syntax role modifiers

[jurgenvinju]

This PR implements syntax role modifiers.

SyntaxModifiers.pdf

[codecov]

Codecov Report

Attention: Patch coverage is 28.35821% with 432 lines in your changes are missing coverage. Please review.

Project coverage is 49%. Comparing base (ab243ce) to head (d6ed722).

Files	Patch %	Lines
src/org/rascalmpl/types/ModifySyntaxRole.java	23%	208 Missing and 27 partials ⚠️
src/org/rascalmpl/ast/SyntaxRoleModifier.java	20%	90 Missing and 1 partial ⚠️
src/org/rascalmpl/library/Type.java	35%	24 Missing ⚠️
src/org/rascalmpl/ast/Type.java	34%	20 Missing and 1 partial ⚠️
src/org/rascalmpl/types/RascalTypeFactory.java	43%	12 Missing and 1 partial ⚠️
...org/rascalmpl/values/parsetrees/SymbolFactory.java	35%	10 Missing and 1 partial ⚠️
src/org/rascalmpl/ast/ShellCommand.java	0%	7 Missing ⚠️
src/org/rascalmpl/ast/NullASTVisitor.java	0%	6 Missing ⚠️
src/org/rascalmpl/types/RascalType.java	0%	6 Missing ⚠️
src/org/rascalmpl/types/TypeReifier.java	0%	4 Missing ⚠️
... and 9 more

Additional details and impacted files

@@           Coverage Diff            @@
##              main   #1835    +/-   ##
========================================
- Coverage       49%     49%    -1%     
- Complexity    6226    6266    +40     
========================================
  Files          662     665     +3     
  Lines        59419   60050   +631     
  Branches      8613    8726   +113     
========================================
+ Hits         29236   29474   +238     
- Misses       27996   28357   +361     
- Partials      2187    2219    +32     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[DavyLandman]

I'll start reviewing this soon. or are you still writing it @jurgenvinju ?

[DavyLandman]

Impressive, if you ignore the generated code, it's actually quite compact.

[DavyLandman]

is it also legal to write parse(#syntax[A], "a") ? and how does this relate to start[A]

[jurgenvinju]

Excellent question!

syntax[X] is equivalent to X in any context. So #X produces the same as #syntax[X] as long as X is not ambiguous.

If X is ambiguous then it could happen that #X produces an ADT definition instead of a syntax definition.

For X you can substitute other non-terminal forms; like syntax[start[X]] and syntax[A*] but that is implemented on a different branch. In the current branch you can type start[syntax[X]] to get start[X] while syntax[start[X]] is undefined. Also syntax[A*] is undefined here. With -ea asserts enabled, during the tests such an application will fail. In a deployed interpreter, without -ea, the unsupported combinations will be idempotent. In other words, syntax[A*] will simply reduce to A*, and it will behave as it as always behaved.

On the other branch, after we have this one working, and merged into it, we can use syntax[A*] to introduce layout symbols in all the right places, and also to pattern match on lexical lists, mutually exclusively with context-free lists: i.e. int size(syntax[&T*] l) and int size(lexical[&T*] l) would be two different overloads to handle the different way of counting elements in either kind of list.

[DavyLandman]

ah interresting. I would have expected syntax[start[A]] to be valid indeed. good that it's incoming as well.

[jurgenvinju]

I'll start reviewing this soon. or are you still writing it @jurgenvinju ?

The util::Explode code is actually more of a integration test; the rest should be almost done. Because of bootstrapping issues, where the standard library needs to be type-checking correctly, before we can release the new code, I think I have to park that use of the new feature until after we have extended the type-checker.

IOW: I'll merge this without util::Explode for bootstrapping purposes. The next step is extending and releasing rascal-core with this feature if we think we are done here.