Add DataGateway and PaNOSC Modes

[MRichards99]

Description:
As DataGateway API and our implementation of the PaNOSC Search API will be developed in this repo, there needs to be a way of dealing with this. For the first release, I think it'll be best to add a configuration option for datagateway and panosc modes.

The datagateway mode will expose the current endpoints and the panosc mode will (upon implementing) expose the 11 endpoints for the search API. This config option will have to be made mandatory and can be done in config.py.

To do this, I need to add a config option (mode?) and use that option in main.py to branch off to the existing calls to api_start_utils.py and the new panosc function calls to setup the Flask app for that mode. To avoid api_start_utils.py from getting too long and confused, I might create a search_api_start_utils.py to have the endpoint definition and general app infrastructure setup there. At which point, it would probably make sense to rename the existing file to datagateway_api_start_utils.py.

In the future, we should look at supporting both at the same time (i.e. where a single instance can be run for both DataGateway API and PaNOSC Search API) by adding prefixes (perhaps something like /datagateway/ENDPOINT and /panosc/ENDPOINT) but that should come later.

Acceptance criteria:

• Add a way to switch between DataGateway and PaNOSC modes, which should be configurable
• Add mode config option to config.json.example
• Add test in to cover the changing of modes

[MRichards99]

[{"datagateway-api": "datagateway"}, {"panosc-api": "panosc"}]

[{datagateway-api": ""}]

no prefix (blank string) is all good as long as length of array is 1

Check that no prefix strings match across the list

{"datagateway": "datagateway-api", "panosc": "panosc-api"}

Could flip them round to avoid collisions and have multiple instances of the same API? Could require further reworking of config if you wanted to point two DG API at two different ICAT instances.

[MRichards99]

My current suggested config format:

{
  "datagateway_api": {
    "extension": "/datagateway-api",
    "backend": "python_icat",
    "client_cache_size": 10,
    "client_pool_init_size": 3,
    "client_pool_max_size": 20,
    "db_url": "mysql+pymysql://icatdbuser:icatdbuserpw@localhost:3306/icatdb",
    "icat_url": "https://localhost.localdomain:8181",
    "icat_check_cert": false
  },
  "search_api": {
    "extension": "/search-api",
    "icat_url": "https://localhost.localdomain:8181",
    "icat_check_cert": false,
    "client_pool_init_size": 3
    "client_pool_max_size": 20
  },
  "flask_reloader": false,
  "log_level": "DEBUG",
  "log_location": "path/to/datagateway-api/logs.log",
  "debug_mode": true,
  "generate_swagger": true,
  "host": "127.0.0.1",
  "port": "5000",
  "test_user_credentials": {"username": "root", "password": "pw"},
  "test_mechanism": "simple"
}

This separates specific config for DataGateway API and the search API (in the relevant objects) and keeps the general config (e.g. flask_reloader) in the same base object as before. The extension key-value pair allows configuring of the end part of the URL. Not sure if extension is the correct term but you get the idea. If this config option is present, perhaps the extensions I've used above could be used as defaults? Depends how easy it is to implement that I guess.

Having DataGateway API and search API config in separate objects means having one API or both will be easy to implement. If neither are present in the config, then the only endpoint that should be present is the openapi one.

The behaviour where optional config isn't needed (i.e. no need for client config if using DB backend for DG API) should remain.