Re: Lexical Conundrum

[dewar@merv.cs.nyu.edu (Robert Dewar)]

JPR says

<<This a known (and presumably only) small difficulty in parsing Ada. And yes,
compilers may take advantage that there is no one letter attribute to make
it easier... although it's not really hard to solve. For example, have a
look at the very simplistic parser used in "normalize" (a utility available
from Adalog's Web site at http://perso.wanadoo.fr/adalog).
>>

It is a difficulty in lexical analysis I would say, rather than parsing.
If you absorb the lexical rules into the grammar accepted by the parser,
there are no ambiguities, and no need for look ahead.

The only "difficulty" arises in doing separate lexical analysis, and as
I showed from the GNAT code, this difficulty is so trivial to resolve
that I see no reason to even describe it as a difficulty, hence my
quotes.

As for one character attributes, not so fast, when you have wide characters,
the test for whether the thing following a quote is or is not a valid
character literal is not that simple, and indeed that test would probably
be more work than the simple test done in the GNAT lexical analyzer.

There are no one character attributes in GNAT currently, but if we ever
found a case where a one character name was the right choice, we could
add this without any difficulties or special processing being required.