Bay 12 Games Forum

I suggest doing it incrementally. First write read_/write_raw that communicates with dfhack, then change the actual manipulations one by one.

Also, I suggested once to write df-structures using real C++ headers. Dunno what came of that, but I think it would make this easier (or not?).

most classes already exist in DT (Dwarf, Squad, Fortress), but they use read-write data in memory by offset with OS-specific implementations.

i want use protobuf as transport protocol, current state of Basic.proto and BasicApi.proto can fill more then 70% оf requirements for DT. But yes, I must do some work for extend this proto (or develop custom plugin for DFHack) for fulfill all requirements.

After that DT don't need know about what version of DF/DFHack running, because we don't now know anything about memory layout, structures and c++ headers.

Current main work it's replace directly work with memory on rpc calls.

after you fix one place and can display something, you must fix another place
(I use DF osx64, DT are not working for me and he simple crash on a try work with memory)
but mostly it's routine work for replacing read-write data one by one

Yes, it's replacement because current model of working DT very depend on direct work with memory.

Main benefits:
1) DFHack add support new version of memory layout more faster then DT
2) DT don't require admin/root permissions for work with DF

It not possible keep current state and simple add new backend,
all classes use QPointer and memory vector for direct read-write in memory
for support both backend in the same time we need for first will do big refactoring and cut working with memory in separate module
but for me it's unnecessary work which will gone in trash soon.