mORMot Open Source

It seems NormVideo is not void ?

Procedure NormVideo;
{
  Set normal back and foregroundcolors.
}
Begin
  TextColor(7);
  TextBackGround(0);
End; 

I just tried to add "Crt.NormVideo" at the end of the above .lpr under FPC and it worked.

After searching the internet I am still not sure. For example, on some distributions the default terminal color is dependent on the "theme". Perhaps it is better to leave it as is...

Many thanks for your kind help !

Many thanks !

> IMHO it is up to the consumers (other objets) to make their own private copy of the original dynamic array.
In principle you are right.

Still, it does not hurt anyone to be defensive, when the overhead of copying the primitive array is negligible.

> Using such a getter is slower when you access the array items,

Iteration will not be done on the property but always on a local var instead. "ArrByRef" is provided in case "consumer" is reading only.

> , and also probably misleading the serializer expectations.

Apparently the serializer is "definitely" misled. The value of "Arr" is "[]", which should be the same as the value of "ArrByRef" "[0]" or "[0,1,..]". Therefore, could you help to comment whether it is technically possible for mORMot to serialize "Arr" property correctly here ?

Dear mpv, I am trying to access the "mainForm" exported by the host SpiderMonkey45Binding.exe from the plugin mathModule.dll. The reason behind my attempt is that if this is indeed possible, this looks like a good way of manipulating "central business service" exported by the host .exe from plugin .dll. Complex additional manipulation referencing the "central business service" could be compiled into plugin .dll without changing the host .exe.

Modified files are uploaded onto gist : uMathModule.pas belongs to "01 - Dll Modules\math-module\src\" and the rest six files belong to "02 - Bindings".

However, I have met several problems in trying with the above files:

(1) The plugin methods have to reference the business class. That is to say, ufrmSM45Demo has to be used in uMathModule.pas. But then the type incompatibility issue arises. For example, in function try_work_tform1_1, InstRec.instance is TfrmSM45Demo fails. I wonder whether it is safe or dangerous to hard cast as done in the code ?

(2) Further attempts try to access the ability of the mainForm , which is obtained by hard-cast, to evaluate JavaScript. In other words:
  (a) host evaluates JavaScript string, where host exports mainForm
  (b) in the JavaScript string, host calls plugin JS method with the mainForm exported
  (c) in the plugin native method, the mainForm obtained by hard-cast is asked to evaluate JavaScript

  However, the attempts all fail: function try_work_tform1_2 does not completes 100% correct, and function try_work_tform2_2 gives access violation; function try_work_tform1_3 and try_work_tform2_3 both give "A component named Form? already exists."

(3) The above problems are the same no matter FastMM4 or ShareMem is used for host or plugin or both.

Could you help to check the code and comment the reason ? 

Furthermore, if I am taking the wrong way, could you help to comment the correct solution to access the "central business service" of host .exe within the plugin dll ?

Dear Orel,

  thank you very much for your kind help and instructive sample code !

  PS: It seems that in your above post and also your #26 post, one should pay strict attention to the rooting and the exception handling.

  Could you help to comment whether you would suggest that I revise the code below (also shown in my #35 post) accordingly ? Many thanks !

  From

function TfrmSM45Demo.NewForm1(cx: PJSContext; argc: uintN; var vp: JSArgRec): Boolean;
var
  Frm1: TForm1;
begin
  Frm1 := TForm1.Create(Self);
  vp.rval := CreateJSInstanceObjForSimpleRTTI({FEngine.}cx, Frm1, FEngine.GlobalObject);
  Result := True;
end;

  to

function TfrmSM45Demo.NewForm1(cx: PJSContext; argc: uintN; var vp: JSArgRec): Boolean;
var
  frm1: TForm1;
  rootObj: PJSRootedObject;
begin
  try
    frm1 := TForm1.Create(Self);
    try
      rootObj := cx.NewRootedObject(CreateJSInstanceObjForSimpleRTTI({FEngine.}cx, frm1, FEngine.GlobalObject));
      vp.rval := rootObj.ptr.ToJSValue;
    finally
      cx.FreeRootedObject(rootObj);
    end;
    Result := True;

  except
    on E: Exception do
    begin
      Result := False;
      vp.rval := JSVAL_VOID;
      JSError(cx, E);
    end;
  end;

More over, if the above revision is correct, could you help to comment why there is cx.FreeRootedObject(rootObj); ? I mean, the return value of this JSNative function, vp.rval, will be needed afterwards. The vp.rval stores rootObj.ptr.ToJSValue, but rootObj is freed ? The GC Rooting Guide does not help me understand here. It really puzzles me

Really great news ! Really fabulous work !!

