Extension Manager (ExtMan)

[mnesarco]

Hi Friends, vocx suggested to create a dedicated thread here for the Extension Manager.

- https://github.com/mnesarco/FreeCAD_ExtMan

This is a work in progress possible replacement for the AddonManager. It requires qt5 and python3.

screenshot_4.png (108.54 KiB) Viewed 3711 times

screenshot_5.png (103.58 KiB) Viewed 3711 times

screenshot_6.png (168.61 KiB) Viewed 3711 times

screenshot_7.png (269.48 KiB) Viewed 3711 times

[vocx]

This project started from the Add ons manager - development and bugs topic thread.

In particular, it is mentioned around pages 10 to 20. To avoid polluting that thread, which refers to the current Addon Manager, this new thread was created.

[vanuan]

I'm excited to see more and more stuff using the web technology for UI instead of Qt. This is the only way forward!

A few questions:
1. Are you generating html for each action from python or using dynamic javascript pages?
2. Which protocol do you use for interaction between html and FreeCAD? Some sort of http API?
3. Are you familiar with React and GraphQL (developed by Facebook)?
4. Have you seen my POC?
https://github.com/Vanuan/FreeCADjs
https://forum.freecadweb.org/viewtopic.php?f=34&t=48611

[mnesarco]

1. Are you generating html for each action from python or using dynamic javascript pages?

html is generated dynamically using a python template manager built by me for this project.

2. Which protocol do you use for interaction between html and FreeCAD? Some sort of http API?

All communications are done directly using QtWebEngine APIs. Not http involved in that part.

3. Are you familiar with React and GraphQL (developed by Facebook)?

Yes I am a ReactJS developer, look at this: https://marzguitars.com/marz-designer/
But I have not used React in this project to keep it simple.

I am also a GraphQL developer, look at this: https://mnesarco.github.io/blog/2020/02 ... s-mutation
but I see no relation with this project.

4. Have you seen my POC?
https://github.com/Vanuan/FreeCADjs
https://forum.freecadweb.org/viewtopic.php?f=34&t=48611

It looks interesting, but I have had no time to look at it.

[vanuan]

All communications are done directly using QtWebEngine APIs. Not http involved in that part.

Interesting. How does it work exactly? So you don't catch DOM Events, but rather use <a href=""> everywhere?
Or there is a way to handle DOM events directly from Python?

Yes I am a ReactJS developer, look at this: https://marzguitars.com/marz-designer/
But I have not used React in this project to keep it simple.

Nice!
I'd say using React is simpler than developing a custom html template engine

I am also a GraphQL developer, look at this: https://mnesarco.github.io/blog/2020/02 ... s-mutation
but I see no relation with this project.

Cool! Do you think it is viable to do the same for FreeCAD API? That is expose it as a GraphQL endpoint?

It looks interesting, but I have had no time to look at it.

Got it. I thought we could join our forces and push Web UI integration into FreeCAD together. It seems that Extension Manager fits into Web UI vision nicely. There are at least 4 people on the forum who are interested in using HTML/JS/CSS for building parts of FreeCAD UI.

[vanuan]

Ah, I see it uses a custom protocol scheme instead of http:

Code: Select all

 extman://...

So it uses a custom protocol. Still, pretty much similar to http. I don't get why you said that no protocol is being used.

[mnesarco]

Ah, I see it uses a custom protocol scheme instead of http:

Code: Select all

 extman://...

So it uses a custom protocol. Still, pretty much similar to http. I don't get why you said that no protocol is being used.

I never said that. I said no http is used.

[mnesarco]

Nice!
I'd say using React is simpler than developing a custom html template engine

I decided to not use react because most of the freecad extensions are python centric, and my goal was to make it easy for other freecad developers to participate... You will get a lot of resistance if try to change too many things too fast.

[vanuan]

I decided to not use react because most of the freecad extensions are python centric, and my goal was to make it easy for other freecad developers to participate... You will get a lot of resistance if try to change too many things too fast.

I'd say the main benefit of introducing a well known technology is attracting more developers rather than trying to mimic a workflow of existing developers. With custom html template engine and custom protocol I'd say you limit developer participation even further without any clear benefit.

Do you have any intention on upstreaming QWebEngine-based MDI window into the main FreeCAD repository?

[mnesarco]

Cool! Do you think it is viable to do the same for FreeCAD API? That is expose it as a GraphQL endpoint?

In an ideal world, we could have an extensions portal with graphql endpoint... and indexing, automated quality tests, compliance tests, etc... but in the real freecad world it will not happen any soon. I have learned in this forum that things move very slowly here, small improvements over existing things are more appreciated than big changes. Sometimes it is frustrating but is the way it is. The first step to improve Extension management is to gain consensus about many things like metadata, file format, naming conventions, etc... And even if you gain consensus, you willl have to maintain backwards compatibility with the past and it makes things very hard to maintain, and if you are the proponent of the change, you are the maintainer.

So, my personal approach is to take an easy path, improve things step by step, in slow motion. if you are following the discussions about release 0.19 stable you know that it will be released some day in one or two decades, so we have time