Scaling Xtext
summary
Xtext scales to 100s of files and individual files up to ~90k lines by replacing lazy EMF behavior with effectively immutable models, enabling parallel parsing, linking and validation across multiple CPU cores, and adding resource caching to deserialize binary models instead of reparsing source. A custom concurrent editor separates UI from compiler services so huge files stay responsive, and a single Builder abstraction plus generated VS Code artifacts preserves feature parity across Eclipse and LSP—most changes are custom and kept out of upstream Xtext.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 4:24Prerequisites
- 7:38Clustering Builder
- 11:51Massage EMF
- 13:41Performance
- 15:52File sizes
- 19:33Resource sharing
- 20:31Language complexity
- 23:52EMF switches
- 26:16Language inheritance
- 27:42Shared runtime module
- 28:24Juice overrides
- 28:58Sigasi Studio
- 33:31Xtext in Sigasi
- 34:30Why not upstream