Great to know that !

Thank you for your efforts and helpful comments !

Dear mpv,

Could you help to comment about returning Delphi object from JavaScript function, using mORMot\SyNode\Samples\02 - Bindings as an example:
(1) add TForm1 to the project, as shown in #35 post of this thread, and modify the btnEvaluateClick as below

procedure TfrmSM45Demo.btnEvaluateClick(Sender: TObject);
var
  res: jsval;
  Frm1V: Variant;
  Frm1: TForm1;
begin
  // evaluate a text from mSource memo
  if mSource.SelText <> '' then
    FEngine.Evaluate(mSource.SelText, 'mSourceSelected.js', 1, res)
  else
    FEngine.Evaluate(mSource.lines.Text, 'mSource.js', 1, res);

  res := FEngine.CallObjectFunction(FEngine.GlobalObject, 'NewForm1AsJSFunc', []);
  Frm1V := res.asSimpleVariant[FEngine.cx];  <---- ESMException in function jsval.getSimpleVariant(cx: PJSContext): Variant;
  Frm1 := TVarData(V).VPointer; // Seems feasible to hold object in Variant according to Barry Kelly's comment in this SO post http://stackoverflow.com/questions/366329/why-cant-delphi-variants-hold-objects
  Frm1.Caption := 'hello ! Delphi Object from JavaScript!';
end;

(2) when running, in the left memo, type in the following code and evaluate. Unfortunately, ESMException occurs in function jsval.getSimpleVariant(cx: PJSContext): Variant;

function NewForm1AsJSFunc()
{
    var f = mainForm.NewForm1()
    return f
}

Could you help to finish the TODOs concerning JSTYPE_OBJECT and JSTYPE_FUNCTION in function jsval.getSimpleVariant, so that Delphi code can call JavaScript functions which return customized Delphi object ?

Dear AOG, does this mean that we can download zipped archive from that github link and replace the c:\NewPascal\ccr\mORMot ?

If yes, can you comment whether you can keep mORMot in your github link above always sync with http://synopse.info/fossil/timeline ?

Many thanks !

Could you help to comment a few more small problems that I meet when trying mORMot\SyNode\Samples\02 - Bindings:

(1) when running, in the left memo, calling fs.writeFileSync fails complaining writeFileNew is undefined, calling fs.appendFileSync fails complaining appendFileNew is undefined.

(2) when running, in the left memo, calling alert fails complaining alert is undefined...

(3) when running, in the left memo, calling const console = require('console') or const console = require('console').Console fails complaining Console expects a writable stream instance...

(4) when running, in the left memo, adding four lines below will not trigger function TStringsTextRead at all(breakpoint does not stop)... Is TStringsProto used at all in this sample ?

let msg1 = mainForm.results.Text
let msg2 = mainForm.results

var msg3 = mainForm.results.Text
var msg4 = mainForm.results

(5) when running, in the left memo, var x = new TStrings() is evaluated without error, even after the aEngine.defineClass(mSource.lines.ClassType, TStringsXProto, aEngine.GlobalObject); line in the source code is commented out. Why is this possible ? It seems that SyNode does not automatically call defineClass by itself.

(6) when running, in the left memo, after the code below is evaluated, the right memo gives "HelloWorld" without the line break ? More importantly,  neither function TStringsTextRead nor function TStringsWrite is triggered (breakpoint does not stop)...

var x = new TStrings()
x.Text = "Hello\nWorld"
mainForm.toLog(x.Text)

var y = new TMemoStrings()
y.Text = "Hello\nWorld"
mainForm.toLog(y.Text)

[7] when running, in the left memo, the code below exemplifies the usage of TStringsProto to me !

