Eclipse Community Forums: TMF (Xtext) » Data type rules - Can someone explain this concept to me?

Help

Home

Home » Modeling » TMF (Xtext) » Data type rules - Can someone explain this concept to me?

Show: Today's Messages :: Show Polls :: Message Navigator

Data type rules - Can someone explain this concept to me? [message #685770]

Sun, 19 June 2011 18:50

john doe

Messages: 74
Registered: June 2011

Member

I have read the tutorial on data types here (www.eclipse.org/Xtext/documentation/latest/xtext.html#datatypeRules), however I still don't really get why you would use a data type rule instead of a terminal rule.

If you look at the link above, why could you not just do the following instead?

terminal QUALIFIEDNAME :
  ID ('.' ID)*
;

Why can't we declare this as a lexer (terminal) rule instead of a parser rule? What exactly is a data type rule?

Thanks

Report message to a moderator

Re: Data type rules - Can someone explain this concept to me? [message #685771 is a reply to message #685770]

Sun, 19 June 2011 18:58

Christian Dietrich

Messages: 14665
Registered: July 2009

Senior Member

Hi,

terminal rules go into the lexer. leing is done in the first phase of the parsing process and is done context free => ID ('.' ID)* will be always parsed
as that and something like parent=ID "." child=ID won't be possible. datatype rules are like other parser rules context sensitive.

~Christian

Twitter : @chrdietrich
Blog : https://www.dietrich-it.de

Report message to a moderator

Re: Data type rules - Can someone explain this concept to me? [message #685774 is a reply to message #685771]

Sun, 19 June 2011 19:28

john doe

Messages: 74
Registered: June 2011

Member

Thanks for the reply. When you say that

parent=ID "." child=ID

won't be possible, why is that?

And why would I not just define a terminal rule instead? Such as
that outlined in my first post?

Thank you!

[Updated on: Sun, 19 June 2011 19:42]

Report message to a moderator

Re: Data type rules - Can someone explain this concept to me? [message #685775 is a reply to message #685774]

Sun, 19 June 2011 19:34

Christian Dietrich

Messages: 14665
Registered: July 2009

Senior Member

hi, if the lexer finds a x.y in the text it will be always parsed as a fqn and never as and ID followed by a . followe by a ID.
the lexers thus creates a token stream of the file and handles this to the parser.
so if the parser wants and ID and a . and another ID but always gets and FQN you ll get an unexpected token error.

So the basic problem is that there is first lexing (context free) and then parsing.

=> Terminal rules are for the lexer, Datatype rules for the parser
=> Terminal rules are context free, datatype rules are context sensitive.

~Christian

Twitter : @chrdietrich
Blog : https://www.dietrich-it.de

Report message to a moderator