Original Lynxer surface: status

The original interpreter exposed a larger built-in and module surface than the standalone C++ runtime. This page catalogues it so programs and notes written against the original have somewhere to look, and records what has since been brought across. limitations documents current runtime and standard-library constraints.

Three statuses are used:

• Implemented — the name works today; the notes point at the current built-ins.
• Not implemented — the name is recognised and fails with <name>() is not supported in Lynxer yet.
• Not planned — a deliberate, permanent non-goal.

Pointers and raw addresses

The original had C-style address and function-pointer built-ins.

Addresses are plain integers, as the original documented, and the raw forms are implemented. nativeCall restricts signatures to the shapes the runtime supports and cannot make an arbitrary ABI safe — an invalid address can still crash the process. Prefer the structured built-ins when they fit — typed blocks (memoryBlockAllocate/memoryBlockGet/memoryBlockSet), native structs (nativeStructAllocate/nativeStructGet/nativeStructSet over a layout string), and owned handles (nativeHandleAllocate/nativeHandleAddress/ nativeHandleFree); see builtins.

Advanced memory layout

The original could model typed blocks and C structs directly in the allocator. All of it is implemented, including page protection and atomic/volatile access.

A layout is a comma-separated type name list, e.g. "int32 id, float64 score"; see builtins.

Native synchronization

Caveat: Lynxer evaluates on one interpreter thread under a single lock, and these handles sit on top of the cooperative nativeThread* model. A blocking wait releases the interpreter lock so another thread can run, but a wait on a resource no other thread will release deadlocks. See builtins.

Async

The async* family is implemented: asyncRun, asyncGather, asyncSleep, the asyncPoll* set, and the timer/wakeup built-ins. The async language syntax also exists — a local async name(params){...} definition called as async.name(args) — but it is eager, not a coroutine. asyncRun starts thread-backed tasks; await joins a task handle, and asyncGather joins the handles it receives. Tasks interleave at blocking or yield points, while the global interpreter lock prevents simultaneous Lynxer evaluation. See async and builtins.

Python bridging

Why: Lynxer does not ship or link a Python runtime. Anything that depended on rawPy { } — including the original tkinter modules and tuple interop — has no direct equivalent. turtle is separately implemented over graphics.

FFI and native-module handles

Native code is reached through the documented native-module ABI: an .so exporting lynxer_module_init_v1 is imported by name (importAs("libx.so", "x")), or loaded explicitly with nativeModuleLoad and inspected by name. Both share the same loader and registration rules; the ffi* family is the plain-symbol alternative. See builtins.

Modules not ported

Toolchain surface removed

Quick map: original → now

See also

• limitations — current runtime and standard-library constraints.
• builtins — every built-in the interpreter does implement.
• native-module-abi — calling native code.