mainForm.toLog("line 1");
mainForm.toLog("line 2\r\nline 3");
mainForm.toLog("line 4");

let l_memo = mainForm.results /* results is the published property of TMainForm and thus accessible automatically ? */
let l_memostrings = mainForm.results.lines /* Lines is the published property of TMemos and thus accessible automatically, as implemented in SyNodeSimpleProto.pas:477: ? */
var v_memo = mainForm.results /* Points to the same reference as l_memo */
var v_memostrings = mainForm.results.lines /* Points to the same reference as l_memostrings */

/* TStringsTextRead */
mainForm.toLog("l_memo.lines.text: " + l_memo.lines.text) 
mainForm.toLog("v_memo.lines.text: " + v_memo.lines.text) 

/* TStringsTextWrite */
l_memostrings.text = "line 1\r\nline 2\r\nline 3\r\nline 4"
mainForm.toLog("v_memostrings.text: " + v_memostrings.text) 

Many thanks !!

Dear mpv and Orel,

Could you help to comment how to use SyNode (SpiderMonkey) to expose a whole set of Delphi class/object hierarchy to JavaScript ?

As shown in ADOBE ILLUSTRATOR CC 2015.3 SCRIPTING REFERENCE: JAVASCRIPT, Adobe Illustrator has exposed its whole CPP class/object hierarchy to JavaScript. Consequently, we can create arbitrary Illustrator CPP instances within JavaScript, and call methods of these Illustrator CPP instances within JavaScript.

#include "lib/lodash.3.10.1.js"
    var docRef = Application.Documents.Add();  // Create an empty Document with default settings
    ...
    var groupRef = docRef.layers[0].groupItems[0];  // Manipulate the properties of the newly-created Document
    _.each(groupRef.pathItems, function(pathItemRef)
    {
      if(pathItemRef.strokeDashes.length >= 1)
      {
        pathItemRef.strokeWidth = 2.0;
        pathItemRef.strokeDashes = [2, 4];
      }     
    })

Nonetheless, as shown in the DoOnCreateNewEngine callback in mORMot\SyNode\Samples\02 - Bindings, it seems that only existing Delphi instance is exposed to the JavaScript global object as property.

 
  aEngine.GlobalObject.ptr.DefineProperty(aEngine.cx, 'mainForm',
    // proeprty value is a wrapper around the Self
    CreateJSInstanceObjForSimpleRTTI(aEngine.cx, Self, aEngine.GlobalObject),
    // we can enumerate this property, it read-only and can not be deleted
    JSPROP_ENUMERATE or JSPROP_READONLY or JSPROP_PERMANENT
  );

Furthermore, as shown in 22.3.2. Calling Delphi code from JavaScript, it seems that only method of existing Delphi instance is exposed to the JavaScript global object as method.

procedure TTestServer.DoOnNewEngine(const Engine: TSMEngine);
...
  // add native function to the engine
  Engine.RegisterMethod(Engine.GlobalObj,'loadFile',LoadFile,1);
end;
function TTestServer.LoadFile(const This: variant; const Args: array of variant): variant;
begin
  if length(Args)<>1 then
    raise Exception.Create('Invalid number of args for loadFile(): required 1 (file path)');
  result := AnyTextFileToSynUnicode(Args[0]);
end;

Lastly, I could not figure out how to do this from https://developer.mozilla.org/en-US/docs/Mozilla/Projects/SpiderMonkey/JSAPI_Cookbook

I am therefore rather lost how to achieve similar effect with SyNode (SpiderMonkey). For the hierarchy shown below, could you help to comment how to expose not only the instance of the TMainForm class but also the TWorkerForm and TWorker classes (instances may not be created yet), so that TWorderForm can be created  in JavaScript, and methods of TWorker can be called in JavaScript ?

Many thanks !!

Dear Orel, thank you very much for your knowledgeable posts ! (PJSRootedObject does not have isObject so that mResult.Lines.Add(BoolToStr(jsArr.isObject, True)); line has to be commented out.)
It then seems that we could not get Delphi array directly from jsval but has to call ptr.GetElement(cx, index, jsArrElement); manually.

Could you help to comment why the "JSObject.Is***Array" returns false in the code below ? The array element is apparently numeric. Many thanks !

PS: The function can be appended to SpiderMonkey45Binding: ufrmSM45Demo.

