Dev container suggestions

[eimrek]

Some suggestions for the dev container setup (happy to prep a PR if agreed):

venv.sh static tasks

As venv.sh is ran every time a container is created, static (and potentially slow) tasks are better to be moved to the Dockerfile, these include:

• setting up the venv
• Installing pip, node, rust, ipykernel, ...

Building C++ separately from python

At the moment

QDK_UARCH=native pip install -v .[all]

builds C++ and python bindings together.

But it might be nicer already here to install C++ separately and link the python build to it, which should allow for faster development.

(At the moment I'm basically doing this manually anyway and ignoring the current install step.)

[wavefunction91]

I'm generally in favor of streamlining with the dev container with two caveats:

1. The intention of the dev container is to incorporate the current version of the code into the working container. This is standard practice. That said, baking it into the Dockerfile seems antithetical. In fact, I had tossed around the idea to remove the auto build of the code entirely, leaving it as an optional add on given pre-defined commands defined within the container. That way there is no "canonical" version installed in the container by default, the developer can choose when, how, and what version of the code they install.
2. Similarly, I'm in favor of separating out the C++/python builds for the reason you state, but with the same level of flexibility / lack of environment pollution - this is harder for libraries, particularly shared, as the runtime path will influence behaviour in unexpected ways.

[eimrek]

thanks @wavefunction91

Some clarification:

1.

Sorry, I didn't mean to move the install qdk-chemistry to the docker image. I fully agree that it should be done at container creation.

But what i'm proposing is to keep only qdk-chemistry installation inside venv.sh and move other tasks (e.g. node and rust) to the image.

2.

Wouldn't it be ok to just follow the 2 steps we document:

• https://github.com/microsoft/qdk-chemistry/blob/main/INSTALL.md#building-the-c-library
• https://github.com/microsoft/qdk-chemistry/blob/main/INSTALL.md#linking-to-an-existing-c-installation
  ?

In general, i think at the moment it's not super clear how to quickly get onboarded to develop/contribute to qdk-chemistry (at least it wasn't for me), and I'd be happy to try to reduce this barrier. With the default setup, if i change any of the code, and call pip install again, it takes ages. But also, maybe best to for me to tackle this in the future, when i have more experience with the build process.

[wavefunction91]

1. Ah, gotcha. That makes sense, feel free to PR that change
2. I'm generally fine with people reading the documentation, but we'd like to make this as automated as possible. This is why I'm advocating for pre-baked bash commands to automate the process when executed in the base docker image.