Re: LL(1) vs LALR(1) parsers

maatwerk@euronet.nl (M.M._van_der_Laan) writes:
>I want to add two ideas to this discussion:


>About quality of error recovery:
>Most errors are NOT detected in the parser, but afterwards. Errors
>detected by the parser are of the kind 'missing something" and in
>those cases there is no need for clever recovery - the program is
>simply wrong, so just skip the statement. Therefore, in my
>opinion error recovery is no argument for selection of the
>right type of grammar.


>For those in favor of handwritten parsers: I also like them
>because they are so easy to understand and fast........


About syntactical error-recovery in the various parsers (hand-written
or automatically generated). I remember that, some 10 years ago, I
saw an automatically generated LL(1) parser for Pascal(generated into
Pascal, easy to read, fast etc.), that has some capabilities that many
(very clever) hand-written parsers don't have.


Consider the following experiment:
Take any correct Pascal program and replace every semicolon (';') by a
space. Then, see what your favourite compiler does with this program.
A good parser will generate the same amount of `` ';' inserted'' error
messages, as the program contained non-redundant semicolons.


Watch and shiver!


I have seen seen compilers, that skip all declarations because the
semicolons are missing, and produce a lot of exotic semantic error
messages.
--

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