xref: /xsrc/external/mit/xedit/dist/lisp/TODO

$XFree86: xc/programs/xedit/lisp/TODO,v 1.8 2002/11/23 08:26:47 paulo Exp $

LAST UPDATED:	$Date: 2008/07/30 04:16:22 $

  Small todo list

o Change function/macro body to know if a &key or &optional argument was not
  provided, and initialize to the default value in the function, for
  interpreted and builtin functions it is better done before the function is
  called, but for bytecode it is better in the function.
o Following the previous idea, change function definitions in the format:
  (defun afun (a &aux b (c (some-code))) ...)
  to
  (defun afun (a) (let* ((b (c (some-code)))) ...))
  This can significatively reduce bytecode size, and also simplify function
  calls.
o Optimize text redisplay in Xaw, instead of allocating a copy of the buffer
  for the paint-list, should use the text in place, and only allocate small
  buffers when required, i.e. displaying control characters, etc.
o Add an interface to create new object types dinamically.
o Add a special string object to simplify and avoid too many copies of
  portions of the text buffers. This special string should be read-only
  and not gc-collected.
o Make the bytecode compiler smarter to detect some constructs like:
  (builtin-or-bytecode-function-call arg1 arg2 (return))
  this will not properly restore the internal stacks.

o When an Init function is present in every file, call LispAddBuiltin from
  that initialization function.
o Cleanup the code, make it optional to build some uncommon features (as well
  as the entire interpreter?). Implement more functions directly in lisp.
o Finish an "specification" for loadable modules. Write a FFI interface.
  Without the possibility of using already existing libraries, the interpreter
  won't be too much useful due to a poor library. It is very desirable to
  "auto-generate" directly from C header files the interface to the lisp
  code, and from that, dlload a shared library. In some cases, it is required
  to link statically with a new interpreter binary, make it easy.
o Implement a better string type. That should support characters larger than
  8 bits, and that should allow embeded nuls.
o Implement a richer set of math functions. This, if properly done can be
  made a loadable module.
o Optmize mathimp.c, comparing a double with a bignum should never cause an
  error. Implement mp?_initsetXXX functions?
o Finish missing features in read.c, and simplify it.
o (close) probably should not send a signal when closing a pipe.
o Implement "real" vectors, they exist to make access to field elements
  at constant time, implementing vectors as lists may be very slow.
o Use float and double as floating points formats.
o Implement support for vectors of "atomic" types. Vectors of floats would
  be useful for example if a OpenGL binding is done.
o Implement a multiple precision floating point format. Either a 128 bits
  (or configurable length?) IEEE 754 like number, or some format using
  ratios, a epsilon for rouding, rounding modes, exact/inexact flag, a good
  amount of guard digits, etc.
o Write more functions and optimization for bignums. Try to make the code
  as reusable as possible, either by other software or at least by the
  different number types.
o Instead of using mathimp.c for a large amount of functions, implement a
  "generic number" type and implement the calculations directly in the
  mp library.
o Add more missing Common Lisp features, like &allow-other-keys for function
  definitions, the missing structure features, CLOS etc.
o Fix the Postgresql module, make it compile everywhere.
o Add support for multi-threaded applications?
o Make it possible to know if a object has only one reference, this is
  required to make "inplace" changes of variables. Very useful for things
  like (incf) and (decf), and also when dealing with bignums.
o Maybe have a freelist for objects depending on the type. Bignums can be
  reused, and having some large ones expecting to be freed by the gc can
  consume a lot of memory.
o Internationalization. Support ',' in floats? Correctly treat characters
  for {up,down}-casing.
o Synch the Xaw text code for supporting tables, text alignment/paragraphs,
  etc, and add bindings to the interpreter. Add support for incremental
  changes of those properties, currently it is "read-only".
o Write some type of "hyperlinks", this is the only feature missing to even
  allow writting a web browser inside xedit.
o Write some "demos" (maybe a file manager or a simple mail reader) using the
  Xt/Xaw bindings (needs modules working in all systems).
o Remove all calls to the macros GCDisable and GCEnable. This is unsafe
  and should be removed.