[Enhancement proposal] Configuration profiles

[aitorarjona]

Currently, Lithops can be configured:

• by a config YAML file in ~/.lithops/config
• by ENV vars
• by directly passing configuration path or dict to a FunctionExecutor constructor

The standard and most convenient is using a configuration file. The problem with having only one config file is that, for a multi-cloud configuration, all parameters reside in the same config file, potentially being a mess.

I propose to create configuration profiles. The "default profile" is the ~/.lithops/config file, but users could create new config files with different names (e.g. ~/.lithops/aws_lambda_testing), and select the appropraite profile in the FunctionExecutor constructor (FunctionExecutor(config_profile="aws_lambda_testing")).

WDYT @JosepSampe

[JosepSampe]

Yes, good idea @aitorarjona
Do you have an example of what the configuration file with profiles should look like?

[aitorarjona]

The idea is to keep ~/.lithops/config file as the default profile for backwards compatibility, which will look the same as always, for example:

lithops: 
    backend: localhost
    storage: localhost

localhost:
    worker_processes: 1

for testing.

Another profile for, for example, IBM Cloud in file ~/.lithops/ibm would look like this:

lithops: 
    backend: ibm_cf
    storage: ibm_cos

ibm_cf:
    endpoint : ...
    namespace : ...
    api_key : ...

ibm_cos:
    storage_bucket: ...
    region : ...

and to select this configuration profile:

FunctionExecutor(config_profile="ibm")