Support env file

[benlangfeld]

Replaces #481 and #479 addressing the issues raised in those tickets.

[AnalogJ]

I would love to see this merged into fig. I know its only been a few days, but any eta on when that might happen? Environment management is a bit of a pain-point for me right now

[benlangfeld]

I'd also obviously love to see this merged. Feedback for improvements is welcome and I'll get them done as fast as possible to make this merge-ready.

[cevaris]

+1

[dnephin]

I think using self.options here would mean that override_options is missed. I guess this doesn't technically break anything right now, but I don't think the way the -e param is implemented for fig run is correct, and this would make it harder to fix (it should be using override_options instead of setting service.options).

We may also want to add a command line arg for --env-file at some point.

Could you make this take a param of options instead? This would also have a nice side benefit that self would be unused, and the method could be moved off the class to just a module level helper.

[dnephin]

Thanks for the PR! I think this is looking good, I just left a few suggestions. A git squash on some of the commits would be good too, once it's all ready.

[benlangfeld]

I'll address your suggestions tomorrow and then request another round of review :)

[benlangfeld]

I think I've now resolved all of your suggestions. I'll leave the commits separate for one more round of review, then squash for merge later.

[mbleigh]

👍 this would be great

[Jasperswaagman]

+1

[dnephin]

LGTM (with a squash)

[marksteve]

🎉

[benlangfeld]

Rebased as requested. Merge-ready.

[dnephin]

hmm, looks like the latest test run hit some errors, not sure if that was a problem on werkers end or what happened.

I've put this into the milestone for the next release.

[benlangfeld]

Now this is ready :)

[benlangfeld]

@dnephin Could we get this merged? It's being discussed as a failing of Fig over at moby/moby#9459 (comment).

[dnephin]

We need another LGTM from a fig maintainer. I believe they are busy with dockercon EU, so it may be a week or so.

@aanand @bfirsh when you have a minute

[bfirsh]

I wonder if we could support the normal use case where it isn't an array? Should just be an isinstance(..., list) right?

[benlangfeld]

Done :)

[bfirsh]

Looks awesome, thanks @benlangfeld! Needs a rebase though.

[benlangfeld]

Rebased and ready for merge :)

[bfirsh]

LGTM

[bfirsh]

Thanks @benlangfeld!

[marksteve]

🎉

[eudaimos]

Looking at the code, it appears as if env_file is only an option from fig.yml, is this correct?

I've been using the --env-file option to docker run to choose an environment file for "dev", "qa", "prod." Is it possible to extend this as an option to fig up and fig run which would make it act exactly like it does for docker run?

[benlangfeld]

You are correct @eudaimos. It would be possible to add this as a command-line override as suggested by @dnephin here. Do you think you could add this?

[eudaimos]

Ha, didn't see that part of the earlier comment. I can give it a try when I free up, but it would be hacky at best and take me a while to get to testing since I've never written anything in python before. @benlangfeld if there's someone else with python experience that wants to take it on before all of that happens, I'd be grateful.

[benlangfeld]

I'm in precisely the same position as you @eudaimos. This is the first Python code I've ever written :)

[geowa4]

Any idea when this will be delivered?

[abesto]

Looking forward to this a lot. Any chance of a minor release soon?

[johanhaleby]

I'd also be really happy for this. This is a showstopper.

[petebrowne]

👍 Our team is waiting for this one as well.

[Starefossen]

+1 we are waiting on this too!

[dnephin]

This feature is available in the latest docker-compose RC releases https://github.com/docker/fig/releases