if I want to save a data structure,
I have two options now:
1. WRITE to a file
both have drawbacks:
1. WRITE uses ascii text which may be slow to
read and write, and it may fail for large
highly circular data because of STACK.
2. MEM cannot be loaded at will: if I want to run
something on 3 datasets, I must start 3 processes.
I suggest that we add object save/retrieve facility
with a berklely-db backend -
probably in the BDB module.
HASH Berkely-DB tables appear to be the right choice.
pointer objects are to be written as records keyed on
as_oint(obj) and with value being the memory to which
the object pointer refers, with other pointer objects
immediate slot values are to be written as is.
1. thus must be platform-independent
(including 64-bit vs 32-bit).
2. will this require me to write a separate routine
for every single CLISP built-in type?
Log in to post a comment.