[patch] [cla] DTL variables in attribute nodes not replaced when empty

[carpie]

When a DTL variable is used for an attribute of a node, (e.g. inside the "value" attribute for an input node), and the variable resolves to an empty string, then the DTL variable is not replaced. If the variable resolves to null, then the DTL variable is replaced with the string 'null'. A regular Dijit widget replaces both of these with an empty string. If the DTL variable is not an attribute value, then both cases work correctly. Here's an example (full example file attached):

dojo.declare("DijitWidget", [dijit._Widget, dijit._Templated],
{
   foo: '',
   bar: null, 
   templateString: '<div><div>Clear text: foo=${foo}, bar=${bar}</div>' +
      '<div>Attr Value: foo=<input value="${foo}" /></div>' +
      '<div>bar=<input value="${bar}" /></div></div>',
});

dojo.declare("DTLWidget", [dijit._Widget, dojox.dtl._DomTemplated],
{
   foo: '',
   bar: null,
   templateString: '<div><div>Clear text: foo={{ foo }}, bar={{ bar }}</div>' +
      '<div>Attr Value: foo=<input value="{{ foo }}" /></div>' +
      '<div>bar=<input value="{{ bar }}" /></div></div>',
});

The dijit widget displays: Clear text: foo=, bar= Attr Value: foo=[] bar=[]

Teh DTL widget shows: Clear text: foo=, bar= Attr Value: foo=[{{ foo }}] bar=[null]

Assuming the DTL widget should work like the Dijit widget, I created a patch that fixes the two issues. I'm new to this code base, so there may be more appropriate fixes.

The empty string issue is caused by the fact that in AttributeNode's render function (dom.js) if the current contents of the node is "" and the resolved value is "", then setAttribute is not called to overwrite the current attribute value of {{ foo }}. My patch checks for "" and allows setAttribute to be called. It may be more appropriate to call setAttribute only if getAttribute doesn't match the resolved value. I'll defer to the maintainer's knowledge of which would be less costly.

The null issue happens because a null gets returned from _Filter's resolvePath routine. There seems to be a check to convert null to an empty string in the for loop that loops through parts. However, in the case above, this loop is entered so the check is skipped. My patch performs the check (and conversion) at the point the return value is being returned. Again, there may be a more appropriate way. The patch is against r19854.

Hope this helps...

Lee Carpenter

[carpie]

Issue demonstration

[carpie]

Patch to resolve the issue

[dante]

is there a CLA on file for you carpie? name?

[carpie]

Yes, I have a CLA. Lee Carpenter.

[bill]

1.5 is released, moving remaining tickets to 1.6

[bill]

(sadly) punting seemingly abandoned ticket and meta tickets to future

[ben hockey]

i tried to apply this patch but it seems to break several tests. do you have a patch against current trunk that doesn't introduce any new errors in the tests?

[trac-o-bot]

Because we get so many tickets, we often need to return them to the initial reporter for more information. If that person does not reply within 14 days, the ticket will automatically be closed, and that has happened in this case. If you still are interested in pursuing this issue, feel free to add a comment with the requested information and we will be happy to reopen the ticket if it is still valid. Thanks!