Originally created to support context-sensitive and regex supporting grammars. It consisted of a recursive descent parser with several optimizations for localized caches and on the fly tokenization (rather than an initial tokenization step). As a result it was fairly fast for a recursive descent parser, only about 3 times slower than ANTLR. It was able to parse more easily certain types of texts and would not run the same regex evaluation more than once from the same point in the parsing process. It also exposed a lot of what is needed to create functionality around autocompletion to the point that I was able to create several smaller DSLs with IDE like support. The main problem with it was that although decently fast for what it did it was not fast enough for a real time typing experience.
To help with that I started using ANTLR 4. Although ANTLR is faster and able to handle much bigger inputs, it lacked functionality around auto-completion and most suggestions involved simple token recommendations. I tried C3 for auto-completion but also noticed it needed some hand holding in order to know what things it should look for auto-completion. I wanted ANTRL to provide a similar experience as the initial parser I had created provided. This library was enhanced with several utilities to allow ANTRL 4 usage that makes it easier to be used for the items below among others:
- Auto-completion (based on a recursive descent implementation that leverages ANTLR 4 ATN states)
- auto-completion suggests both tokens and entire parse rules. There is no need to tell it which ones you are looking for, you are just responsible for providing the auto-completion text for the parse rules recommended. This allows you to provide auto-completion suggestions/examples that completes the entire parse rule.
- Highlighting (Currently HTML based, browser focused)
- Formatting
- Parse Tree Printing in ASCII format
- FastTokenRewriter
- While using the existing implementation of ANTRL TokenRewriter it was clear it was very slow for the use above, specially for real time highlighting. After examining its implementation and points of slowness I created this version that is significantly faster
- This also supports providing text change computations. Do you want to highlight and format a file using different listeners focused on doing each action individually? with computations, you can, you can merge them and create the final output that contains both sets of changes. This allows you to not have to write code only used for highlighting, another one for formatting and a third that correctly does both at the same time.
ANTLR Completion with Indentation
ANTRL Highlighting and Formatting
public key for validating dependency available from keyserver.ubuntu.com