Use multi-stage docker builds

[smitpsanghavi]

• Update Docker build to multi-stage build.
• Use python's venv feature to isolate the dependencies to one location.
• Add a .dockerignore file to ignore files not required within the final image.
• Cache python dependencies locally on file-system for faster local development.
• Linted README.md

[pranjal-joshi]

@smitpsanghavi

Thanks for your contribution. 🎉

Let's discuss/resolve few comments that I've added before merging this.

Also, could you provide the image size after multistage build?

[smitpsanghavi]

This is what I see in my Docker Desktop locally.
Your image vs the one built using this Dockerfile.

The size on Docker hub for v2.24 is 2.04 GB as it is compressed. I don't know much the updated one's size would be after compression.

[smitpsanghavi]

I don't see any comments of yours anywhere. Where did you add them?

[pranjal-joshi]

@smitpsanghavi Let me try creating a docker image with the modified dockerfile and compare the size on Dockerhub as well

The comments are added as a code review which might be visible in the 'Files' section of this PR

[smitpsanghavi]

Nothing in the conversations section of the PR.

Did you submit the comments after reviewing them?

[pranjal-joshi]

@smitpsanghavi I was not able to observe a significant reduction in size either on my end or on dockerhub repo after a push. Seems like dockerhub image size is increased after multistage.

[smitpsanghavi]

With regards to the reduction in size, I think something went wrong while building the image in your machine.

I analysed your image and saw that when the command COPY . . is run, the whole git repo (including the complete git history) was added into the image that you pushed. I also see a few other files/folders in the image which I explicitly ignored via .dockerignore. I am not sure why .dockerignore did not have any effect on your system. Hence, you did not see any size reduction.

Here's a comparison:

Your image

My image

I would suggest checking why .dockerignore didn't work for you and fixing it. Once done, you should see massive improvements in the pushed image.

[pranjal-joshi]

@smitpsanghavi Got it. I quickly cloned the updated docker file on my local for testing without creating any other files. Let me try building this branch itself.

[pranjal-joshi]

There is a substantial reduction in the size after a multistage build! Merging this change after running a functional test @smitpsanghavi

[smitpsanghavi]

I didn't expect a 50% reduction on Docker Hub! 😲
The magic was in the .dockerignore actually.
Multistage also helps and is generally good practice to help your docker image cleaner by removing build time dependencies.

You can reduce this further if you remove any other unnecessary apt-get installations (e.g. git). I'll leave that to you.