So, it seems the bug is partly saying `| constraint`, because any non-final
attribute is, in theory, constrainable (there may be some that don't work, but
that would be a bug in either the declaration or the implementation).
Secondly, I think we would like to support the XSD declaration for non-negative
number, so we ought to teach convertAttributeType about that.
Finally, we don't currently have a way to express that restriction in
Javascript (although perhaps some day we will). So for now, it is really just
a descriptive declaration.
I'm willing to accept just changing the declaration to `number`, but if so, we
should file an improvement to support the `xsd:float {minInclusive="0"}`
declaration.
On 2010-05-26, at 07:56, André Bargull wrote:
> The schema compiler changes any enumeration [1] (like "foo | bar | baz") to
> "string" type, because enumerations are only used (intended to be used?) for
> string values [2]. Therefore declaring <animatorgroup>#duration's type to
> 'xsd:float {minInclusive="0"} | constraint' is just wrong. We could change
> the type to 'xsd:float {minInclusive="0"}', but then the schema compiler
> needs to change it to "number" again, because the schema reader doesn't
> understand that type [3].
> We just don't have any facility like in xsd to constrain attributes to valid
> values. I don't even know if such a facility is feasible at compile time due
> to constraints.
>
>
> [1] except for boolean | 'inherit' which becomes inheritableBoolean, see
> SchemaBuilder#convertAttributeType()
> [2] complete list of enumeration types from lfc.lzx schema file:
> <node>#initstage, <animatorgroup>#motion, <animatorgroup>#process,
> <datapath>#axis, <datapath>#replication, <dataset>#querytype, <dataset>#type,
> <script>#when, <view>#align, <view>#backgroundrepeat, <view>#stretches,
> <view>#valign, <include>#type
> [3] cf. SchemaBuilder#convertAttributeType()
>
> On 5/26/2010 1:29 PM, P T Withington wrote:
>> Sorry, this got lost in my inbox.
>>
>> My question is: clearly the intent is to have the schema indicate (and if
>> you have a verifier, verify) that an animation duration must be
>> non-negative. So the real bug lies in the schema compiler, not in the
>> declaration. So shouldn't we really be fixing that?
>>
>> To change the XML type declaration to "number" is acceptable if there is no
>> way to fix the schema compiler.
>>
>> On 2010-05-07, at 12:36, André Bargull wrote:
>>
>>> Change 20100507-bargull-4X2 by barg...@bargull02 on 2010-05-07 18:18:33
>>> in /home/anba/src/svn/openlaszlo/trunk
>>> for http://svn.openlaszlo.org/openlaszlo/trunk
>>>
>>> Summary: change lz.animatorgroup#duration lzxtype to "number"
>>>
>>> New Features:
>>>
>>> Bugs Fixed: LPP-8980 (http://jira.openlaszlo.org/jira/browse/LPP-8980)
>>>
>>> Technical Reviewer: ptw
>>> QA Reviewer: (pending)
>>> Doc Reviewer: (pending)
>>>
>>> Documentation:
>>>
>>> Release Notes:
>>> lz.animatorgroup#duration type has been changed to "number"
>>>
>>> Overview:
>>>
>>>
>>> Details:
>>> This is a change for the @lzxtype for lz.animatorgroup#duration. The
>>> lzxtype was set to 'xsd:float {minInclusive="0"} | constraint' which was
>>> treated as an enumeration type by the schema compiler and therefore the
>>> actual type was "string". This means any value for the duration attribute
>>> written in the<animatorgroup> tag was changed to a String in script.
>>> That's for sure not the desired result for a numeric type like duration.
>>> With this change, the @lzxtype is updated to "number". There is only one
>>> potential issue: Users will have to update<attribute name="duration"
>>> value="..." type="string"/> to<attribute name="duration" value="..."
>>> type="number"/>. Although I doubt anyone is really using that form.
>>>
>>>
>>> Tests:
>>> testcase from bugreport
>>>
>>> Files:
>>> M WEB-INF/lps/lfc/controllers/LzAnimatorGroup.lzs
>>>
>>> Changeset:
>>> http://svn.openlaszlo.org/openlaszlo/patches/20100507-bargull-4X2.tar
>>>
>>
>>
