[antlr-interest] Re: SQL grammar tree construction problem

[Joshua Davis]

Just a quick follow up.  The Parser did not build an AST because the 
rule that had fired did not specify which token should be the root of 
the sub-tree.  For example:

fromClause : FROM identifier (AS identifier)?

The AST builder seems to require that one of the tokens be designated 
as the root.  I changed it to:

fromClause : FROM^ identifier (AS identifier)?

and it worked.

--- In antlr-interest at yahoogroups.com, "Joshua Davis" 
<joshua.davis at k...> wrote:
> My example is even simpler.  It's *supposed* to generate a node for 
> every rule if the buildAST is set to true isn't it?  I was under 
the 
> impression that explicitly creating nodes is not necessary if 
> buildAST is enabled.
> 
> --- In antlr-interest at yahoogroups.com, Gustav Boström 
<i93gusbo at h...> 
> wrote:
> > Hi!
> > 
> > Thanks for the help!
> > 
> > I think I'm trying the same thing as you, but also without luck.
> > I try to do:
> > sql_stmt : 
> > 	  sql_data_stmt  { #sql_stmt = #([SQL_STMT,"sql_stmt"], 
> > #sql_stmt); }
> > 
> > , but I never find this node in the tree. The code seems to be 
> > generated, but has no effect. Could it be that no root node is 
ever 
> > created?
> > 
> > I also tried this:
> > sql_script : 
> > 	 (sql_stmt)? { #sql_script = #
> > ([Token.MIN_USER_TYPE,"sql_script"], #sql_script); } ( SEMICOLON 
> > (sql_stmt)? )*
> > 
> > "sql_script" is the first rule and now a node is in fact created, 
> but 
> > it has no children. I thought that all the subsequent nodes would 
> be 
> > added as children automatically.
> > 
> > Qoute from doc:
> > "Setting the return tree is very useful in combination with 
normal 
> > tree construction because you can have ANTLR do all the work of 
> > building a tree and then add an imaginary root node such as: 
> >   
> > decl : ( TYPE ID )+
> >        { #decl = #([DECL,"decl"], #decl); }
> >      ;
> > ANTLR allows you to assign to #rule anywhere within an 
alternative 
> of 
> > the rule. ANTLR ensures that references of and assignments to 
#rule 
> > within an action force the parser's internal AST construction 
> > variables into a stable state. After you assign to #rule, the 
state 
> > of the parser's automatic AST construction variables will be set 
as 
> > if ANTLR had generated the tree rooted at #rule. For example, any 
> > children nodes added after the action will be added to the 
children 
> > of #rule." 
> >  
> > This is exactly what I want, but I must be doing something wrong.
> > 
> > 
> > 
> > /Gustav
> > 
> > 
> > 
> > --- In antlr-interest at yahoogroups.com, "Joshua Davis" 
> > <joshua.davis at k...> wrote:
> > > ANTLR newbie question:
> > > 
> > > I've been having a similar problem I think.  I made a very 
simple 
> > > grammar based on L. V.'s grammar.   It seems to be always 
> producing 
> > > only one root node, with no children.  What steps should I take 
> to 
> > > debug this?  I have tried adding actions to the rules, and it 
> looks 
> > > like the rules are being triggered correctly, but no AST is 
> > produced.
> > > 
> > > --- In antlr-interest at yahoogroups.com, "lgcraymer" <lgc at m...> 
> wrote:
> > > > Gustav--
> > > > 
> > > > You cannot make a rule a root--that's what ANTLR is 
complaining 
> > > about. 
> > > >  Rules may represent subtrees, so only nodes with token ids 
> (TEXT 
> > > or 
> > > > STRING, for example) can be followed with ^.
> > > > 
> > > > --Loring
> > > > 
> > > > 
> > > > --- In antlr-interest at yahoogroups.com, Gustav Boström 
> > > <i93gusbo at h...> 
> > > > wrote:
> > > > > Hello!
> > > > > 
> > > > > I'm trying to use Lubos Vnuk's (Thanks for doing all the 
> work!) 
> > > SQL-
> > > > > grammar in order to do some 
> > > > > processing on SQL-statments. I would for example like to 
find 
> > out 
> > > > > which columns are affected by a DML-statement and which 
> > > parameters 
> > > > > correspond to which columns.
> > > > > 
> > > > > I'm doing this is in Java , so I've changed the grammar 
> > slightly.
> > > > > I've removed the C++ actions or replaced them with Java 
> > versions.
> > > > > 
> > > > > My thought was to add AST-construction and then navigate 
the 
> > tree 
> > > to 
> > > > > find out the stuff I need. This is where I run into 
problems.
> > > > > I tried to annotate the rules with ^ to construct my tree, 
> but 
> > it 
> > > > > keeps giving me "unexpected token:^" when I try to dothis 
for 
> > > > example:
> > > > > sql_stmt : 
> > > > > 	  sql_data_stmt^ 
> > > > > ;
> > > > > 
> > > > > I've set the options to buildAST=true;
> > > > > I'm using version 2.7.2.
> > > > > 
> > > > > Any ideas on what can be wrong?
> > > > > 
> > > > > Cheers,
> > > > > 
> > > > > Gustav Boström


 

Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/