Configuration

Every Frost program reads one YAML configuration file at startup. It supplies the things that would otherwise be hard-coded in .lf files: where each component's data model lives, how loudly each one logs, and any per-component parameters.

Keeping them here means the same Lingua Franca sources can be run against different plants, different models or different logging levels without editing code.

Where the file is found

Frost reads the path from the FROST_CONFIG environment variable, defaulting to resources/frost_config.yml:

FROST_CONFIG=resources/frost_config.yml bin/Main

If the file does not exist, Frost falls back to built-in defaults (time_precision: NSECS, logging_level: WARNING) and every component uses an empty data model path — which then fails when FrostNode tries to load it. A missing configuration file is therefore worth ruling out first when a program stops immediately at startup.

Paths inside the file are resolved relative to the process's working directory, not to the file's own location.

The file

resources/frost_config.yml

time_precision: NSECS
logging_level: INFO

reactors:
  frost_link:
    logging_level: WARNING
    parameters:
      data_model_path: "resources/data_model/link.yml"
    reactors:
      message_filter:
        logging_level: WARNING

  device_1:
    logging_level: INFO
    parameters:
      data_model_path: "resources/data_model/device_1.yml"
      peer: "device_2"
    reactors:
      message_filter:
        logging_level: WARNING

Top level

Key	Meaning
time_precision	The unit logical timestamps are formatted in. One of NSECS, USECS, MSECS, SECS, MINUTES, HOURS.
logging_level	The default level for every component. Standard Python names: DEBUG, INFO, WARNING, ERROR.
reactors	Per-component sections, keyed by component name.

Per component

Key	Meaning
logging_level	Overrides the global level for this component.
parameters	Entries set as attributes on the reactor before its own startup reactions run.
reactors	Sections for reactors contained in this one.

Lookup follows the component's name as a dotted path, which is why a contained reactor nests. Every Frost component contains a MessageFilter named <component>.message_filter, and it is the usual thing to quieten — it logs every message it sees at DEBUG.

parameters

Entries under parameters are applied with setattr, so they can set any attribute the reactor has — and any it does not.

data_model_path is the one every component needs. It is how FrostNode finds its YAML model.

Anything else is yours. Declare a state variable with a placeholder, supply the real value here:

reactor Device extends FrostReactor {
    state peer = ""          # from parameters.peer
}

Setting a key the reactor does not have still works, and logs a warning:

Parameter peer not found in reactor device_1. Creating dynamic parameter peer = device_2.

Declaring the state variable is worth it: it documents what the component expects, and keeps the log clean.

Timers and widths cannot be set here

parameters are applied in a startup reaction, which runs after the reactor has been built. Anything used to construct the reactor — a timer period such as update_interval, or a multiport width — must be passed at instantiation instead:

link = new FrostLink(name="frost_link", width=3)

Names must agree

Three names have to line up for a component to be reachable:

1. the name passed at instantiation — new Device(name="device_1");
2. the key under reactors: in this file — device_1:;
3. the name: field of the data model that component loads.

The first two must match or the component gets no configuration, and will fail to load a data model. The first and third must match or the component registers under one name while everyone addresses it by another — because FrostNode reports the data model's name as the component's identity.

Logging

self.logger is a standard Python logger, and Frost installs a formatter that prefixes each line with the current logical time in time_precision units. That is what makes a Frost log readable: entries are ordered by simulated time, not by when they happened to be flushed.

Turn a single component up without drowning in the rest:

logging_level: WARNING
reactors:
  device_1:
    logging_level: DEBUG