Inspect generated output

After the tutorial. Use this page once you have run Generate your first class. It explains the files you already produced; there is no additional installation step.

Keep these artifacts separate:

Artifact	Meaning	What to inspect
classes.yidl	Authored definition authority	Property names, resource shape, target paths
runtime.py	Generated schema/assembly library	Record constructor storage names, build_Module, ASSEMBLY_* metadata
model.py	Final generated Python	Direct slots/stores, signature, surviving markers
Executed class	User-visible behavior	Values, callbacks, ordering, exceptions

If you used the tutorial's --out build/first command, open build/first/runtime.py and build/first/model.py from the repository root. If you chose another output directory, use that directory instead. ast.unparse(composable.to_executable_ast()) provides readable Python without requiring you to evaluate a second handwritten code sample. composable.emit_commented() is another inspection surface; comments and hygiene spellings can differ from final AST display.

The tutorial records the actual engine selection in its JSON output. A variable requesting native execution is not evidence of native dispatch. Keep package versions and import origins with reproduction logs.

Questions to answer before refreshing a golden

Does each field have a slot? Does each selected parameter have a store? Are unselected and orphan fields absent? Is x before y despite reversed fact insertion? Can you instantiate the result with the expected signature?

The canonical documentation_classes.py fixture calls the tutorial's behavioral assertions and snapshots its model source. A snapshot matching yesterday is not an independent oracle. Intentional changes must first satisfy the behavior requirement, then receive an inspected output refresh.

For failures, preserve the traceback chain and locate its stage with debugging assembly. There is no promised all-in-one trace flag.

Keep authority, remove disposable output

runtime.py and model.py are generated files in your selected --out directory. Remove that destination if it belongs only to this experiment; retain classes.yidl and generate.py, which are the authored inputs needed to rebuild it. The installation cleanup steps also cover the created environment and wheel directory without deleting authored files.

Choose what to learn next

For a second walkthrough, follow a field from input to output in Matching and data flow. For a generator spread across files, use Split a project and regenerate. The example catalog offers runnable variations. Choose the page that fits your next task; these are not prerequisites for one another.