Component initialization in constructor vs separate configure function

[csmith608]

Pattern

Explain why the pattern in F Prime breaks with the standard C++ pattern "RAII" to use configure functions rather than initializing the component in the constructor.

Rationale

Explanation that was provided to me:
In F Prime we do a configure call because one-argument constructors work much better with FPP as FPP can then instantiate the component without custom code to supply arguments. If you change the interface past a single argument, then FPP cannot automatically construct it for you and this would require require phase-code to fix that. Initializing a component in the constructor uses the standard C++ pattern "RAII" which says that a class holding a resource should be invariant over the life of the class (e.g. from construction to destruction). https://en.wikipedia.org/wiki/Resource_acquisition_is_initialization . This is inconsistent with a deferred configure call. While we like RAII patterns a lot, they are inconsistent with a Global singleton (i.e. an F prime component as normally constructed in global space) requiring non-global input (e.g. CLI arguments). For this and the above reason of zero-arg constructors, we use the deferred configure pattern.