Open asmodehn opened 1 year ago
I recently rebased this onto the recent makefile_improvements
.
Things are working fine.
I can make develop
to work from a project-based venv.
I can also build a wheel and install it via pip:
$ pip install dist/opengeode-4.0.2-py3-none-any.whl
Processing ./dist/opengeode-4.0.2-py3-none-any.whl
Requirement already satisfied: pygraphviz in /home/alexv/.local/lib/python3.10/site-packages (from opengeode==4.0.2) (1.10)
Requirement already satisfied: pyside6 in /home/alexv/.local/lib/python3.10/site-packages (from opengeode==4.0.2) (6.4.2)
Requirement already satisfied: shiboken6==6.4.2 in /home/alexv/.local/lib/python3.10/site-packages (from pyside6->opengeode==4.0.2) (6.4.2)
Requirement already satisfied: PySide6-Addons==6.4.2 in /home/alexv/.local/lib/python3.10/site-packages (from pyside6->opengeode==4.0.2) (6.4.2)
Requirement already satisfied: PySide6-Essentials==6.4.2 in /home/alexv/.local/lib/python3.10/site-packages (from pyside6->opengeode==4.0.2) (6.4.2)
Installing collected packages: opengeode
Attempting uninstall: opengeode
Found existing installation: opengeode 4.0.2
Uninstalling opengeode-4.0.2:
Successfully uninstalled opengeode-4.0.2
Successfully installed opengeode-4.0.2
One thing I don't really like is the fact that I have to activate the virtual env from the makefile at every line there (make will create a new shell process for each line, to mitigate side-effects damage, and therefore we loose the virtualenv).
Which is why I would like to move the "non-obvious" steps outside the makefile if possible (in a later PR):
pyside6-rcc opengeode.qrc -o opengeode/icons.py
manually, or automate it in a different way, and put icons.py in repo ?...
Let me know if you have any idea about those, or if you prefer the PR as is now.One thing I don't really like is the fact that I have to activate the virtual env from the makefile at every line there (make will create a new shell process for each line, to mitigate side-effects damage, and therefore we loose the virtualenv).
Why don't you use &&
between each command to avoid this ?
Also I am not sure to understand - the install
rule depends on develop
. Does it mean that it will be installed twice - first in the venv then on the system ?
Which is why I would like to move the "non-obvious" steps outside the makefile if possible (in a later PR):
- dealing with antlr3_python3_runtime as a proper pip dependency, somehow...
Yes the solution would be to upload this runtime on PyPi, making it available properly to everyone
* doing `pyside6-rcc opengeode.qrc -o opengeode/icons.py` manually, or automate it in a different way, and put icons.py in repo ?...
In general I don't want to put generated code on the repo
One thing I don't really like is the fact that I have to activate the virtual env from the makefile at every line there (make will create a new shell process for each line, to mitigate side-effects damage, and therefore we loose the virtualenv).
Why don't you use
&&
between each command to avoid this ?
Sure, I could do this, but it is perhaps less clear which command will embed which shell (and how errors will be back-propagated), and anyway I see it as a hint that doing so many steps in the Makefile is not the way to go about it...
Also I am not sure to understand - the
install
rule depends ondevelop
. Does it mean that it will be installed twice - first in the venv then on the system ?
Correct.
Note I only install it on the venv to be able to generate the resources with pyside6-rcc
, so that the file is present when installing afterwards...
Which is why I would like to move the "non-obvious" steps outside the makefile if possible (in a later PR):
- dealing with antlr3_python3_runtime as a proper pip dependency, somehow...
Yes the solution would be to upload this runtime on PyPi, making it available properly to everyone
For later reference, there seems to be a way to bring it along from opengeode setup.py directly, called direct references : https://peps.python.org/pep-0440/#direct-references This would allow to install it when installing opengeode, without having to make it a pip package by itself. Disclaimer, I havent tried that feature yet.
* doing `pyside6-rcc opengeode.qrc -o opengeode/icons.py` manually, or automate it in a different way, and put icons.py in repo ?...
In general I don't want to put generated code on the repo
Here is an interesting option: do it in the setup.py : https://gist.github.com/ivanalejandro0/6758741 This way we could hook it to existing commands. Ref: https://setuptools.pypa.io/en/latest/userguide/extension.html#customizing-commands or maybe even via a build sub-command since we just deterministically build a file here: https://setuptools.pypa.io/en/latest/userguide/extension.html#supporting-sdists-and-editable-installs-in-build-sub-commands
- dealing with antlr3_python3_runtime as a proper pip dependency, somehow...
Yes the solution would be to upload this runtime on PyPi, making it available properly to everyone
For later reference, there seems to be a way to bring it along from opengeode setup.py directly, called direct references : https://peps.python.org/pep-0440/#direct-references This would allow to install it when installing opengeode, without having to make it a pip package by itself. Disclaimer, I havent tried that feature yet.
What about python3 -m pip install --user https://download.tuxfamily.org/taste/antlr3_python3_runtime_3.4.tar.bz2
?
That seems to work as well.
This depends on #89
Adds a target to the makefile, to deploy to a project based venv. This allows to gather dependencies into a
requirements.txt
file.Also installing in a clean virtual env from the wheel revealed missing dependencies in
setup.py
Possible future improvements: