Limitations of Operator Precedence Parsing?

In several places I have read that operator precedence parsing is less
powerful than more sophisticated methods, which seems pretty obvious,
but I am puzzled by the remark -- I think it was in the dragon book --
that operator precendence "grammars" are not tightly coupled with the
languages they describe. I gather that this means that they will
either refuse what ought to be legal constructs or that they will
accept what ought not to be legal constructs. Am I understanding this
correctly?


Unfortunately, most of the books on compiler design that I've been
able to lay my cash-starved hands on simply dismiss operator
precedence parsing out of hand without bothering to give concrete
examples of its limitations. This is a pity, since knowing why an
algorithm is inadequate can be quite instructive in knowing why others
are adequate.


I would really appreciate it if one of the resident experts here could
provide an example or two.


-e.


--Eric
[They accept too much. There's no way in an OP parser to say that two
operators can't be composed with each other. -John]

Return to the comp.compilers page.
Search the comp.compilers archives again.