feat(core): Add support for working with env files

[Tranquility2]

Fix: #687

Users should not be required to load each env var manually if they have an env file.
Added support for loading a dot-env file.

Usage:

with DockerContainer("nginx:latest").with_env_file("my_env_file"):
    # We now have a container with the relevant env vars from the file

This is an implementation of docker run --env-file <file> ...

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Please upload report for BASE (main@8f1165d). Learn more about missing BASE report.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #684   +/-   ##
=======================================
  Coverage        ?   81.27%           
=======================================
  Files           ?       12           
  Lines           ?      614           
  Branches        ?       91           
=======================================
  Hits            ?      499           
  Misses          ?       88           
  Partials        ?       27           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[sasa-buklijas]

@Tranquility2

It is great, but not sure that it will be useful, to me at least.
I will explain my workflow, because maybe I am using testcontainers-python in non optimal way and somebody can correct me.

I start testcontainers-python via pytest and have 3 containers running: redis, postgres and my own FastAPI app.
I am accessing redis, postgres from FastAPI app and from pytest directly(for simulating rest of the system)
What I have found that different IP and ports are needed if my own FastAPI app(container) is accessing redis and postgres containers or if pytest is accessing redis and postgres container.
IP and port also changes if I am using Linux(docker engine) or OSX (Docker Desktop).

Do to all this I can not use one dot-env file.

If I am doing something wrong, please correct me.

[Tranquility2]

Let me try an sort this out :)

1. Your workflow is fine, you need dynamic data that your are obtaining for previous runs, that's not the use case for env files.
2. This PR relates to the first part of your question where you raised a very interesting point regarding using dotenv and sending current/some os.environ to the Container and thats where env file can be used. (You simply reflected something deeper in the question. maybe it was not intentional)
   For example, if I would have an Redis service with some config:

# Redis Bike Commpany Demo Application: Example .env file.
REDIS_URL=redis://localhost:6379/?decode_responses=True
BIKE_INDEX_NAME="idx:bikes"
STORE_INDEX_NAME="idx:stores"
REDIS_KEY_BASE="redisbikeco"
BIKE_KEY_BASE="redisbikeco:bike"
STORE_KEY_BASE="redisbikeco:store"
FLASK_ENV=development

We would now be able to load it without evoking with_env 7 times + keeping a single source of truth for the env config.

TL;DR your current solution is great as its dynamic, we also need to support static config from file (As this is the common case for quite a few services)

[sasa-buklijas]

@Tranquility2 OK, thanks for explanation.

Is there some way, one line, to get container internal IP address ?

Only solution that I have found is:

import docker
client = docker.from_env()

# redis
redis_container_id = redis.get_wrapped_container().id
redis_container = client.containers.get(redis_container_id)
redis_container_ip_address = redis_container.attrs['NetworkSettings']['Networks']['bridge']['IPAddress']

[Tranquility2]

I assume you mean redis as the model in testcontainers, you can use this:

from testcontainers.redis import RedisContainer

with RedisContainer() as redis:
    redis_container_ip_address  = redis.get_docker_client().bridge_ip(redis._container.id)

(This should provide the same data as in the example you found)

[Tranquility2]

Some more details (if you are interested)

Usually get_container_host_ip is good enough (it is based on client.api.base_url from the docker API)
but as you can see in the code,

testcontainers-python/core/testcontainers/core/container.py

Line 141 in 4912725

# if inside_container() and not os.getenv("DOCKER_HOST") and not host.startswith("http://"):

some of the DinD code was disabled, but you can clearly see
self.get_docker_client().gateway_ip(self._container.id)
and self.get_docker_client().bridge_ip(self._container.id) are very useful for some cases.

[sasa-buklijas]

@Tranquility2 thanks

P.S.
from my experience on Linux(docker engine) I need bridge_ip if one container want to communicate with other.
On OSX(docker Desktop) I just replace localhost with host.docker.internal

[alexanderankin]

im not sure about this use case. the issue linked sets hostnames as variables. shouldn't those be constants in code for easy updating? like youre passing the same hostname in the same function (same module at least) to both set and pass to other containers. im not convinced.

[Tranquility2]

I see this is confusing, let me open a new issue :)

[Tranquility2]

@alexanderankin I've updated the related issue to #687 (sorry for the mix up) just to be extra clear this is a docker feature, we are just adding support to do the same as docker run --env-file <file> ...