter]

On 18 Nov 2004 04:05:48 -0800, Ronald Fischer <ro*****@eml.cc > wrote:

Assume the following JavaScript function:

function bracketize(s)
{
return '['+s+']';
}

This function which doesn't assume anything about its argument
except that it must be convertible to a string.

This causes a problem because *everything* can be converted to a string.
That conversion may not be particularly useful, but it can be done
nevertheless.

If you want to know if something *is* a string, you can use the expression

'string' == typeof <variable>

which will only evaluate to true if <variable> is a string.

If you did actually mean "convertibl e to a string", you'll have to be more
specific. What characterises something that, in your opinion, cannot be
converted?

[snip]

Mike

--
Michael Winter
Replace ".invalid" with ".uk" to reply by e-mail.

[Duncan Booth]

Ronald Fischer wrote:

function bracketize(s)
{
return
(`s can be converted to string`)
? '['+s+']'
: null
;
}

How can I write the condition `s can be converted to string` in
JavaScript?

Given that an object might implement its own toString method but fail to
convert because of some error, the only guaranteed way to know whether it
can be converted to a string is to do the conversion and handle the error:

function bracketize(s) {
try {
return '['+s+']';
} catch(e) {
return null;
}
}

N.B. This function will produce different output on different browsers. For
example, if an object's toString method throws an error Mozilla will return
null but IE will catch and ignore the error from toString so you get '[]'.
You get a more stable result of you call toString yourself:

function bracketize(s) {
try {
return '['+(s.toString?s .toString():s)+ ']';
} catch(e) {
return null;
}
}

[Ronald Fischer]

"Michael Winter" <M.******@bluey onder.co.invali d> wrote in message news:<opshnubpq 7x13kvk@atlanti s>...

On 18 Nov 2004 04:05:48 -0800, Ronald Fischer <ro*****@eml.cc > wrote:

Assume the following JavaScript function:

function bracketize(s)
{
return '['+s+']';
}

This function which doesn't assume anything about its argument
except that it must be convertible to a string.

This causes a problem because *everything* can be converted to a string.
That conversion may not be particularly useful, but it can be done
nevertheless.

This is what I thought at first too, but it turned out to be not true. Maybe
I should give some background on this:

We have a general utility function which tests wheather a string is
empty:
return s==null || s=='';

This function ends up to be called in a context where s is sometimes not
a string, but still this test is done for any object s - this may seem
to be weird, but assume for the moment that this makes perfectly sense
in this context.

Now the problem is that JavaScript occasionally complains

"Can't convert s to string"

as I can see in the logfiles. So there must be cases where a conversion
to string is not possible.

I can imagine one, although very pathological, case where this could
happen: It must be that s is an object where no "toString" method exists.
Can this be possible? Can such a property, which is inherited from
Object, be dropped at run-time?

But maybe more realistic cases are possible:

Of course it is conceivable that the error message does not come
from the JavaScript engine, but from a custom toString method attached
to the object s (which is likely an object from a third-party library
we are using); but when I researched the discussions in the
usenet, I did find a few threads mentioning too the "can't convert ...
to string" message, so it seems to me that it indeed comes from
JavaScript.

Ronald

[Michael Winter]

On 19 Nov 2004 00:17:26 -0800, Ronald Fischer <ro*****@eml.cc > wrote:

[T]here must be cases where a conversion to string is not possible.

I can imagine one, although very pathological, case where this could
happen: It must be that s is an object where no "toString" method
exists. Can this be possible? Can such a property, which is inherited
from Object, be dropped at run-time?
Potentially. For an object to be converted to a string, it must possess
either a toString or valueOf method. They are checked in that order. If
neither are objects or, when evaluted, one is an object that doesn't
implement the internal [[Call]] method, an exception is generated.

var o = new Object();
String(o); // "[Object object]"
o.toString = void 0; // Assign (undefined)
String(o); // Error

As you said, it would be a very unusual situation where someone would
intentionally remove the toString method. I can see reason to create your
own (I do when needed), but not remove it entirely.

[snip]
Of course it is conceivable that the error message does not come from
the JavaScript engine, but from a custom toString method attached to the
object s (which is likely an object from a third-party library we are
using);

Might I ask what environment you are using?

[snip]

Mike

--
Michael Winter
Replace ".invalid" with ".uk" to reply by e-mail.

[Grant Wagner]

Ronald Fischer wrote:

"Michael Winter" <M.******@bluey onder.co.invali d> wrote in message news:<opshnubpq 7x13kvk@atlanti s>...

On 18 Nov 2004 04:05:48 -0800, Ronald Fischer <ro*****@eml.cc > wrote:

Assume the following JavaScript function:

function bracketize(s)
{
return '['+s+']';
}

This function which doesn't assume anything about its argument
except that it must be convertible to a string.
This causes a problem because *everything* can be converted to a string.
That conversion may not be particularly useful, but it can be done
nevertheless.

This is what I thought at first too, but it turned out to be not true. Maybe
I should give some background on this:

We have a general utility function which tests wheather a string is
empty:
return s==null || s=='';

This function ends up to be called in a context where s is sometimes not
a string, but still this test is done for any object s - this may seem
to be weird, but assume for the moment that this makes perfectly sense
in this context.

Now the problem is that JavaScript occasionally complains

"Can't convert s to string"

as I can see in the logfiles. So there must be cases where a conversion
to string is not possible.

