Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!purdue!decwrl!decvax!ima!compilers-sender
From: kjell@saturn.ucsc.edu (Kjell Post)
Newsgroups: comp.compilers
Subject: Error report in multipass compilers
Message-ID: <3547@ima.ima.isc.com>
Date: 23 Mar 89 22:34:24 GMT
Sender: compilers-sender@ima.ima.isc.com
Reply-To: Kjell Post <kjell@saturn.ucsc.edu>
Lines: 39
Approved: compilers@ima.UUCP

I am currently writing a multipass compiler where the parser builds
an Abstract Syntax Tree which is later traversed in several passes.
I would like to keep track of the linenumbers so I can refer to a
specific line when an error is detected during the tree traversal.

There are two approaches I can think of:

	1. Store the linenumbers in the leaf nodes only.
	   When an error occur, go first right then all the
	   way down the left spine to find a leaf.
	   This doesn't work in all cases; for instance 

           stmt ::= RETURN ';' { ... }
  
           doesn't have a leaf node and no line number can be
           associated with that construct.

        2. Store the linenumbers in all nodes.
           The problem here is how to synthesize a linenumber from
           two subtrees in an non-ad-hoc way.  Example

	   expr ::= expr '+' expr { ... }

	   The linenumber stored in $$ should probably be the linenumber 
	   where the '+' appeared.  But some productions have right-hand
	   sides without any terminals and so $$'s linenumber must be
	   synthesized from one of the nonterminals.  Empty subtrees must
	   of course be avoided so this method requires careful thinking.

Is there some clean, obvious way that I  have missed?

--Kjell (kjell@saturn.ucsc.edu)
[At parse time, I'd set the line number in a node from the current line number
kept by the lexer.  Crude, but as likely as anything else to be right.  -John]
--
Send compilers articles to ima!compilers or, in a pinch, to Levine@YALE.EDU
Plausible paths are { decvax | harvard | yale | bbn}!ima
Please send responses to the originator of the message -- I cannot forward
mail accidentally sent back to compilers.  Meta-mail to ima!compilers-request