Not-unrelated to #71 and Flow-IPC/ipc#85 -- probably should be looked-at at the same time:
Flow-IPC (generally speaking, written after Flow, except for some parts that were written in parallel) is pretty good about this (possibly outside non-ipc::session::shm SHM-jemalloc-land, but that's orthogonal), but Flow might be less so, or less consistently so:
Have a to-string output for all object types. So far we've been using ostream<< for this. (Maybe also look into how fmt deals with this, as it might be relevant for #78 -- does it performantly and nicely play with to-string APIs written as ostream<< impls?)
The formatting/order/style/whatever should follow some consistent convention; and since we share DNA with Flow-IPC should be consistent with Flow-IPC/ipc#85 work.
Not-unrelated to #71 and Flow-IPC/ipc#85 -- probably should be looked-at at the same time:
Flow-IPC (generally speaking, written after Flow, except for some parts that were written in parallel) is pretty good about this (possibly outside non-ipc::session::shm SHM-jemalloc-land, but that's orthogonal), but Flow might be less so, or less consistently so:
Have a to-string output for all object types. So far we've been using ostream<< for this. (Maybe also look into how fmt deals with this, as it might be relevant for #78 -- does it performantly and nicely play with to-string APIs written as ostream<< impls?)
The formatting/order/style/whatever should follow some consistent convention; and since we share DNA with Flow-IPC should be consistent with Flow-IPC/ipc#85 work.