Re: foreign interface



Michael,
>         I really don't see what you're adding either in terms of
> performance or complexity reduction.

Complexity reduction for language bindings is my primary objective. 
Using the current interfaces, my foreign data structures have to be 
pressed into C data-structures. This is done doing a lot of heap 
allocation, even for primitives. Data trees have to be traversed twice, 
once for conversion and second for marshalling, the same on response.

On marshalling the data tree is traversed to marshall its leafs. Doing 
demarshalling leafs and nodes have to be created. Generic tree 
traversing and tree construction can be encapsulated using interfaces, 
design patterns of wrapper and factory. Additional heap allocation would 
  be obsolet, primitve conversion could be done in space.

Performance could be an add on in terms of

T(interface interaction)
--------------------------- < 1
T(previous heap allocation)

but this is not expected.

orbit-foreign is thought for scripting languages, NOT for C, C++, 
Modula, etc.

Regards, Frank
-------------------------------------------------------------
  Frank Rehberger <frehberg@cs.tu-berlin.de>
-------------------------------------------------------------
  "Global order can arise from local interactions."
  [A.Turing, 1952]
-------------------------------------------------------------
  "when all nodes have reached that [stable] state, the whole
  graph is as dead as a doornail and the diffusing computation
  is defined to have terminated" [E.W.Dijkstra, 1980]
-------------------------------------------------------------





[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]