I'd like to see a small, working code sample that demonstrates this behaviour, because I've never seen it.
For example:

<script type="text/javascript">
function test(s) {
document.write( '[' + s + ']');
}
test();
</script>

outputs: [undefined]
Of course it is conceivable that the error message does not come
from the JavaScript engine, but from a custom toString method attached
to the object s (which is likely an object from a third-party library
we are using); but when I researched the discussions in the
usenet, I did find a few threads mentioning too the "can't convert ...
to string" message, so it seems to me that it indeed comes from
JavaScript.

I can't duplicate the exact error, but if you attempt to access something you do not have access to, I
read some references to the error you are receiving on google. Something like:

<iframe name="foo" src="http://www.microsoft.c om" onload="test(); "></iframe>
<script type="text/javascript">
function test() {
var bar = window.frames['foo'].location;
alert(bar);
}
</script>

Gives "Access is denied" in IE and "Error: uncaught exception: [Exception... "Could not convert JavaScript
argument arg 0 [nsIDOMWindowInt ernal.alert]" nsresult: "0x80570009 (NS_ERROR_XPC_B AD_CONVERT_JS)"
location: "JS frame :: file:///c:/DOCUME~1/grantw/LOCALS~1/Temp/hs~new.htm :: test :: line 13" data: no]"
in Firefox.

Maybe other browsers, other circumstances, it would produce "Can't convert ... to string".

You could _try_ "return '[' + String(s) + ']';" but I'm guessing that in the circumstances in which you
receive the error, there is nothing you can do to prevent the problem, since as my first example
demonstrates, even an undefined value gets converted to a valid string, so anything else should.

--
Grant Wagner <gw*****@agrico reunited.com>
comp.lang.javas cript FAQ - http://jibbering.com/faq

[Ronald Fischer]

"Michael Winter" <M.******@bluey onder.co.invali d> wrote in message news:<opshpws3x sx13kvk@atlanti s>...

On 19 Nov 2004 00:17:26 -0800, Ronald Fischer <ro*****@eml.cc > wrote:

Of course it is conceivable that the error message does not come from
the JavaScript engine, but from a custom toString method attached to the
object s (which is likely an object from a third-party library we are
using);

Might I ask what environment you are using?

The Broadvision framework (including a (slightly modified) JavaScript), running
under HP-UX.

Ronald

[Ronald Fischer]

Grant Wagner <gw*****@agrico reunited.com> wrote in message news:<41******* ********@agrico reunited.com>.. .

I'd like to see a small, working code sample that demonstrates this behaviour, because I've never seen it.
So would I. The problem is that this behavious is not reproducible in the
lab. I know its existence only from our trace files, but whenever I try
to run similar examples manually, everything works as expected.
I can't duplicate the exact error, but if you attempt to access something you do not have access to, I
read some references to the error you are receiving on google. Something like:

<iframe name="foo" src="http://www.microsoft.c om" onload="test(); "></iframe>
<script type="text/javascript">

I searched on Google, but all cases relate to client-side JavaScript only
(which we never had a problem with), and they didn't seem to be related
to my problem anyway....

Ronald

[Grant Wagner]

Ronald Fischer wrote:

Grant Wagner <gw*****@agrico reunited.com> wrote in message news:<41******* ********@agrico reunited.com>.. .

I'd like to see a small, working code sample that demonstrates this behaviour, because I've never seen it.

So would I. The problem is that this behavious is not reproducible in the
lab. I know its existence only from our trace files, but whenever I try
to run similar examples manually, everything works as expected.

I can't duplicate the exact error, but if you attempt to access something you do not have access to, I
read some references to the error you are receiving on google. Something like:

<iframe name="foo" src="http://www.microsoft.c om" onload="test(); "></iframe>
<script type="text/javascript">

I searched on Google, but all cases relate to client-side JavaScript only
(which we never had a problem with), and they didn't seem to be related
to my problem anyway....

I wasn't aware you were referring to server-side JavaScript under BroadVision. Yes, I've seen BroadVision return
an object that can not be converted to a string. I can't remember the specific circumstances, nor what I did to
resolve it. However, you have to be aware of circumstances when you are returning BVI_ objects and when you are
returning JavaScript objects. For example:

// returns a BVI_Value object:
var aBviValueObject = aBviTable.get(0 , "column_nam e");
// which you should access using:
Response.write( aBviValueObject .stringValue);

However:

// returns a String:
aBviTable.curso r = 0;
var aString = aBviTable.colum n_name;
// which can be accessed simply by:
Response.write( aString);

--
Grant Wagner <gw*****@agrico reunited.com>
comp.lang.javas cript FAQ - http://jibbering.com/faq

[Ronald Fischer]

Grant Wagner <gw*****@agrico reunited.com> wrote in message news:<41******* ********@agrico reunited.com>.. .

I wasn't aware you were referring to server-side JavaScript under BroadVision. Yes, I've seen BroadVision return
an object that can not be converted to a string. I can't remember the specific circumstances, nor what I did to
resolve it. However, you have to be aware of circumstances when you are returning BVI_ objects and when you are
returning JavaScript objects. For example:

// returns a BVI_Value object:
var aBviValueObject = aBviTable.get(0 , "column_nam e");
// which you should access using:
Response.write( aBviValueObject .stringValue);

This would indeed fit into the context where my function is actually called!

Do you see a way to check at runtime what type an object actually is,
or wheather it contains a toString method?

Ronald