Hi Max,
Great to hear! Any chance we could get some numbers on the performance
changes? What are you running?
- Don
On Mar 17, 2011, at 12:00 PM, Max Carlson wrote:
> I thought I'd try this out - and the performance boost is amazing! I was
> wondering why the generated app code was slightly larger. For example,
> components/demos/house.lzx?lzt=object&lzr=dhtml has tons declarations like
> this:
>
> (function($0){
> with($0)with($0.prototype){
> {
> LzNode.mergeAttributes({$CSSDescriptor:{},$attributeDescriptor:{types:{aaactive:"boolean",aadescription:"string",aaname:"string",aasilent:"boolean",aatabindex:"number",align:"string",backgroundrepeat:"string",bgcolor:"color",cachebitmap:"boolean",capabilities:"string",classroot:"string",clickable:"boolean",clickregion:"string",clip:"boolean",cloneManager:"string",contextmenu:"string",cornerradius:"string",cursor:"token",datapath:"string",defaultplacement:"string",fgcolor:"color",focusable:"boolean",focustrap:"boolean",font:"string",fontsize:"size",fontstyle:"string",frame:"numberExpression",framesloadratio:"number",hasdirectionallayout:"boolean",hassetheight:"boolean",hassetwidth:"boolean",height:"size",id:"ID",ignoreplacement:"boolean",immediateparent:"string",inited:"boolean",initstage:"string",isinited:"boolean",layout:"css",loadratio:"number",mask:"string",name:"token",nodeLevel:"number",opacity:"number",options:"css",parent:"string",pixellock:"boolean",placement:"string"
> ,playing:"boolean",resource:"string",resourceheight:"number",resourcewidth:"number",rotation:"numberExpression",selected:"boolean",shadowangle:"number",shadowblurradius:"number",shadowcolor:"color",shadowdistance:"number",showhandcursor:"boolean",source:"string",stretches:"string",styleclass:"string",subnodes:"string",subviews:"string",tintcolor:"string",totalframes:"number",transition:"string",unstretchedheight:"number",unstretchedwidth:"number",usegetbounds:"boolean",valign:"string",visibility:"string",visible:"boolean",width:"size","with":"string",x:"numberExpression",xoffset:"numberExpression",xscale:"numberExpression",y:"numberExpression",yoffset:"numberExpression",yscale:"numberExpression"}},onselected:LzDeclaredEvent,selected:false,selectionmanager:null},$lzc$class__m2f.attributes)
> }}})($lzc$class__m2f)
> };
>
> It seems like the with($0) shouldn't be there. Awesome work so far!
>
> On 3/16/11 2:06 PM, Donald Anderson wrote:
>> Change dda-20110301-0dx by [email protected]
>> <mailto:[email protected]> on 2011-03-01 14:23:55 EST
>> in /Users/dda/laszlo/src/svn/openlaszlo/trunk-c/WEB-INF/lps/schema
>> for http://svn.openlaszlo.org/openlaszlo/trunk/WEB-INF/lps/schema
>>
>> Summary: Eliminate the need for with(this) from most typical cases
>>
>> New Features:
>>
>> Bugs Fixed: LPP-8751 Eliminate the use of `with(this)`
>>
>> Technical Reviewer: ptw (pending)
>> QA Reviewer: henry (pending)
>> Doc Reviewer: (pending)
>>
>> Documentation:
>>
>> Release Notes:
>>
>> Overview:
>> [Mar 16 review - updated review comments indicated in [] brackets below]
>>
>> with(this) can be eliminated from a function only when either of these
>> conditions occur:
>> - we have an complete knowledge of the class heirarchy.
>> - there are no free references in the function
>> and when this condition does not occur:
>> - another 'with(varname)' appears in the function.
>> We address the first condition via changes to the js2doc as it
>> generates the lfc schema, the tag compiler as it generates extra
>> 'reference' JS classes that contain attributes and methods with
>> empty implementations, and the script compiler, as it uses the
>> reference classes to create a (usually) complete picture of the class
>> heirarchy for most cases.There's nothing we can do about the
>> second condition [a change from previously].As for the third
>> condition, there are ways of mitigating it, but we don't bother for now
>> on the assumption that a function coded containing 'with(varname)' is not
>> a function that is coded for speed.
>>
>> There may be some future work that may require a complete set of
>> reference classes, or method args, etc..These are mentioned in
>> comments below as 'FUTURE WORK'.For most of these, preliminary
>> explorations were done, and abandoned for expediency.
>>
>> Details:
>> [these 3 files changed since last review: added more tests]
>> test/lztest/lztest-class-impl.lzx:
>> test/smoke/smokecheck.lzx:
>> test/smoke/compiler.lzl:
>> Added some simple tests that illustrate how with(this) is inserted
>> in certain cases (before this change).By inspecting JS, used to
>> verify that 'this.' insertion and 'with(this)' removal is working.
