gimp-python-development

Some ideas and tools to develop Python 3.8 plugins for GIMP 2.99.4. GIMP 2.99.4 is the latest unstable pre-release of GIMP 3. It suppots Python 3, however it's documentation is rather poor, and also one thing that always annoyed me was it uses the system Python distribution and adds on top of it some libraries. Also, a GIMP plugin must run inside GIMP... So, let's hack to have a proper developing environment!

Install GIMP 2.99.4

GIMP 2.99.4 comes with pre-compiled binaries in a flatpak distribution. So, first of all if you don't have flatpak, assuming you are on a Debian-based Linux distro:

$ sudo apt install flatpak

Flatpak is a distribution system, like Apt, but focused on end-user applications and isolation. Like Docker, it isolates the compiled application in its own Sandbox, however it is not actually virtualized. Let's install GIMP 2.99.4:

flatpak remote-add --if-not-exists flathub https://dl.flathub.org/repo/flathub.flatpakrepo
flatpak remote-add --if-not-exists flathub https://dl.flathub.org/beta-repo/flathub.flatpakrepo
flatpak install --user https://flathub.org/beta-repo/appstream/org.gimp.GIMP.flatpakref

Say yes to permissions, etc. Then you will be able to run GIMP from flatpak:

flatpak run org.gimp.GIMP//beta

Understanding which Python is using GIMP

So, in order to start developing we must understand which Python executable is using GIMP. This was tricky to me at first, but it actually has a very nice outcome. As I mentioned, GIMP uses the system Python executable, if we go to the menu entry Filters > Development > Python-Fu > Python Console a window will be prompt with a Python console that will say:

GIMP 2.99.4 Python Console
Python 3.8.6 (default, Nov 10 2011, 15:00:00) 
[GCC 10.2.0]
>>> 

My main issue here to locate this Python was that I didn't have this Python installed in my system, because I'm using Python 3.9. Although, I don't have any experience with Flatpak, Docker experience suggested that was using the Python provided by its isolation environment. So, searching the Flatpak documentation I found this approach:

flatpak run --command=python org.gimp.GIMP//beta

And magic happened:

Python 3.8.6 (default, Nov 10 2011, 15:00:00) 
[GCC 10.2.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>>

Then, if we use this trick using --comand=bash instead of Python executable we will be having the shell of the Sandbox. So, when you thought it this is very good news. Because, we finally have an isolated Python environment for GIMP, a tricky one, but we have it!!

Prepare the environment

As it is thought as a minimal environment, it lacks some basic tools for developing Python code, let's install some of them. First, things firsts, let's ensure pip is installed:

$ python -m ensurepip

Once installed, update pip using pip to check pip works!

$ python -m pip install -U pip

One pip is updated let's grab at least IPython:

$ python -m pip install ipython

Now, from inside the plug-in Python Console, we can verify that we can import IPython:

>>> import IPython
>>> IPython.embed()
Python 3.8.6 (default, Nov 10 2011, 15:00:00) 
Type 'copyright', 'credits' or 'license' for more information
IPython 7.20.0 -- An enhanced Interactive Python. Type '?' for help.

In [1]: 

However, it will crash inmediadtely. This is because the plug-in messes up with sys.stdout and sys.sterr objects, needed to be able to reproduce a 'console' application. So, there is possibility to improve the Python Console plug-in to become an IPython Console plug-in working on the Python code of this plug-in and adapt the way it manages those two objects.

Getting a minimal setup to develop code

Having an IPython terminal is an advantatge over Python, because we have some extra instructions and access to shell, using !. However, what we really want is to be able to run code externally from GIMP, to be able to rapidly prototype code, specially in Python. A possible solution could be use a PyDev server - client scheme. PyDev was originally developed to be used inside Eclipse, but also PyCharm has a port of its own. Where PyCharm or the IDE will act as a server for debugging and the app will act a client of that debugger.

So, let's do that. In PyCharm go to Edit Configurations ... > + > Python Debug Server, there you will see the instructions to run this scheme in a Python project. Mines are:

• Install pydevd-pycharm corresponding to a current PyCharm version:

pip install pydevd-pycharm~=203.7148.72

• Run this two lines inside your application (a.k.a GIMP):

import pydevd_pycharm
pydevd_pycharm.settrace('localhost', port=9000, stdoutToServer=True, stderrToServer=True)

If you try this two lines inside the plug-in Python Console you will end with the same problem than IPython, these two solutions rely on sys.stdout and sys.stderr, but we are not giving up. So the main idea is to execute those two lines. So easy, let's hack into the plug-in files and put them. However, for the sake of the demo, let's use an even simpler plug-in, there is a bright new plug-ins templates for several languages in a new location Filters > Development > Goat Exercices > Excercise a goat and a python. Which file is located (from the sandbox point of view) at /app/lib/gimp/2.99/extensions/org.gimp.extension.goat-exercises/goat-exercise-py3.py and from the host point of view (and the editable one) at ~/.local/share/flatpak/app/org.gimp.GIMP/lib/gimp/2.99/extensions/org.gimp.extension.goat-exercises/goat-exercise-py3.py

This plug-in is a simple plug-in where the current loaded GIMP image (thus you need one opened), is color inverted. We are going to hook to our server using this plug-in.

We can define a function to wrap the trace:

def set_trace(): return pydevd_pycharm.settrace('localhost', port=9000, stdoutToServer=True, stderrToServer=True)

So the question to ask now is where the hell I put this call. Obviously you want it to run when the plug-in starts, there is a line where the main method is defined:

def run(self, procedure, run_mode, image, drawable, args, run_data):

One can add the set_trace() call just there, however we will not know if it is actually running, and what is worse, we will freeze GIMP! So, we need one last leap of faith and multiprocessing, more multiprocessing than faith. So, we have to add this import:

import multiprocessing as mp

And just before the while (True) loop begins, we can add:

th = mp.Process(target=set_trace)
th.start()

Before any return, this should be called th.terminate() to avoid another freeze. And it works! We are providing an isolated Python terminal to the debugger, which is running inside GIMP process, so it has access to GIMP bindings. My case:

Modify the file

There is a modified version of the plug-in in the repo, you can use it to copy it to the app folder. However, notice that you need to use you actual environment and not the Sandbox, so, assuming that you are in the repo's root folder:

$ cp plug-ins/pydev/goat-exercise-py3.py ~/.local/share/flatpak/app/org.gimp.GIMP/current/active/files/lib/gimp/2.99/extensions/org.gimp.extension.goat-exercises

Update: first version of PyDev plugin

To enable it, just declare in GIMP plugin paths your path to plug-ins folder of this repo. The way to do so is in Edit > Preferences > Folders > Scripts. The reboot GIMP and look if there is a new item under Filers > Development > Python > PyDev Client. Also notice you will need to install the plugin dependencies in the Python environment GIMP is using, in my case the Sandbox one, check previous sections of this readme. Before, launching the plug-in you should have started the PyDev server inside PyCharm.

Then you can run Python code directly from PyCharm, happy hacking.

TODO list

• Create a minimal Python plug-in to create a PyDev client. → First version available.
• Fork Python Console plug-in to be able to run IPython. Only forked, for the moment states the same functionality.
• Create a minimal Python package which depends on IPython and pydev, and other dev stuff, to install it to be able to use the new plugins.
• Create a pip wrapper to show useful info like pip list and also install PyPI packages from GIMP to GIMP Python environment.
• Define proper licence in the repo. → GNU GPL v3 added.