Should env defined variables override .env?

[nvartolomei]

This is what I expected actually.

[ddollar]

I think this could easily lead to unexpected scenarios. Could you provide an example of this please?

[tomeon]

Although this issue is closed, it's perhaps worth mentioning that what the OP expects is dotenv's default behavior, justified on the assumption that

the deployment environment has more knowledge about configuration than the application does.

Maybe this behavior could be optional -- say, via a flag like --preserve-env? I would be happy to submit a PR adding this feature.