'framebuf' display simulation

[fma38]

Hi!

Is there a graphical library available for Micropython Unix port ? Like tkinter/PIL...

I would like to simulate a framebuf display (like SSD1306), so I can develop and test a GUI lib without using a real microcontroller/screen/buttons...

Thanks,

Frédéric

[2dof]

Try pygame as main "display engine" for create simulator, or try luma.emulator

edit: luma.oled

[fma38]

Thanks for your answer! luma.emulator looks nice...
But how do I install pygame or luma in the micropython unix port?

[2dof]

sorry, my mistake, to I thought that You are looking for some emulator for python (not micro python) Rasspbbery pi platform, or even in Macos/win.

I used luma for oled 0.91 display emulation for design menu system for esp32. ; python pynput.keyboard for keyboard (buttons) interaction ( or just call callback function for keys - even better since all can be done from program).

[fma38]

Well, my goal is to work on a micropython menu/gui lib which uses displays like ssd1306 and similar, which are based on framebuf. That's why I wanted to use micropython unix port.

But I may be able to use regular Python, by emulating the entire framebuf module...

[peterhinch]

There is no support for this way of working. You'll have to use real hardware or write your own support.

I guess you could write an emulator using (e.g.) pygame and CPython, but then you'd need to write a FrameBuffer class for CPython (the MP class is written in C). A nontrivial task. I'm not convinced an emulator would save much time in practice. With the right tools (mpremote, rshell) development with real hardware can be quite quick.

[2dof]

luma has a lot of SSD family drivcers simulators, in my case i used to emulate ssd1306 oleds 0.91/0.96 inch do design menu system for esp32 micropython, but all in python (luma was deeded to all fancy 'icons' drawnings).

Menu/GUI system can be 'split' for : display rendering and for interaction with peripherals ( buttons, etc) -> all this part need to be implemented as some display rendering functions and some callback functions for interaction ( callback "connected" to button action). In simplest way we do not even need physical button, just call a callback function responsible for assigned button action.
Most of work is just some classes implementing and can be done in python (just need to know difference betwen micropython-vs-python) for faster R&D .

edit. I forgot to mention than luma will not work on windows system ( no ioctl).
other method for modelling oled display is using python PIL lib

[fma38]

The problem is to have acces to the hardware: I would like to be able to develop from anywhere, when I have some free time (in train, for example)...
Is it complicated to recompile a micropython module written in C for regular Python?

[peterhinch]

The whole thing is a substantial task. Rewriting C code in Python is straightforward but time consuming. Then you have to add the pygame stuff.

I would compare the effort to buying a display which supports a Pico plugged in the back and mounting it in a box. Plug the USB lead into your laptop and develop anywhere. That is my approach, but the choice is yours.

[fma38]

Yes, I'm not sure to want to dig into C that much!
As you say, it is more fun to build a real hardware stuff ;o)
Thanks,

[peterhinch]

The file modframebuf.c is quite substantial. While an emulator would be an interesting project it would take quite a lot of writing and testing. If the emulation wasn't spot-on it would be more trouble than it's worth.

My view of MP development is to use real hardware with simple command-line tools.

[rkompass]

I'm assuming you use the Linux OS. The framebuffer memory is a chunk of RAM which is pixel-wise arranged.
You may query its address from within MP.
I naively think it wouldn't be too difficult to write a C-program using the Cairo library, for example, that reads this chunk of RAM, re-codes the pixel values, if necessary, and displays them in a window. Including permanent refreshes, of course.
But I'm not sure. These would be different processes. Can they access the memory of each other?

[fma38]

It works great, thanks!
I need to dig into your examples, but I'm pretty sure it will be very easy to adapt them to my use case.
Again, great job!

[mattytrentini]

That's awesome work @russhughes, thanks for your efforts! Could definitely have used this in the past.

Not to take anything, at all!, away from your excellent work, but it sparked a few (some old) thoughts and I thought I'd share:

• It would be interesting to see if it were possible to make this a native module to reduce the build friction and ease deployment.
• I wonder if this could also be implemented with ffi. Could possibly be written entirely in Python. Even easier to deploy.
• I've wanted a similar solution, but to render the output to an HTML canvas, with data transferred over a web socket. The benefit would be that it could also be used easily from within a container.

[fma38]

+1 for native module!

[russhughes]

@mattytrentini @fma38 I'll look into it.

[russhughes]

@mattytrentini @fma38 I don't see a way to link an external library like SDL2 with a native .mpy module without modifying MicroPython. I'm temped to try adding dlopen, dlsyn, dlclose, and dlerror to the mp_fun_table to see if that is a way around natmod linking limitations.

[massimosala]

Hi

I was intrigued by this problem and it is a feature useful to me too.
So, between Monday and yesterday, I develop a virtual driver.

At the moment I use to replace the real ssd1306 driver.
It supports mono and color displays, of any size.
To not overload the micro memory, the framebuffer is created on the web page :-)

So it works only with web applications, I mean micropython programs with a web server.

It consists of two files:

• the micropython module, about 5 Kb (1.3 Kb compiled to MPY);
• the javascript module (rendering the display on the web page), about 10 Kb.

I am thinking to publish it as open source, but I have to improve some features.
Any of you want to be a beta tester?
Write to [email protected]