Le Wed, 13 May 2009 11:52:57 -0300, Jeremy Banks firstname.lastname@example.org s'exprima ainsi:
To someone who's a novice to this, could someone explain to me why it has to be an existing keyword at all? Since not identifiers are valid in that context anyway, why couldn't it be a new keyword that can still be used as an identifier in valid contexts? For example (not that I advocate this choice of keyword at all):
def foo(bar reinitialize_default ): # <-- it's a keyword here reinitialize_default = "It's an identifier here!"
That would be a syntax error now and if it were defined as a keyword only in that context it wouldn't introduce backwards compatibility problems and wouldn't force us to reuse an existing keyword in a context that may be a bit of a stretch.
Is there a reason that this wouldn't be a viable approach?
My opinion on this is you're basically right. Even 'print' (for py<3.0) could be an identifier you could use in an assignment (or in any value expression), I guess, for parse patterns are different: print_statement : "print" expression assignment : name '=' expression So you can safely have "print" as name, or inside an expression. Even "print print" should work !
But traditionnally grammars are not built as a single & total definition of the whole language (like is often done using e.g. PEG, see http://en.wikipedia.org/wiki/Parsing_Expression_Grammar) but as a 2-layer definition: one for tokens (lexicon & morphology) and one for higher-level patterns (syntax & structure). The token layer is performed by a lexer that will not take the context into account to recognize tokens, so that it could not distinguish several, syntactically & semantically different, occurrences of "print" like above. As a consequence, in most languages, key word = reserved word
There may be other reasons I'm not aware of.
Denis ------ la vita e estrany