I'm going to add this check and then commit the change:
if (!ScriptCompiler.isIdentifier(when)) {
throw new CompilationError(
"'when' attribute for passthrough element must be a valid
javascript identifier",
element);
}
On 5/18/2010 5:44 PM, André Bargull wrote:
One think we could (should?) do, is to check whether the value is a
valid javascript identifier. Otherwise people could write <passthrough
when="(function () {return false})()" > or any other complex expression.
On 5/18/2010 4:31 PM, P T Withington wrote:
Ok. Henry, can you do what André proposes? This seems like the best
plan.
I am pretty sure that if you end up with a back-end compile where the
conditional is not defined that the back-end will signal an error;
but I think that is the best we can do (since the tag compiler has to
permit compiling to runtime-neutral .lzo, which has no runtime
specification).
On 2010-05-18, at 10:22, André Bargull wrote:
I'm not sure whether it's possible test for valid compile-time
conditionals at parse time. For example think of custom conditionals
which are only used/available when compiling the actual application.
So, I'll only change two things:
- rename "runtime" to "when"
- remove prepending the "when" value with '$' in ClassModel.java,
this means you need to write<passthrough when="$as3"> instead
of<passthrough when="as3">. This makes<passthrough>'s "when"
compatible with<switch>'s "property" attribute.
On 5/15/2010 3:31 PM, P T Withington wrote:
Ok, you convinced me. Reopen the bug and lets change it to
`when="..."`. As a bonus, you can verify that the when condition
is a valid compile-time conditional, but I'm pretty sure the script
compiler will complain if it is not.
On 2010-05-15, at 09:13, André Bargull wrote:
But valid (= currently supported) runtimes are only 'swf8',
'swf9', 'swf10' and 'dhtml'. That means 'as3' is not a valid
runtime identifier, but I think we still want to have a pass
through for both, swf9 and swf10, to avoid duplicate code.
(I know I'm a bit nit-picking right now, but this is my full
intention, because from the user's point of view the current
rationale might be confusing.)
On 5/15/2010 3:00 PM, P T Withington wrote:
In the case of passthrough, where its intent is to "pass through"
to a particular target runtime, the runtime argument seems most
appropriate to me. I suppose we should add some sanity checking
that the runtime argument is a valid runtime.
If you really need a passthrough conditional to other properties,
you can do so in a<script> tag.
---
The problem we are trying to solve here is that<switch> cannot
be used in a binary library (because its semantics are that it is
evaluated at parse time). Another approach would be to change
switch to have an 'evaluation time' property, but that seemed
like generality we didn't need.
On 2010-05-15, at 08:10, André Bargull wrote:
We had first:
<switch>
<when runtime="swf10">
<passthrough>
import flash.utils.*;
</passthrough>
</when>
</switch>
"runtime" was only allowed to be: 'swf7', 'swf8', 'swf9',
'swf10' or 'dhtml'
The first form was deprecated in favor of:
<switch>
<when property="$swf10">
<passthrough>
import flash.utils.*;
</passthrough>
</when>
</switch>
"property" can be any of the compile time constants: '$swf7',
'$swf8', '$swf9', '$swf10', '$dhtml', '$as3', '$as2', '$debug'
(left out some unused constants)
And now we're back with "runtime", although "runtime" is just
any string which will be transformed to the compile time
constants from above (by prepending "$" to the string).
<passthrough runtime="swf10">
import flash.utils.*;
</passthrough>
But also possible:
<passthrough runtime="as3">
import flash.utils.*;
</passthrough>
or:
<passthrough runtime="debug">
import flash.utils.*;
</passthrough>
So, maybe<passthrough>'s new "runtime" attribute should rather
be "property" like in<when>?
On 5/14/2010 10:46 PM, P T Withington wrote:
Approved!
On 2010-05-13, at 17:19, Henry Minsky wrote:
Change 20100513-hqm-p by [email protected] on 2010-05-13
17:17:23 EDT
in /Users/hqm/openlaszlo/trunk1
for http://svn.openlaszlo.org/openlaszlo/trunk
Summary:<passthrough> should take a `runtime="..."` property
New Features:
Bugs Fixed: LPP-8995
Technical Reviewer: ptw
QA Reviewer: (pending)
Doc Reviewer: (pending)
Documentation:
Release Notes:
Overview:
Details:
Tests:
test case below, passthru should disappear in DHTML output:
<canvas width="100%" height="80%">
<debug fontsize="12"/>
<class name="foo">
<passthrough runtime="swf10">
import flash.net.*;
</passthrough>
<method name="foo">
Debug.info('foobar');
</method>
</class>
</canvas>
Files:
M
WEB-INF/lps/server/src/org/openlaszlo/compiler/NodeModel.java
M
WEB-INF/lps/server/src/org/openlaszlo/compiler/ClassModel.java
Changeset:
http://svn.openlaszlo.org/openlaszlo/patches/20100513-hqm-p.tar
