mORMot Open Source

Nice, now demo project pass build. 

By the way, 2020-04 lazarus-2.0.8-fpc-3.2.0_fixes-44680 not works.
I can't run empty form project due error (don't work if Application.DoubleBuffered).
If i disable it, error appear in another module. I have many problems with empty project.
Suppose that issue comes from compiler itself.

Will test another version.
---
https://synopse.info/forum/viewtopic.ph … 695#p31695
FPC: "fixes3.2" and Lazarus: "fixes2.0" works.

My thoughts was about html documentation.
Currently, md format is primary documentation format, used everywhere.
For example, https://angular.io/guide/architecture
And https://docusaurus.io is just one of many services that can help with that.

Thanks!

I see, thanks!

I'm looking for predefined functions for RawUTF8, similar to "LeftStr(str, count)", "Copy(str, start, len)".
Seems SynCommons does not contain what I'm looking for.
Am i wrong?

Hello!

When delphi function returns js object (via vp: JSArgRec) with some fields, what is correct steps to be sure that JS garbage collector will not destroy object too early?
1. create local rooted object on delphi.
2. fill.
3. assign to vp.rval and call FreeRootedObject?

Or RootedObject should be used only for objects that was passed to delphi as parameters (to add reference and prevent object destruction while working with it)?

Hello, MPV!

Do you have compiled files for version 52.4.1?

I have code sample that cannot be executed in SyNode, but run without any problems inside node.js or pure ES2015
Seems iterator object for "arguments" is missing.
Where that internal member initialize?)

"use strict";
function test(...args) {
    console.log(`param count is: ${arguments.length}`);
}
function testExecutor(p1, p2) {
    test(...arguments); // <<<<<<<< arguments is not iterable
}
testExecutor('param1', 'param2');

What about synode.js, i was implemented automatic core modules search with index.js support (when module has index.js it may be loaded as folder name).
With that change, new modules will not require changes in synode.js. Pretty handy isn't it?

What changes was necessary? Version 1.6.0 of debugger work, only issue was path comparing that case sensitive (afaik it will be insensitive soon).

And you can cooperate (suggestions, ideas or may be pull requests) with debugger plugin developer
To make both projects better to work together.

With debugger developer we done some work to provide compatibility.
I patched SyNode sources to get work debugger correctly, modules placed in folder Units_PatchedModules
I'm not sure about one place, where i get thread id (in SyNodeRemoteDebugger.pas), it may be incorrect thread id (i want pass to debugger thread id that processing corresponding smEngine).
Also i patched JS code (Win64\core_modules\). Beyond compare may help evaluate differences.

Here is compiled project with sources (created in delphi 10.2).
Project include use cases of both SyNode demos with additional feature demonstration.
In addition it include vscode project written in typescript (Win64\web.apps\app1\).
Debugging work, even more, source maps work too.

@mpv, take a look, some (or almost all) changes may be imported to SyNode, as well as demo project.

Thanks, i figured out

Need little help, i get strange results while debugging Debugger.js.

    constructor (actorID, source, parent) {
        super(actorID, parent);
        parent._sourcesMap.set(source, this)
        this._source = source;
        this._breakpoints = {};
    }
    ...
    get _resp(){
        let source = this._source;// || this.generatedSource;
        consoleLog(stringify(source));   // -> [object undefined]
        consoleLog(source.url);          // -> ModuleLoader.js
        ...

How that possible?

With JSON.stringify i get:

    get _resp(){
        let source = this._source;// || this.generatedSource;
        consoleLog(JSON.stringify(source)); // -> {}
        consoleLog(source.url);             // -> ModuleLoader.js
        ...

I use output log to debug Debugger.js, is better way exists?

Should break points work in current implementation?
Looks like vscode debug plugin trying to attach to tab instead of attaching to addon. That's how it seen from wireshark.
In addition, firefox can attach, i think that issue on vscode plugin side. Well, i'll post here if get solution.

Hi, mpv!

What is the best approach to add "console" emulation for JS context?
Should i create a class with all required methods, instantiate it and then call aEngine.GlobalObject.ptr.DefineProperty?
Is there a way to group few methods in one variable (like console.log, console.clear) without class instance on delphi side?
PS: dll extension can provide such behavior, but i'm interested to implement console internally.

That's it, tnx.

Valuable information, thanks.

Would you like to add ReHash method directly to TRawUTF8ListHashed class?

TRawUTF8ListHashed = class(TRawUTF8List)
...
public
...
    /// manual rehashing
    function ReHash: boolean;
...
end

function TRawUTF8ListHashed.ReHash: boolean;
begin
  result := fHash.ReHash;
  fChanged := not result;
end;

Hello!

TObjectDictionary perform search faster than TRawUTF8ListHashed, while TRawUTF8ListHashed is faster for add operation.
Is it expected behavior?
Also, if max "i" value is increased up to 20 000 000, out of memory occurs with TRawUTF8ListHashed, while TObjectDictionary works.

