Path: news.daimi.aau.dk!olm
From: olm@daimi.aau.dk (Ole Lehrmann Madsen)
Newsgroups: comp.lang.beta
Subject: Re: Extensions to the Beta Grammar
Date: 12 Feb 1997 21:23:03 GMT
Organization: DAIMI, Computer Science Dept. at Aarhus University
Lines: 44
Message-ID: <5dtcbn$rvc$1@gjallar.daimi.aau.dk>
References: <3301C7BF.1C3A@daimi.aau.dk>
NNTP-Posting-Host: clematis.daimi.aau.dk
X-Newsreader: NN version 6.5.1 (NOV)
Xref: news.daimi.aau.dk comp.lang.beta:10963

In <3301C7BF.1C3A@daimi.aau.dk> J|rg Olaf Schneider <josch@daimi.aau.dk> writes:

>While experimenting with extensions to the Beta grammar, I observed that
>the bobsit tool concludes, that the Beta grammar is not LALR(1). This
>implies that I can't generate a correct parser for an extended Beta
>grammar. Can anybody tell me, why the Beta grammar is not LALR(1), as
>required by the metaprogramming system. The grammar used is in
>'~beta/grammars/beta/v2.5'. How do I fix this?

The reported LALR(1) conflict is due to conflicts between so-called
error productions. Error productons are used by the parser to recover
after syntax errors in the input string.

The reported conflict is harmless and a parser is in fact constructed.
(I think that the message from bobsit actually says that a parser is constructed?)

BOBS always constructs a parser, even when there are LALR-conflicts.
When a conflict is met during parsing, a choice is made between one
of the conflicts. If the conflict is due to a simple ambiguity this will
work. If the conflict is not a simple ambiguity but the choice depends on
input symbols further ahead in input the choice may be wrong and a parse
error is reported. It could be fixed if the parser did some backtracking, but
this is not the case.

In general, of course, one should fix LALR-conflicts, since the parser may
otherwise be incorrect. However for simple ambiguities, the paser may be OK.
To decide this may be hard unless you are familiar with LALR-grammars.

As said, the reported conflicts for the BETA grammar are harmless.

I hope this helps. It is difficult explaining LALR-parse theory
in a message like this. I will be happy to give a more detailed 
explanation if you come by my office:-)

---olm

>-- Jorg

>-------------------------------------------------------------------
>/ Jorg Schneider, Computer Science Department, Aarhus University  /
>/ Ny Munkegade 116, Bld. 540, DK-8000 Aarhus C, DENMARK           /
>/ Email: josch@daimi.aau.dk                                       /
>/ WWW: http://www.daimi.aau.dk/~josch	                          /
>-------------------------------------------------------------------