OK, what looks better. The OLD tcl-oo or the NEW tcl-oo?

The NEW tcl-oo is:

much EASIER to use than the OLD tcl-oo
can be compiled into pure C code using the TCL compiler
follows the "command-first" philosophy of Tcl
can be used from Tcl 8.0...
much faster than the OLD tcl-oo

https://www.facebook.com/permalink.php?story_fbid=pfbid02s4Wo3VokUzCXfuXzqdMepTcrhDtWHghuUggQ7Hsr8BtvbaZPZrtqAyvQ8V7733GAl&id=100069563501101

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

libmyoo initial release - project setup, coding, testing and documentation.

libmyoo is an attempt to rethink the topic of Object-Oriented-Programming (OOP) in TCL, offering a maximum of
Object-Oriented-Programming (OOP) functionality with minimal effort.

The aim is to offer an identical API via both C and TCL, taking into account the integration in the meta-code-compiler
(META-COMPILER) as well as in the TCL-COMPILER.

project page including code -> http://thedev.nhi1.de/theLib/main/index.htm

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

NEWS FROM: 23 Jan, 2025

Adding the libmyoo library as C/binary code and revising the API.

The performance-critical part was rewritten in C, whereby the TCL and C implementations are equal and interchangeable.
The validation of the myoo extension was carried out using tcltest and thus the API was finally defined at the same time.

Now the time has come: myoo is not only available as tcl-only but also as c-tcl-only.

http://thedev.nhi1.de/theLib/main/index.htm

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

something for tcl internal specialists ... who has the fastest 'my'

http://thedev.nhi1.de/theLib/main/myoo-performance-variants-my.htm

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

I'm still struggling with myself as to whether "myoo" should be reference-based or namespace-based.
after doing some research with the tcl-only API, I've now switched to C. Unfortunately, the public namespace C API is rather
weak and only "string" based, which has now prompted me to switch to the semi-public "Int" API, where at least there is usable
"namespace" support. After the first analysis, I can now say that "namespace-based" is currently faster.

package require libmyoox

1.0

::myooX::ClassN ::MyClassN {
proc MyClassN {myNs num} {
namespace upvar $myNs my my
set my(num) $num
}
proc get {myNs} {
namespace upvar $myNs my my
set my(num)
}
}

::MyClassN::cls

::myooX::ClassN ::MyClassR {
proc MyClassR {myRef num} {
upvar $myRef my
set my(num) $num
}
proc get {myRef} {
upvar $myRef my
set my(num)
}
}

::MyClassR::cls

set ns1 [::myooX::NewN ::MyClassN 1]

::MyClassN::MyClassN-1

set ref2 [::myooX::NewR ::MyClassR::cls 2]

::MyClassR::MyClassR-1::my

::MyClassN::get $ns1

1

::MyClassR::get $ref2

2

time {::myooX::NewN ::MyClassN 1 } 10000

10.244 microseconds per iteration

time {::myooX::NewR ::MyClassR::cls 2 } 10000

15.086 microseconds per iteration

time {::MyClassN::get $ns1 } 10000

1.6015 microseconds per iteration

time {::MyClassR::get $ref2 } 10000

2.4201 microseconds per iteration

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Array versa Namespace
=====================

‍I struggled with whether myoo should be array-reference-based or namespace-reference-based.

After doing some research with the tclmyoo TCL-API and libmyoo C-API I switch to namespace-reference-based

The following reasons led to the switch to namespace-reference :

- With TclGetNamespaceFromObj from the tcl internal Int API, a very powerful tool is available for resolving a namespace.
- It has been confirmed that "namespace upvar NS my my" is significantly faster than "upvar NS::my my".
- To call a sub-method, the local-reference no longer needs to be resolved because the instance-namespace is always used
for the global namespace (::).
- An external-pointer can also be attached to the tcl internal Tcl_Namespace* pointer, which speeds up the
All-Language-Compiler (ALC) compiler massively.
- The code becomes much smaller and simpler by switching to the namespace-reference, so that I can proudly say that myoo is
by far the smallest Object-Oriented-Programming (OOP) in the world.
- Even with the tcl-only version, tclmyoo is faster than tcloo with over ten years of development work.

read more at: http://thedev.nhi1.de/theLib/main/index.htm

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

First "sign of life" from the merger of "libmyoo" with the "ALC-compiler"

TCL now has a new OO subsystem where the class is defined either in TCL or in C/C++, which means an enormous gain in flexibility
depending on the application. With "theLink" your class is not only a LOCAL part of your application, but connects when required
with the shadow class in another application in another programming language locally or somewhere else in the network.

https://www.facebook.com/permalink.php?story_fbid=pfbid06TEVHK2eTdWA8BKGDhbTnxdgU6VCKiikfJLjpHbTP2MQywKC6YNsyrEGPStiaD4zl&id=100069563501101

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

# This causes a CRASH if after unloading the “atlmkkernel-library” a namespace with “Tcl_Namespace->deleteProc”
# is still active that points to the deleted "atlmkkernel" library.
# -> By default, when "unloading" ONLY the class-NS was deleted, BUT now also the instance-NS.
# Deleting the instance belonging to a class is more complicated than you think, because NOT ONLY the instance
# directly belonging to the class must be deleted, but also all instances that have this class as a base class.

https://www.facebook.com/permalink.php?story_fbid=pfbid0qA1aDKaiUBMAjButCH62oR3LDe78Djg1yPyxHtt1WYETkYs4NSXLpTeAaBA9n8kUl&id=100069563501101

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

This example looks simple at first, but differs in detail.

Where tclOO internally falls back to "c_str", "myOO" consistently uses "Tcl_Obj*" as a representation. **Strict** care was taken
to ensure that a "Tcl_Obj*" is not **repeatedly** constructed from a "c_str" so that an already resolved reference is
consistently reused.

https://www.facebook.com/photo/?fbid=938449471817161&set=a.805828751745901

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

On 10.01.25 09:43, aotto1968 wrote:

OK, what looks better. The OLD tcl-oo or the NEW tcl-oo?

The NEW tcl-oo is:

much EASIER to use than the OLD tcl-oo
can be compiled into pure C code using the TCL compiler
follows the "command-first" philosophy of Tcl
can be used from Tcl 8.0...
much faster than the OLD tcl-oo

https://www.facebook.com/permalink.php? story_fbid=pfbid02s4Wo3VokUzCXfuXzqdMepTcrhDtWHghuUggQ7Hsr8BtvbaZPZrtqAyvQ8V7733GAl&id=100069563501101

UPDATE myoo → Optimization is still in progress, but the first results show that the new "namespace-based" TCL-OO (myoo)
extension is 20% faster than the traditional "command-based" TCL-OO (oo) extension, with python-oo still being 20% ​​faster than
the new TCL-OO (myoo) and 40% faster than the traditional TCL-OO (oo)
Purely strategically, however, the new TCL-OO extension (myoo) has the advantage that the "brutal-force" compiler "Tcl2C"
accelerates the "myoo" code to such an extent that we can now talk about "C speed with TCL" and no longer just "PYTHON speed."

→ https://www.facebook.com/permalink.php?story_fbid=pfbid0315uLRXFuajwF5RdBBHEo8XRsHc3q8yQ9knpZWzWTq7Z1q7VBgB3QFUsS46mox8X8l&id=100069563501101

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)