I just upgraded from 2010 B2 to the RC and was greeted by several hundred syntax errors due to custom operator name rules changing. I've having problems understanding how to fix things.

As an example, consider "~@" which was a prefix operator that was working fine and is no longer accepted. Looking at the (what I think are the most recent) docs, it seems like this should still be fine:

_________________________________________________________

Creating New Operators

You can overload all the standard operators, but you can also create new operators out of sequences of certain characters. Allowed operator characters are !, %, &, *, +, -, ., /, <, =, >, ?, @, ^, |, and ~. The ~ character has the special meaning of making an operator unary, and is not part of the operator character sequence

Depending on the exact character sequence you use, your operator will have a certain precedence and associativity. Associativity can be either left to right or right to left and is used whenever operators of the same level of precedence appear in sequence without parentheses.

Other combinations of operator characters that are not listed here can be used as operators and have names that are made up by concatenating names for the individual characters from the following table. _________________________________________________________

So it sounds from the docs as if "~@" should be fine, as a unary "@" operator - but, that only makes since if "~" is part of the op calling sequence (which is how I used in before), because otherwise "@" is binary. (I can no longer call "~@" directly.)

I assume now that what the docs mean is: "The ~ character has the special meaning of making an operator unary if the operator is defined such that it can be either unary or binary, or if the operator is a sequence of only tildes." Also, "Depending on the exact character sequence you use" is pretty vague - specifically, I'm assuming that the first character (other than ".") defines the prec & assoc of a custom (one not currently defined) operator, but this is not specified. (Edit: later on we see, "Typically, the first character of the operator determines whether the operator is an infix or a prefix operator." - but what are the "untypical" examples?) We also see the sentence "The operator characters . and $ are not used to determine whether an operator is an infix operator or a prefix operator; instead, the next character in the operator is used to determine this" - with no explanation anywhere of what "$" is supposed to mean. (We are told that "." is used a op/prec. neutral leading character.)

Even the following statement seems confusing: F# supports custom operator overloading. This means that you can define your own operators

Operator "overloading" means re-defining an existing operator - defining custom operators is another thing entirely, and is the source of the naming confusion.

Now, "!@" seems to work fine. However, "?@" (as "~@") does not, even though "!" and "?" seem to be in the same class of prefix operator characters.
These work: ? !? !@ !~ ~~~ !!? !!!

These do not: ?! ?@ ~@ ~~! ???

None of the names that work do so with a leading ".", despite the docs saying that they are a valid and neutral leading character.

It is true that I have not had any coffee nor have I been to the gym today, but if there's a pattern here can someone point it out? Can this be explained another way in the docs? (Hell, maybe even some code that given a string, returned a DU of |InvalidOperator |Prefix of int |Binary of int, would be good documentation.)

Sorry if I'm missing a key piece to the Grand Unified Operator Name Theory...
thanks!

NB: Given F#'s "LOP" following, I'd love to see an operator system that is simple and complete - such as Prolog's "op(++, xfy, 500). op(not, xf, 800)." style definitions. (No, trust me, I don't want F# to be Prolog - I'm justing saying that it's simple and you can do alot w/it re:DSL's w/o resorting to parsing yourself, especially seeing the reader and the printer both grok the operators.)