=============================================================================

	icalc - a complex-number expression parser

	by Martin W Scott

=============================================================================

	Revision History,
	and Things To Do.

=============================================================================

	To Add
	------

PROBLEM	- function, f say, defined initially, then some other fn g
	  defined using f. f is redefined, incorrectly at first, thereby
	  deleting it from sym-table, until defined correctly. But
	  pointers in g now point to nothing...
	  This indicates need for reference count within symbols...

	- break, continue constructs.

	- print function which takes strings and format values.

	- error construct. Very similar to print...


	Version 2.0
	-----------

	- new and revised scripts - see documents for details.

	- writevars "file" writes all variables and arrays to specified
	  file in such a way that it can subsequently be read by a 'read'
	  command or as an argument to icalc.

	- exec "str" command added, passes string to system()
	  ie. runs command specified in str. eg. exec 'run dme'
	  Also, cd "str" and pwd commands added.

	- local vars available for functions via local statement.

	- 1d-arrays added - see docs for details.

	- clear <SYM> where sym is function or variable (or array)
	  frees all memory, deletes entry in table. Use with care, doesn't
	  check if SYM is referred to elsewhere.

	- startup-file optionally specified in env-var ICALCINIT.

	- read <name> command, so you can read files from other files
	  (like #include) and interactively. Makes writing and debugging
	  of compilcated functions much easier.

	- Functions can take expressions as parameters, ie. parameters need
	  not be evaluated on calling, but only on reference; this can be
	  a powerful tool in function definition - see docs.

	- More freedom about putting new-lines in expressions to aid clarity.
	  See docs for details. Still not free-form though, because of the
	  interactive nature of icalc.

	- if-then and if-then-else constructs added.

	- while-do and repeat-until constructs added - see docs for details.
	  To facilitate infinite loops (which would probably print() values),
	  the constants TRUE and FALSE have been added. Also added INFINITY
	  constant, which represents the largest possible value that can be
	  used.

	- [internal: changed over to Berkeley Yacc from Bison, because
	   the latter didn't seem to understand %nonassoc operator type.]

	- [Bugette: a = 3 && b = 4 is valid - parses as a = (3 && (b = 4))
	   instead (I would have thought) as a = (3 && 4) = b (ie. nonesense).
	   Something wrong with precedence of right-to-left ops in Bison/Yacc?
	   Or a grammar that just can't be dealt with by Bison/Yacc?]

	- [Internal: binary ops have node set to call-function - quicker than
	   case lookup in eval_tree().]

	- Exponentiation operator ^ now sensitive to operands, ie. doesn't
	  just use complex-number formula, but chooses method depending on
	  operands. Thus, realno^realno = realno + 0i. Quicker in most
	  cases now.

	- BUG FIX: files are now closed upon being read, rather than at
	  program termination.

	- BUG FIX: can now assign to parameters in function definition (for
	  what it's worth...).

	- BUG FIX: all arguments to functions are now evaluated from left
	  to right; previously, this was the case for builtin functions,
	  but not user-defined functions.

	- Recursive functions may now be written, but program stack grows
	  quickly - I'm working on reducing it.

	- Addition of ternary operator ?: with usage as in C (see docs).

	- Addition of +=, *= etc. operators.

	- Addition of relational operators > , >=, <, <=, ==, != and logical
	  operators &&, ||, ! (negation). && and || use short-circuit
	  evaluation, ala C.

	- Abort trap now set, so you can ^C out of long calculations :-)
	  Doesn't slow things down too much.


	Version 1.1a
	------------

	- BUG FIX: The inverse trigonometric functions asin, acos, atan now
	  return values in the conventionally regarded PVR's. They should
	  also work all the time now (unlike in the last release). 

	- BUG FIX: Fixed major bug whereby sqrt(z) didn't return values in
	  the correct quadrant for PVR of (-PI,PI]. Not only did this affect
	  the sqrt function, but also all functions defined using sqrt
	  (internally and in scripts), including inverse trig. functions.

	- Addition of max and min functions, which take a variable number
	  of arguments and return respectively the value of the  maximum
	  or minimum REAL part encountered.

	- Addition of a time() builtin. This allows timings of operations
	  to be made. time(t) = current time - t.

	- Addition of sgn() builtin, which returns the sign (-1,0,1) of the
	  REAL part of its argument.

	- Modified definitions of int, floor, ceil to only act on REAL part of
	  their arguments, so to get floor of imag. part, say floor(Im(z)).

	- Revised icalc.init file, with more definitions.

	- BUG FIX: Corrected stupid mistake in definition of polar(r,theta)
	  function in icalc.init.



	Version 1.1
	-----------

	- Example script files modified. Now provided are:
		icalc.init, which has many useful functions (startup-file);
		trig.icalc, which contains less-used trig functions;
		stat.icalc, one-variable statistical analysis.
	  You are strongly recommended to look at these for tips on how
	  best to use icalc.

	- On startup, icalc will now read the file 's:icalc.init' if it
	  exists.

	- Special function multi() added, which evaluates all its arguments,
	  returning the last one as its value. See docs and example scripts
	  for applications. Also added is print() builtin, for use
	  with multi().

	- Special functions Sum(), Prod(), every() and vevery() now added.
	  The repeat construct has therefore been removed. (It was only
	  in there until I got around to adding this type of facility).

	- Expressions and function definitions may now be continued on
	  the next line by use of a backslash '\'.

	- Identifiers may begin with and contain underscores (and, as a
	  side-effect of the implementation, be composed entirely of
	  underscores).

	- Multi-parameter functions now added.

	- You can now declare functions that take no parameters, eg.
		func total() = sum1 + sum2

	- Listing of user functions improved: now shows parameter list;
	  However, I don't think it's worth writing out whole definition.

	- Method of setting precision has changed: now use function
	  prec(digits), which returns ans (last computed result). Initially
	  set to 8.

	- Added function int(z), which returns the real and imaginary parts
	  of z rounded to nearest integers; 0.5 rounds up.

	- Added functions ceil(z) and floor(z), which operate separately
	  on the real and imaginary parts of their argument, in a similar
	  manner to int(z).

	- Now has a prompt to avoid confusion.

	- [Real-valued functions are no longer computed internally as such.]

	- Removed debugging code accidentally left in with version 1.0.



	Version 1.0
	-----------

	- Initial release.
