Monster: evaluate math expressions in strings w/o external programs

This script gradually grew from twenty lines to over 250 AHK code lines. It might not be the best approach, but each time only a few new lines were added, so it was never feasible to start all over for a more general expression parser.

The main point is not providing a standalone calculator, because there are hundreds of free ones available, but to allow evaluating a math expression typed in an editor or word processor. The compiled script can run in a PC, where AHK is not installed. It is easy to add new functions of one variable. Follow the examples of SGN (the sign of its argument), Fib (the n-th Fibonacci number) and FAC (the factorial). Two argument functions are implemented as operators: "9 gcd 6" is the same as it would be in functional form gcd(9,6), which is 3. "4choose2" = 6, the binomial coefficient. Min and Max can be chained: "-1 min 2 min -5" gives -5; "3 max 5 max 2" = max(3,5,2) = 5. Further operators can be easily added to the script.

User variables can be defined and used. They are kept in memory until the script is reloaded, with values assigned before they are used in expressions, like "a:=1; b:=2; a+b". In later evaluated expressions "a" and "b" still have their last assigned values.

User functions (even recursive ones) can also be defined in expressions:
f(x) := x < 2 ? 1 : x*f(x-1)
Here the formal parameter x is not defined as a variable, it is a placeholder for the value given to the function at a call. The function definition also remains in memory until the script is terminated.

As usage examples, two hotkeys obtain the math text from the current document. Ctrl+Win+- replaces the expression with the result, Ctrl+Win+= appends an "=" and the result. If a (multi-line) expression is selected, it gets evaluated. If there is nothing selected, the script searches for the last back-quote in the current line, and uses the text after it, until the insertion point (caret). For example, after pressing Win+Ctrl+= the text

"The area is `125*32"
becomes
"The area is 125*32 = 4000"

If instead you press Win+Ctrl+- at the the original line above, it becomes
"The area is 4000"

There are many features supported in the expressions:

- They can contain HEX (0x1ff), Decimal (123, 1.23e2) and Binary numbers ('1001, first bit = sign),
- arbitrary number of nested parentheses (..)
- variables (a, b), constants (e, pi, inch, foot, mile, ounce, pint, gallon, oz, lb)
- user defined functions
- ternary- (_? _ : _) and logical operators !; ||; &&
- relational operators =,<>; <,>,<=,>=
- special operators: GCD, MIN, MAX, Choose;
- bitwise operators ~; |; ^; &; <<, >>
- arithmetic operators +, - (or #); *, /, \ (or % = mod); ** (or @ = power);
(They are listed in the order of their precedence, from low to high, except the unary operators !,-, + and ~ are of highest precedence. The operators separated by commas are of equal precedence.)
- Functions Abs,Ceil,Exp,Floor,Log,Ln,Round,Sqrt,Sin,Cos,Tan,ASin,ACos,ATan,SGN,Fib,fac

The output format can also be specified.
- If the expression does not contain the format specifier "$", 6 decimal digits are shown in the general form (%0.6g in C), and integers are in decimal
- With ${k}: k (optional integer) decimal digits are shown after the decimal point, in case of floating point result
- With ${k}e or ${k}g: k decimal digits are shown of floating point results, in exponential or general scientific form, respectively
- With $x or $h: Rounded results are shown Hex
- With $b{W}: Rounded results are shown in binary form (LS W-bits; W="": first bit is sign: "1000" = -8, "0111" = 7)

There are no precision limitations in version 1.1 or later. All intermediate results are computed and stored in full 64-bit accuracy.

The ternary operator "condition ? exp1 : exp2" is implemented as two low precedence operators: "a ? b" (which returns "b" if "a" is true, that is nonempty and nonzero. It evaluates to the empty string if "a" is false, that is empty or 0. The other operator "b : c" returns "c" if "b" is the empty string, otherwise it returns "b". (It can also be used to set default values to uninitialized variables.) This saves time, because only the one of "b" or "c" is computed, which is returned. It also helps avoiding arithmetic errors, like in "x=0 ? 0 : 1/x". Here 1/x is not computed if x=0.

There is, however, a little difference between this implementation and the true ternary operator "a ? b : c". In Monster if "a" is true, "b" is returned by the "?" operator. If it happens to be empty (an error in an arithmetic expression), the following ":" operator does not know if "a" was false or this error occurred, and returns "c", instead of the empty "b". It was easy to change this behavior (e.g. "?" could return a special, non-numeric character when "a" is false and ":" checks for this character), but then we'd loose the "default value" function of the ":" operator "x:0".

Function names are internally enclosed by ' characters, operators are enclosed by  and  characters, to prevent problems when their names happen to be the prefix or postfix of another name (like tan/atan/at).

Please report bugs! Regular expressions are heavily used, and they are notoriously difficult to read (and to get them right). 