Re: [office-formula] Syntax loose ends: inline error names,

["David A. Wheeler" <dwheeler@dwheeler.com>]

Andreas J Guelzow says:
> (For obvious reasons) I like the Gnumeric version:
> Sheet1.print_area   refers to the named expression in sheet scope of
> Sheet1 if there is such an expression defined otherwise it refers to the
> expression in workbook scope. 
> .print_area simply replaces Sheet1 with the local sheet.

I like the idea, worry about ambiguity.  What is this?:
  'file:///local'.print_area
Is it an external reference to a file named local, or the sheet named "file:///local"?
SheetNames can have arbitrary chars in them, and thus
sometimes get surrounded with '...'.  IRIs also get surrounded by '...'.
If we simply said "either an IRI or a SheetName" can prefix a namespace
identifier, we'd create a nasty ambiguity.

And if you allow named expressions to be used as functions and attached
to sheets - a future plausible extension that has been discussed - it gets worse.
What is sheet1.myfunc("hello") -- is that a call to the function on sheet1,
or to the function named "sheet1.myfunc" (like Error.type)?

Here's what we have now:
  NamedExpression ::= Source? Identifier
  Source ::= "'" IRI "'" "#"
  FunctionName ::= Identifier
  Identifier ::= LetterXML (LetterXML | DigitXML |
               '_' | '.' | CombiningCharXML)*
  SheetName ::= /* empty */ | '$' ? ( [^\. #$']+ | "'" ([^'] | "''")+ "'" ) | Error

Getting exactly the right solution is non-trivial, esp. if we
want to have syntax available for nested tables.

Syntactically, we could simply say that if an IRI begin with '#',
it's local, and followed by the SheetName in IRI format.
That would be an easy change, but is it too ugly to use?
It's also odd; if I have nested tables,
I would expect an IRI reference to start from the top (root) of the document,
yet I'd want a sheetname to start from the bottom (current location)
and work up to find a matching sheetname. Might get different answers!

Perhaps we need to introduce SheetName-specific references with another
character, so disambiguation is trivial.

Suggestions on the new BNF?

--- David A. Wheeler