froilanq / google-jstemplate Goto Github PK

For data:
var tplData = {username:"Jane User",
                    addresses:{
                      primary: {location:"111 8th Av.", label:"NYC front door"},
                      2: {location:"76 9th Av.", label:"NYC back door"},
                      3: {location:"Mountain View", label:"Mothership"}
                     }
                    };


It should be possible to loop over addresses with jsselect. 
Not sure if $index should contain the property name or they should be 
additional $key variable.

What steps will reproduce the problem?
1. Construct a template using something like <span id="template"
jsvalues="$options:'a;b;c'.split(';')"><span jsselect="$options"
jscontent="$this" /></span>
2. call jstProcess(new JsEvalContext({}), document.getElementById('template'))

You should get a stack of three spans <span>a</span>...b...c...

Instead, this template is not processed because the expression dies at ';',
and what's left is a syntax error.  I assume this is because the expression
is split with ';' with no regard to string boundaries.

I'm using svn r9, FF3/Ubuntu8.04.

I am working around this by using '|' instead.  I can imagine a scenario in
which this workaround would not be practical, though.

$top is very useful, but not mentioned anywhere in the reference
documentation on the website.  It is mentioned in the slideshow, though.

Should add it to the reference documentation.

it is very complicated to read and understand the examples on 
https://code.google.com/p/google-jstemplate/wiki/TemplateProcessingInstructionRe
ference page

the syntax highlighting puts all code on one line

could you pls. fix it ?

util.js has the following comment
    // NOTE(mesch): An alternative idiom would be:
    //
    //   eval('(' + expr + ')');
    //
    // Note that using the square brackets as below, "" evals to undefined.
    // The alternative of using parentheses does not work when evaluating
    // function literals in IE.
    // e.g. eval("(function() {})") returns undefined, and not a function
    // object, in IE.

This is correct, but the solution:
    return eval('[' + expr + '][0]');
creates an immediately discarded array object even on interpreters that do
not have IE's bug.

An alternative which creates fewer objects is:
    return eval('0,(' + expr + ')');

Moving the strings '0,(', and ')' out of the function body will create
fewer string objects too.

Is there really a reason not to provide at least possibility to do it with 
awareness that it should not output nodes with js* parameters ?

thx

are there plans to convert this library into RequireJS AMD 
(http://requirejs.org/) module ?

What steps will reproduce the problem?
 1. use a jsselect inside a <form>, on an <input>.  The <input> should have
a constant 'name' attribute, such that document.formxx.field_name is an
array of fields with that name.  (See attached html file.) 
 2. call jstProcess on the form, to duplicate that field several times.
 3. compare document order and the order in the formxx.field_name array.

Expected: The two alerts should be the same, in my test page.

Actual: As RAtM (and Luke) said, "the last shall be first".


I'm using svn r9, FF3/Ubuntu8.04.


It is very common for apps to use document.formxx.field_name[n] where more
than one field has the same name 'field_name'.  Referenced through
formxx.field_name, these fields are in order by how early they were created
(chronologically), *not* in document order.  jsselect's use of
domInsertBefore shifts the order backwards 1 slot, placing the 0'th element
at the end of the form.name array.