>> Added more tests in response to Andre's testing, and suggestions by
>> Tucker [new]
>> Added tests using classes that inherit from LzURL, since it won't
>> be 'completely known', so will require with(this). [new]
>>
>> WEB-INF/lps/schema/lfc-undeclared.lzx: [changed since last review, added
>> ondestroy, from LzEventable]
>> Added attribute: isinited, marked as private in doc, which is used
>> by old components, and cannot be marked as public.
>> Added method: setAttribute, which actually lives in LzEventable, and is
>> used quite often.We don't current emit reference classes for
>> any classes that are not associated with a tag (it was too hard
>> to get a non-tag class into the tag compiler - FUTURE WORK).
>> Added method: ondestroy, an event that really lives in LzEventable,
>> for the same reason as above. [new]
>>
>> WEB-INF/lps/lfc/core/LzNode.lzs:
>> Added devnote about isinited.
>>
>> WEB-INF/lps/lfc/compiler/Class.lzs:
>> Minor error message improvement that I found helpful once. (optional)
>>
>> WEB-INF/lps/server/src/org/openlaszlo/js2doc/SchemaBuilder.java:
>> - jsname (the javascript name) for a class as well as jsextends
>> (the javascript class it extends) are put into the schema, though
>> not currently used - FUTURE WORK).
>> - omit event names not already defined as attributes
>> - method parameters are now collected and inserted into the schema,
>> but no typing is currently preserved (FUTURE WORK).
>>
>> WEB-INF/lps/server/src/org/openlaszlo/compiler/ClassCompiler.java:
>> Ignore classes in the schema that has no associated tag name (FUTURE WORK).
>> Small change to the protocol of compiling a ClassModel.
>>
>> WEB-INF/lps/server/src/org/openlaszlo/compiler/NodeModel.java:
>> Ignore the jsname, jsextends parts of the schema (FUTURE WORK).
>> Small change to the protocol of compiling a ClassModel.
>>
>> WEB-INF/lps/server/src/org/openlaszlo/compiler/ClassModel.java: [changed
>> since last review: fixed parenthesization, emit reference classes
>> without regard to runtime type]
>> - Changed isCompiled() to be true to its name.
>> - reworked compile() to handle emiting reference classes as needed.
>> - reworked emitClassDeclaration() to work with reference classes.
>>
>> WEB-INF/lps/server/src/org/openlaszlo/sc/ScriptClass.java: [changed
>> since last review: elide, rather than rename, names that are reserved]
>> Changed generation of a class so it won't create attributes named 'with'
>> classes named 'class', etc.Needed when generating all classes for
>> the full schema (possibly optional now, but needed for FUTURE WORK).
>>
>> WEB-INF/lps/server/src/org/openlaszlo/sc/CommonGenerator.java: [changed
>> since last review: renamed predefinedGlobals->globalProperties, added
>> "canvas", "escape", "unescape", removed "prototype"]
>> - rearrange globals and debugGlobals into separate sets, enumerate
>> the members of JS 'Object' class.
>> - ClassDescriptor.getInstanceProperties() modified to return an
>> 'incomplete' set,
>> rather than null.ClassDescriptor.complete is set and needs to be checked
>> by the caller.
>> - ClassDescriptor.incompleteSet() added for informational messages.
>> - ClassDescriptor.toString() for development/debugging.
>> - process reference classes minimally when needed at all.
>>
>> WEB-INF/lps/server/src/org/openlaszlo/sc/JavascriptGenerator.java:
>> [changed since last review: renamed predefinedGlobals->globalProperties,
>> new algorithm for determining if withThis is needed]
>> - create set of forwardGlobals by doing a shallow walk of the AST.
>> - manage the 'possibleInstance' set to try to reduce it and move
>> as many items as appropriate/possible to the 'bindthis' set,
>> which is the set of variables we will actively bind to 'this'.
>> - issue warning if 'warnWithThis' or 'warnUnknownGlobals' is on.
>>
>> WEB-INF/lps/server/src/org/openlaszlo/sc/SWF9Generator.java: [changed
>> since last review - fixed code to ignore reference classes]
>> Make sure reference classes are ignored in this generator.
>>
>> WEB-INF/lps/server/src/org/openlaszlo/sc/DHTMLCompiler.java:
>> set option so that reference classes are accepted.
>>
>> WEB-INF/lps/server/src/org/openlaszlo/sc/Compiler.java: [changed since
>> last review: added "warnUnknownGlobals"]
>> define 'referenceClass' pragma issued by tag compiler.
>> define 'registerReferenceClasses', an option only set by the DHTMLCompiler.