// Test with: 
//   mainForm.JSTestArray1D([1,2,3])
//   mainForm.JSTestArray1D([1.0,2.0,3.0])
//   mainForm.JSTestArray1D([1.1,2.2,3.3])  
function TfrmSM45Demo.JSTestArray1D(cx: PJSContext; argc: uintN; var vp: JSArgRec): Boolean;
var
  jsArr, jsArrElement: jsval;
  Cnt, I: Cardinal;
begin
  jsArr := vp.argv[0];
  mResult.Lines.Add('###########################');
  mResult.Lines.Add(BoolToStr(jsArr.isObject, True));
  mResult.Lines.Add(BoolToStr(jsArr.asObject.isArray(cx), True));
  mResult.Lines.Add(BoolToStr(jsArr.asObject.IsTypedArrayObject, True));
  mResult.Lines.Add(BoolToStr(jsArr.asObject.IsInt8Array, True));
  mResult.Lines.Add(BoolToStr(jsArr.asObject.IsUInt8Array, True));
  mResult.Lines.Add(BoolToStr(jsArr.asObject.IsInt16Array, True));
  mResult.Lines.Add(BoolToStr(jsArr.asObject.IsUInt16Array, True));
  mResult.Lines.Add(BoolToStr(jsArr.asObject.IsInt32Array, True));
  mResult.Lines.Add(BoolToStr(jsArr.asObject.IsUInt32Array, True));
  mResult.Lines.Add(BoolToStr(jsArr.asObject.IsFloat32Array, True));
  mResult.Lines.Add(BoolToStr(jsArr.asObject.IsFloat64Array, True));
  mResult.Lines.Add('---------------------------');
  jsArr.asObject.GetArrayLength(cx, Cnt);
  mResult.Lines.Add(IntToStr(Cnt));
  jsArr.asObject.GetElement(cx, 0, jsArrElement);
  mResult.Lines.Add(BoolToStr(jsArrElement.isInteger, True));
  mResult.Lines.Add(BoolToStr(jsArrElement.isDouble, True));
  mResult.Lines.Add(BoolToStr(jsArrElement.isNumber, True));
  mResult.Lines.Add(BoolToStr(jsArrElement.isSimpleVariant[cx], True));
  mResult.Lines.Add('###########################');

  Result := True;
end;

I should be more precise: characters encoded with UTF8.
I see your point. Now I feel lost again why mpv's fix works...

It seems that the memory footer corruption can be avoided with some modification of TSynTempBuffer to avoid the re-allocation in TSynTempBuffer.Init. The reason is unclear. Could ab or you help to correct the root problem ?

  private
    // tmp: array[0..4095] of AnsiChar; <--- memory footer corruption
    // 
    // for 'mORMot\SyNode\Samples\02 - Bindings\SpiderMonkey45Binding.dpr', 
    // where 'mORMot\SyNode\core_modules\DevTools\Debugger.js' will be loaded
    // = Length of 'mORMot\SyNode\core_modules\DevTools\Debugger.js' * 2 + 1 
    tmp: array[0..118038] of AnsiChar; 
    //
    // if only the files 'mORMot\SyNode\core_modules\node_modules\' are to be loaded
    // = Length of NativeModule.wrap('mORMot\SyNode\core_modules\node_modules\util.js') * 2 + 1 
    // tmp: array[0..33584] of AnsiChar; 
  end;

Dear mpv,

http://synopse.info/forum/viewtopic.php?id=3544
http://synopse.info/forum/viewtopic.php?id=3571

In the links above, you have identified that there should be a conflict between FastMM in FullDebug mode and a nspr memory manager (SpiderMonkey use a memory manager from Netscape Portable Run-time nspr4.dll).

However, FullDebugMode is needed for FastMM to log all memory leaks to a text file, and therefore rather important for finding memory leaks.

In SynSMAPI.pas, there is mentioned "compile mozjs without nspr" under "road map". Does this mean that this incompatibility between FastMM w/ FullDebugMode and SpiderMonkey will be removed in future ? 

Thank you very much for your efforts in researching ! Because FullDebugMode is rather important for FastMM for finding memory leaks, I hope this incompatibility between FastMM w/ FullDebugMode and SpiderMonkey could be removed in future.

Thank you for fixing the code and committing the change to trunk!

About the exception when debugger is off: do you mean ab might know all the details ?