// Add time = 2256 ms
// Search time = 0 ms
procedure TForm1.ButtonTObjectDictionaryClick(Sender: TObject);
var
  Dictionary: TObjectDictionary<String, TTest>;
  i: Integer;
  sw: TStopwatch;
  TestObj: TTest;
begin
  Dictionary := TObjectDictionary<String, TTest>.Create([doOwnsValues]);
  sw := TStopwatch.StartNew;
  for i := 0 to 6000000 do
    Dictionary.Add(i.ToString, TTest.Create(i));
  sw.Stop;
  ShowMessage('add time (ms): ' + sw.ElapsedMilliseconds.ToString);
  sw.Reset;
  sw.Start;
  if not Dictionary.TryGetValue('1500000', TestObj) or (TestObj.Id <> 1500000) then
    ShowMessage('err');
  sw.Stop;
  ShowMessage('search value time (ms): ' + sw.ElapsedMilliseconds.ToString);
  Dictionary.Free;
end;

// Add time = 809 ms
// Search time = 360 ms
procedure TForm1.ButtonTRawUTF8ListHashedClick(Sender: TObject);
var
  Dictionary: TRawUTF8ListHashed;
  i: Integer;
  sw: TStopwatch;
  idx: int64;
begin
  Dictionary := TRawUTF8ListHashed.Create(True);
  sw := TStopwatch.StartNew;
  for i := 0 to 6000000 do
    Dictionary.AddObject(i.ToString, TTest.Create(i));
  sw.Stop;
  ShowMessage('add time (ms): ' + sw.ElapsedMilliseconds.ToString);
  sw.Reset;
  sw.Start;
  idx := Dictionary.IndexOf('1500000');
  if (idx = -1) or (TTest(Dictionary.GetObject(idx)).Id <> 1500000) then
    ShowMessage('err');
  sw.Stop;
  ShowMessage('search value time (ms): ' + sw.ElapsedMilliseconds.ToString);
  Dictionary.Free;
end;

Welcome back to my blog

I updated test/demo project with TDocVariant (with TDocVariantData) use case.
Works good, TDocVariant is very powerful part of framework.. Again, not simple as pie, but check the demo and don't forget that every mORMot method well commented.

Test project updated, now looks more simple, use TObjectList serialization to serialize/deserialize TObjectList<T> objects.

Not like dependency on REST.Json.
So, i've changed test case code to use mORMot functionality for TObjectList<TCustomObjectForGenerics> serialization / deserialization.
There is memory leak in JSONToObject, when json text is not correct.

Hello!

I have simple test case, that probably can help you.
TObjectList - serialization and deserialization works as expected.
Also, i included code that should map fields, but that not works, last line of code (commented).

Personally i have question about generic version of TObjectList<T>.
With similar usage, serialized text not contain objects (that behavior included in test case), while documentation says:

  {$M+}
  TCustomClass = class
  private
    FField1: RawUTF8;
    FField2: RawUTF8;
  published
    property Field1: RawUTF8 read FField1 write FField1;
    property Field2: RawUTF8 read FField2 write FField2;
  end;
  {$M-}

procedure TForm1.Serialize();
var
  i: Integer;
  CustomClassObject: TCustomClass;
  CustomClassList: TObjectList;
begin
  CustomClassList := TObjectList.Create(True);
  // fill test data
  for i := 0 to 4 do
    begin
      CustomClassObject := TCustomClass.Create();
      CustomClassObject.FField1 := 'value ' + StringToUTF8(i.ToString);
      CustomClassObject.FField1 := 'another value ' + StringToUTF8(i.ToString);
      // TObjectList
      CustomClassList.Add(CustomClassObject);
    end;
  // Serialize and show results in memo
  MemoObjectList.Lines.Text := UTF8ToString(ObjectToJSON(CustomClassList, [woStoreClassName, woHumanReadable]));
  // Cleanup
  CustomClassList.Free();
  // Additional generic serialization test
  MemoGenericObjectList.Lines.Text := UTF8ToString(GenSerializationTest.DoSerializeTest());
end;

initialization

// TJSONSerializer.RegisterCustomSerializerFieldNames(TCustomClass, ['field1', 'field2'], ['RenamedField1', 'RenamedField2']);

I hope that should be way to debug js from vscode...
Found special extension that can successfully connect with spidermonkey debugger.
But breakpoints not work.

As i saw in firefox, spidermonkey app determines as firefox addon.
Extension page says that 3 types of addons are supported: addonSdk, webExtension and legacy.
Regarding of addon type, corresponding config file is required.
Which type of addon SyNode simulate for debugging?

Thanks! I appreciate your opinion.

Sometimes it is impossible to completely avoid logical references between entities, that mean, some kind of transactions are required (at least on application level).
Pretty hard to provide freedom for users to customize data structures, without native transactions support from DBMS...

SQLite3 support only single writer and as stated on their site, SQLite cant replace big RDBMS like MSSQL or PostgreSQL.
I expect heavy load to database system from different threads, each with own connection (probably from pool).

PostgresSQL:
Since PostgreSQL manual promise advanced JSON support, it may be a solution that gives freedom and data consistency in same time.
https://www.postgresql.org/docs/9.6/sta … -json.html
https://www.postgresql.org/docs/current … -json.html

MongoDB
I like how looks Mongo site, installer, folder structure, admin UI pretty stylish and handy, CLI access present too, (single process, while PostgreSQL use 5-6 processes to operate), many other advantages.
Only one limitation - no atomic batches, that feature was requested in 2010, hope, someday they add any kind of internal transactions that will work when requested (while in most cases it should not be used).
They really have such plans in JIRA tracker, but without dates.
I still suppose that for now, transactions may be implemented on application layer (when impossible to aggregate some data).

Do you have any negative experience with PostgreSQL?

Also, interesting thing, for storing files they implemented GridFS, does mORMot support it?
-- regarding change log and current source file, answer is - no.

Do you suggest direct access to mongoDB via SynMongoDB (is it enough updated to work with mongoDB 3.4?)
or maybe use somehow with mORMot's ORM firedac connector (which not support GridFS too)?
or would be better for now to use node.js library with all new features like GridFS (from SpiderMonkey)?