Prevent attempting to setup data layer on every call toget_data_layer() when no data layer's defined.

[dokterbob]

Is your feature request related to a problem? Please describe.

Currently, the get_data_layer() function attempts to set up the data layer on every call, even when no data layer is defined. This can lead to unnecessary checks and potential performance issues if the setup logic is executed repeatedly.

Describe the solution you'd like

Modify the get_data_layer() function to return early if the _data_layer is already set. This prevents re-initialization by checking if _data_layer is None before attempting to set it up. The current implementation already includes this logic:

def get_data_layer():
    global _data_layer

    if not _data_layer:
        if api_key := os.environ.get("LITERAL_API_KEY"):
            from .literalai import LiteralDataLayer

            # support legacy LITERAL_SERVER variable as fallback
            server = os.environ.get("LITERAL_API_URL") or os.environ.get(
                "LITERAL_SERVER"
            )
            _data_layer = LiteralDataLayer(api_key=api_key, server=server)

    return _data_layer

Describe alternatives you've considered

We should discriminate between there intentionally not being a data layer setup and there being a default data layer (e.g. LiteralAI). Perhaps we could add a 'null' data layer or something similar?

An alternative could be to refactor the data layer setup logic into a separate initialization function that is explicitly called once during application startup, rather than being embedded within get_data_layer().

Additional context

The current implementation already prevents re-initialization by checking if _data_layer is None before setting it up. If _data_layer is already initialized, it simply returns the existing instance ^[1]. Just when no data layer is used, it will run this logic every time get_data_layer() is called.

[dokterbob]

@dosu Would you help me fill out the full issue report based on the title and the contents of backend/data/__init__.py?

[dosubot[bot]]

To fill out the full issue report based on the title and the contents of backend/data/__init__.py, you can use the following template:

---

Is your feature request related to a problem? Please describe.

Currently, the get_data_layer() function attempts to set up the data layer on every call, even when no data layer is defined. This can lead to unnecessary checks and potential performance issues if the setup logic is executed repeatedly.

Describe the solution you'd like

Modify the get_data_layer() function to return early if the _data_layer is already set. This prevents re-initialization by checking if _data_layer is None before attempting to set it up. The current implementation already includes this logic:

def get_data_layer():
    global _data_layer

    if not _data_layer:
        if api_key := os.environ.get("LITERAL_API_KEY"):
            from .literalai import LiteralDataLayer

            # support legacy LITERAL_SERVER variable as fallback
            server = os.environ.get("LITERAL_API_URL") or os.environ.get(
                "LITERAL_SERVER"
            )
            _data_layer = LiteralDataLayer(api_key=api_key, server=server)

    return _data_layer

Describe alternatives you've considered

An alternative could be to refactor the data layer setup logic into a separate initialization function that is explicitly called once during application startup, rather than being embedded within get_data_layer().

Additional context

The current implementation already prevents re-initialization by checking if _data_layer is None before setting it up. If _data_layer is already initialized, it simply returns the existing instance ^[1].

---

This template should help in clearly communicating the issue and the proposed solution.

To continue talking to Dosu, mention @dosu.