I believe that's the way to go. I was thinking about it recently (we use a groovy dsl for a work project and providing a decent editor would be nicer for the customer, does have very low prio though). I found https://github.com/palantir/language-servers which is using the groovy compiler and exposing it as Language Server, but is abandoned. One of the many great things about using the original compiler would be that you could also use the extensions. However, especially for DSLs, the autocomplete would need more information than currently provided. My initial thoughts are that an ASTNode should be able to contribute to autocomplete and that we need a context sensitive factory for creating ASTNodes. Have not implemented anything yet.

Am So., 17. Feb. 2019 um 21:16 Uhr schrieb Milles, Eric (TR Tech, Content & Ops) <eric.milles@thomsonreuters.com>:

There is also the possibility of a consolidated effort on a Language Server Protocol implementation for Groovy that would replace the current eclipse tooling upon reaching maturity.  This would extend to other platforms like Visual Studio Code as well.



From: Milles, Eric (TR Tech, Content & Ops) <eric.milles@thomsonreuters.com>
Sent: Sunday, February 17, 2019 2:06 PM
To: dev@groovy.apache.org
Subject: Re: [VOTE] Polish the generics type syntax for closure
 

Agreed.  I'd be open to a side project to migrate eclipse patches back into core.  Many of them should be able to make the crossing.  There is also the possibility of conditional code with some sort of signal that it is the IDE, not the compiler or runtime.


There are a few other needs for the editor that the core would need to provide:

  1. offsets, not just line and column for AST nodes
  2. parser recovery so incomplete syntax in one method allows the rest to be parsed and compiled
  3. remove assumptions of short-lived processes like compiler and allow for long-running processes like editor



From: Jochen Theodorou <blackdrag@gmx.org>
Sent: Sunday, February 17, 2019 1:17 PM
To: dev@groovy.apache.org
Subject: Re: [VOTE] Polish the generics type syntax for closure
 
On 17.02.19 18:31, Milles, Eric (TR Tech, Content & Ops) wrote:
[...]
> So, parser recovery is new development.  And interpretation of new
> syntax in the absence of running all AST transforms to completion is new
> development.

and is there a way to make things more easy? I mean I would prefer to be
able to give eclipse an unmodified compiler - or at least one that does
not need to be source patched

bye Jochen