semantic/ROADMAP.md

Semantic Roadmap

Semantic Code produces data around code’s structure and meaning. This is made available via GraphQL APIs, enabling others (both at GitHub and elsewhere) to build features.

High-level goals:

• Production ready in Q2 FY2018.
• Help developers understand, navigate, and improve their code.
• Lower barriers to contributions, e.g. by eliminating conflicts from renaming variables.

GraphQL APIs

• Parse tree.
  • Integration with code search 2.0
  • Semantic notifications
• Diff.
• Repository and/or project-level (semantic index)
  • Dependencies
  • Some sort of persistence to “link” together different parse trees/repositories/etc.

Ongoing Work:

• Extend/improve the data provided via GraphQL. API consumers will have to specifically opt in to any new fields in order to receive them.

LSP integration

Enables:

• Find References from blobs
• Go To definition
• Find workspace symbols

Ongoing Work:

• Hosting language servers
• Hosting LSP servers in containers
• Persistence or caching of requests

Semantic analysis

• À la carte syntax
• Cyclomatic complexity
• Modular abstract interpreters
• Type inference
  • Is this program well-typed?
  • What is the type at any particular node?

Ongoing Work:

• type signatures,
• parse/type errors,
• callers/callees,
• labelling symbol declarations (for e.g. jump to…/ToC as well as code search)
• linking symbol references to declarations.

Language support

• Python

Ongoing Work:

• Extend the set of supported languages, guided by data on language usage.
• Provide/improve tooling for writing/testing tree-sitter grammars.
• Review contributions to open-source grammars.
• Automate as much as possible to keep us focused on the core. Can we reduce/offload this in other ways?

Operability/Production readiness

• Containerization/-as-a-service architecture
• Caching
• Performance
• Security review of tree-sitter
• Resolving tree-sitter error recovery hangs

Ongoing Work:

• Performance improvements.
• Resiliency improvements/maintenance/operational excellence.
• Metrics. How does it perform, how are people using it, what data do people care about, what don’t they use at all.

ToC

• Performance with large parse trees
• Depends on improved operability
• Production ready tree-sitter
• ToC in Enterprise

See also our roadmap project.