Functional API is awesome - the code is easy to read, parallelisation can be achieved with asyncio.gather() etc. Branching is intuitive and “right there” instead of having to constantly jump between what was done in add_edge in another part of code vs what the node is returning and if it’s all still reasonable/expected etc. Also while rearranging things, type checkers and static analysis immediately help. For graph API if node-x expects foo to be computed/populated we must ensure that some node...