Explain the lifecycle of a Truffle language implementation from parsing to execution?

Implementing a language on Truffle follows a fairly standard pipeline, even though the actual interpreter logic is developer-written.

flowchart TD
  A[Source code] --> B[Lexer/Parser]
  B --> C[Truffle AST construction]
  C --> D[Tree-walking interpretation]
  D --> E[Node self-specialization on observed types]
  E --> F{Code path hot?}
  F -->|No| D
  F -->|Yes| G[Graal partial evaluation]
  G --> H[Optimized machine code installed]

1. A parser (often hand-written or generated with ANTLR) turns source text into a Truffle-specific AST, where each node subclasses com.oracle.truffle.api.nodes.Node.
2. Execution begins by tree-walking the AST directly, interpreting each node.
3. As described in specialization, nodes rewrite themselves to type-specialized variants as they observe real operand types.
4. Once the Truffle runtime's profiling detects a call target is hot, it triggers Graal to partially evaluate that AST subtree into optimized machine code, which subsequent calls use directly instead of re-interpreting.

Because this whole pipeline is provided by the Truffle framework, language implementers focus almost entirely on step 1 and 2 - writing the parser and the node classes - while specialization and compilation are largely handled by the framework itself.