Hi guys, I have a question around building a Docker image from a python source. So far I have understood that if I want to publish a python library from a python source, I should be using

python_distribution()

target and for a Docker image it should be

docker_image()

target. Throwing in gRPC and Protobufs in the mixture and there I know that

python_distribution()

target type will resolve the proto dependency at the time of executing

package

goal. Would the target type

docker_image()

do the same? Is

package

the right goal for my use case here as in to build a Docker image?

This might be helpful. https://blog.pantsbuild.org/pants-pex-and-docker/ Generally

pex_binary

is the recommended target for docker image.

This might be helpful.

https://blog.pantsbuild.org/pants-pex-and-docker/

Generally

pex_binary

is the recommended target for docker image.

Ok, what would be the workflow then? Adding the dependency of

pex_binary()

target to the

docker_image()

target?

You don't need to explicitly add dependency on

pex_binary

. It's from the docs.

When you COPY PEX binaries into your image, the dependency on the pex_binary target will be inferred, so you don't have to add that explicitly to the list of dependencies on your docker_image target.

https://www.pantsbuild.org/docs/docker

Yeah, that’s why I was looking for building a Docker image from

python_distribution()

.

It has its cons too like if you have local python distribution (1st party) on which your

python_distribution

depends on then you will have to host them somewhere and then pass the necessary secrets to docker build so that it gets installed inside the image.

pex_binary

bundles all 1st party and 3rd party dependencies without much from your side.

Ok, but right now I am just focused on getting the

package

goal working and see the results. I can think about the target types later once I have a clear picture in mind and I proceed towards implementing the learnings in prod.