>> define 'warnWithThis', a new debugging option that will warn
>> when with(this) is used, and give the reason behind it, e.g.:
>> 'with(this) added in foobar: unknown parts of class hierarchy:
>> [LzTestManagerClass->LzEventable] and unaccounted refs: [LzHTTPLoader,
>> escape]'
>> define 'warnUnknownGlobals', a new debugging option that will warn
>> when we know the inheritance chain, yet there are free references
>> that we cannot identify.
>>
>> WEB-INF/lps/server/src/org/openlaszlo/sc/CodeGenerator.java:
>> Abide by new ClassDescriptor protocol
>>
>> WEB-INF/lps/server/src/org/openlaszlo/sc/GenericVariableAnalyzer.java:
>> WEB-INF/lps/server/src/org/openlaszlo/sc/VariableAnalyzer.java:
>> WEB-INF/lps/server/src/org/openlaszlo/sc/WithThisAnalyzer.java:
>> Refactored VariableAnalyzer so essential parts could be shared
>> with a new 'WithThisAnalyzer'.The new analyzer walks the tree
>> and applies identifier --> this.identifier transformation for
>> any free variables found in a set.It also tracks if any
>> 'with' constructs are found, which is a spoiler for with(this) removal.
>>
>> WEB-INF/lps/server/sc/src/org/openlaszlo/sc/parser/SimpleNode.java:
>> Added indexOf() - a convenience function, helpful in the 'this binding'.
>>
>> Tests:
>> - Ran smoketest (DHTML,SWF10) lztest, runlzunit (includes new tests)
>> - For lztest, spot checked to ensure that 'this.'
>> was inserted, specifically in the 'WithThis' tests
>> (in test/lztest/lztest-class-impl.lzx)
>> -Of the occurrences in lztest that require with(this), we know why they
>> occur: 1) LzTestManagerClass inherits
>> from LzEventable, not a class associated with a tag.This will not happen for
>> pure lzx programs. 2) they often refer to use class names in the LFC:
>> LzHTTPLoader, LzTimeKernel.3) they refer to identifiers 'canvas', 'escape'.
>>
>>
>>
>> Files:
>> Mtest/lztest/lztest-class-impl.lzx
>> Mtest/smoke/smokecheck.lzx
>> Mtest/smoke/compiler.lzl
>> MWEB-INF/lps/schema/lfc-undeclared.lzx
>> MWEB-INF/lps/lfc/core/LzNode.lzs
>> MWEB-INF/lps/lfc/compiler/Class.lzs
>> MWEB-INF/lps/server/src/org/openlaszlo/js2doc/SchemaBuilder.java
>> AWEB-INF/lps/server/src/org/openlaszlo/sc/WithThisAnalyzer.java
>> MWEB-INF/lps/server/src/org/openlaszlo/sc/JavascriptGenerator.java
>> AWEB-INF/lps/server/src/org/openlaszlo/sc/GenericVariableAnalyzer.java
>> MWEB-INF/lps/server/src/org/openlaszlo/sc/VariableAnalyzer.java
>> MWEB-INF/lps/server/src/org/openlaszlo/sc/CommonGenerator.java
>> MWEB-INF/lps/server/src/org/openlaszlo/sc/SWF9Generator.java
>> MWEB-INF/lps/server/src/org/openlaszlo/sc/DHTMLCompiler.java
>> MWEB-INF/lps/server/src/org/openlaszlo/sc/Compiler.java
>> MWEB-INF/lps/server/src/org/openlaszlo/sc/CodeGenerator.java
>> MWEB-INF/lps/server/src/org/openlaszlo/sc/ScriptClass.java
>> MWEB-INF/lps/server/src/org/openlaszlo/compiler/ClassCompiler.java
>> MWEB-INF/lps/server/src/org/openlaszlo/compiler/NodeModel.java
>> MWEB-INF/lps/server/src/org/openlaszlo/compiler/ClassModel.java
>> MWEB-INF/lps/server/sc/src/org/openlaszlo/sc/parser/SimpleNode.java
>>
>> Changeset: http://svn.openlaszlo.org/openlaszlo/patches/dda-20110301-0dx.tar
>>
>>
>>
>> --
>>
>> Don Anderson
>> Java/C/C++, Berkeley DB, systems consultant
>>
>> voice: 617-306-2057
>> email: [email protected] <mailto:[email protected]>
>> www: http://www.ddanderson.com <http://www.ddanderson.com/>
>> blog: http://libdb.wordpress.com
>>
>>
>>
>>
>>
--
Don Anderson
Java/C/C++, Berkeley DB, systems consultant
voice: 617-306-2057
email: [email protected]
www: http://www.ddanderson.com
blog: http://libdb.wordpress.com
