Stibium - Language Support Library for Antimony

This is a language support library for Antimony that supports parsing, static analysis, completion, and more. Stibium provides an simple Python API for performing the above tasks and handling the parse tree. Lark is used as the parser backend.

Stibium requires Python 3.7+.

Features

• Parse Antimony source files and generate a navigable tree.
• Error recovery during parse.
• Diagnostics: obtain syntax and semantic errors
• Obtain definition of a name (symbol).
• Obtain the name (symbol) at the provided cursor location.
• Autocomplete reaction rate law (e.g. mass-action)
• Basic autocompletion of species names

Planned Features

• Extend the syntax to parse compartments, events, and models
• Parser Caching
• Better context-sensitive autocompletion, possibly based on the word fragment (prefix) before the cursor. The current autocompletion is generally pretty dumb -- with the exception of rate law completion, Stibium does not check for things like scope, keywords, or the context in general.

Building

• Install the dependencies from requirements.txt.
• python setup.py sdist to build the package for distribution.

Testing

• pytest is required for testing. Install it from dev-requirements.txt.
• pytest test to run all the tests.

Notes on Error Recovery

For now error recovery is based solely on threshold set by simple_stmt. Namely, when an error token is encountered, I backtrack to the last fully parsed simple_stmt node as the last valid state. By inspecting the (LALR) parser stack, all nodes/tokens encountered between the last fully parsed simple_stmt and the error token are formed into an ErrorNode, and the errored token is formed into an ErrorToken. The parser state is then set to as if it just finished parsing a simple_stmt.

In the future, when models and functions are introduced, we will need to modify parse.py so that instead of backtracking to the last simple_stmt, we instead go to the last full statement/model/ function, whichever rule that is.