args.ts

These extend the CLI args for tasks. Anything can be assigned to a task's args. It's just a mutable POJO dictionary. Downstream tasks will see args that upstream events mutate, unless invoke_task is called with modified args. Upstream tasks can use listeners to respond to downstream events and values. It's a beautiful mutable spaghetti mess. cant get enough The raw CLI args are handled by mri - https://github.com/lukeed/mri

Parses user input args with a Zod schema. Sets the correct source of truth for no- versions of args, to the opposite of the unprefixed versions when not included in unparsed_args. This is needed because CLI args don't have a normal way of setting falsy values, so instead the args parser mri will pass through the truthy versions of args without the no- prefix. When we declare task args schemas, we need include both versions with their defaults to get correct --help output. Parsing like this also ensures data consistency for both versions because mri only creates one. A simpler implementation could replace mri, but it handles some finicky details well.

Serializes parsed Args for CLI commands.

Parses process.argv for the specified command, so given gro taskname arg1 --arg2 -- eslint eslintarg1 --eslintarg2 -- tsc --tscarg1 --tscarg2 the command 'eslint' returns eslintarg1 --eslintarg2 and 'tsc' returns --tscarg1 and --tscarg2.

Gets all args after the first -- without assuming a command name. This is useful for tasks that want to forward args directly to a tool without requiring users to specify the tool name explicitly. Optionally strips a specific command name if present for backward compatibility.

Gets the array of raw string args starting with the first --, if any.

Parses task_name and args from process.argv using mri, ignoring anything after any --.