Playing with Marble for my own fun project, and looking at the implementation of get, I was wondering if you'd consider adding an API that lets uses provide their own callback for buffer allocation?
In my case I would like to be allocating into my own bespoke buffer pool, which is based vaguely/roughly on the techniques described in the Umbra paper (https://db.in.tum.de/~freitag/papers/p29-neumann-cidr20.pdf). And I would prefer not to have an extra copy/move step into those buffers after doing my work with Marble.
I'm imagining something like being able to pass something like an FnMut(usize)->Pin<&mut [T]> as an alloc_fn or similar? Marble still might need to read and compose the header_buf separately, but the work done with uninit_boxed_slice could be done instead by the user's provided buffer allocation routine.
Not sure if this makes sense for your overall design or not, but something to consider.
Playing with Marble for my own fun project, and looking at the implementation of
get, I was wondering if you'd consider adding an API that lets uses provide their own callback for buffer allocation?In my case I would like to be allocating into my own bespoke buffer pool, which is based vaguely/roughly on the techniques described in the Umbra paper (https://db.in.tum.de/~freitag/papers/p29-neumann-cidr20.pdf). And I would prefer not to have an extra copy/move step into those buffers after doing my work with Marble.
I'm imagining something like being able to pass something like an
FnMut(usize)->Pin<&mut [T]>as analloc_fnor similar? Marble still might need to read and compose theheader_bufseparately, but the work done withuninit_boxed_slicecould be done instead by the user's provided buffer allocation routine.Not sure if this makes sense for your overall design or not, but something to consider.