Tooling for DX have skyrocketed these last years.
Point is it is still hard to develop these tools: what if there is a standard? "What if your users could start writing lint rules immediately?".
As an example of Tree-sitter, what if such a compiler could include every build framework such as salsa, turbo [opinionated thought], VSCode, etc...
All of these tools are what I’d call meta-tools. They allow programming language implementors to increase their productivity in a comparable manner to how developer tools increase the productivity of regular developers.
- Meta-tools must offer such a compelling suite of features that it becomes difficult to refuse.
- The experience of using a meta-tool should also be first-class.
- Meta-tools must fundamentally be extensible: being able to implement theirself 1% needed.
What keeps Rust from being an Inria PhD student’s thesis that nobody uses is its focus on tooling, good compiler diagnostics, and ecosystem
- raise the bar: maybe it’s better code generation for scaffolding projects. Maybe it’s extremely quick deployment based on incremental compilation. Maybe it’s seamless interop with other languages.