Eclipse Community Forums: TMF (Xtext) » Limitation on recursive objects?

Home » Modeling » TMF (Xtext) » Limitation on recursive objects?(Limitation on recursive objects?)

Limitation on recursive objects? [message #1855498]

Tue, 18 October 2022 08:07

Eclipse User

Hi,

we have some kind of sample ifstatement in our project

IfStatement:
'if' condition=ValueExpression then=BlockAnnotatedStatement => ('else' else=(BlockAnnotatedStatement|IfAnnotatedStatement))?;

now the problem as if we have a lot of else if statements it seems that the build keeps hanging.
Is there any limitation on the amount of depth we can go into a xtext object?

Kind regards,

Re: Limitation on recursive objects? [message #1855504 is a reply to message #1855498]

Tue, 18 October 2022 14:35

Eclipse User

I always recommend to actually measure (profile) where the time in spent

Re: Limitation on recursive objects? [message #1855510 is a reply to message #1855504]

Wed, 19 October 2022 02:36

Eclipse User

In attachment i included partial of the jprofiler

Attachment: ifIssueJprofiler.jps
(Size: 6.16MB, Downloaded 102 times)

Re: Limitation on recursive objects? [message #1855524 is a reply to message #1855510]

Wed, 19 October 2022 10:17

Eclipse User

unfortunately i dont have a jprofiler lic.
looking with the eval version
i dont see any profiling from within eclipse there.
so i assume the answer is yes. if you nest too deep, then calling the parser will be slow.
the question is: is the problem in eclipse the parser or something else

i see there is 35% spent in the lexer.

Re: Limitation on recursive objects? [message #1855525 is a reply to message #1855524]

Wed, 19 October 2022 10:42

Eclipse User

btw did you try to use a First Token Set Predicate (->) instead of the Syntactic Predicates (=>)

IfStatement:
'if' condition=ValueExpression then=BlockAnnotatedStatement (->'else' else=(BlockAnnotatedStatement|IfStatement))?;

Re: Limitation on recursive objects? [message #1855542 is a reply to message #1855525]

Thu, 20 October 2022 01:46

Eclipse User

Hi

The underlying technology of Xtext is LL and so right recursion is good and left recursion bad. Through its * operator Xtext can support a simple effective left recursion. But I suspect that the underlying conversions/back-tracking are troublesome and quadratic.

For OCL, I found some ridiculously nested expressions in stylized auto-generated code took ages and so reformulated the grammar to parse a 'wrong' Concrete Syntax that Xtext could create efficiently. The 'wrong' CS is corrected during the CS to Abstract Syntax conversion.

Regards

Ed Willink

Previous Topic:	Create and configure instances
Next Topic:	whether Xtext/Xtend have some ways to develop DSL debugger

Goto Forum:

-=] Back to Top [=-

Current Time: Sun Apr 20 17:51:03 EDT 2025

.:: Contact :: Home ::.

Breadcrumbs

Sign up to our Newsletter