Inheritance

Dialect and dataset definitions support inheritance to reduce duplication. Use the inherits_from key to base a new definition on an existing one and override only the properties that differ.

Dialect inheritance

If you have a dialect that is nearly identical to another, you can create a new dialect inheriting from it. For instance, suppose nextgems is derived from cmip6 but uses a different specs_file and some overrides:

TOML CONFIG

[drs_settings.dialect.nextgems]
inherits_from = "cmip6"
specs_file = ["variable", "time", "level"]
defaults = { project = "NEXTGEMS" }

When the config is loaded, the parent dialect (cmip6) is copied and then the child’s fields are merged on top. Empty tables in the child override the parent’s tables; nested dictionaries are merged recursively. Multiple levels of inheritance are supported.

Dataset inheritance

Datasets can also inherit from other datasets. Use this when two datasets share the same root or storage options but differ in defaults or dialect choice. For example:

TOML CONFIG

[cmip6-cycle3]
inherits_from = "cmip6-fs"
root_path = "/data/cmip6/cycle3"
defaults = { version = -2 }

In this case all fields from cmip6-fs (root_path, fs_type, storage_options, defaults, etc.) are copied; only root_path and defaults.version are overridden. Using inheritance reduces duplication in complex configuration files.

Note that cyclic inheritance is not allowed. If a parent dialect or dataset cannot be found the loader will raise an error.