What causes "Function body is empty" in commaai/openpilot?

Put an expression or a function body into the code file, e.g. 'return a * 2' or a bare expression using env variables

commaai/openpilot · warning · ValueError

Function body is empty

Error message

Function body is empty

What it means

jotpluggler's math_eval evaluates user code from the per-metric code file. _evaluate_user_code() strips whitespace and refuses an empty body - there is literally nothing to eval or wrap in a generated function. This is a fast, clear validation failure before exec/eval machinery runs.

Source

Thrown at openpilot/tools/jotpluggler/math_eval.py:38

def _write_vector(path: str, values: np.ndarray) -> None:
  np.asarray(values, dtype=np.float64).tofile(path)


def _resample_to_reference(ref_t: np.ndarray, src_t: np.ndarray, src_v: np.ndarray) -> np.ndarray:
  ref_t = np.asarray(ref_t, dtype=np.float64).reshape(-1)
  src_t = np.asarray(src_t, dtype=np.float64).reshape(-1)
  src_v = np.asarray(src_v, dtype=np.float64).reshape(-1)
  if ref_t.size == 0 or src_t.size == 0 or src_v.size == 0:
    return np.empty_like(ref_t)
  indices = np.searchsorted(src_t, ref_t, side="right") - 1
  indices = np.clip(indices, 0, src_v.size - 1)
  return src_v[indices]


def _evaluate_user_code(code: str, env: dict):
  stripped = code.strip()
  if not stripped:
    raise ValueError("Function body is empty")

  expr = stripped
  if expr.startswith("return "):
    expr = expr[7:].strip()
  try:
    return eval(expr, env, env)
  except SyntaxError:
    pass

  function_src = "def __jotpluggler_eval__():\n" + textwrap.indent(code, "    ")
  exec(function_src, env, env)
  return env["__jotpluggler_eval__"]()


def main() -> int:
  if len(sys.argv) != 6:
    print("usage: math_eval.py <manifest.json> <globals.py> <code.py> <out_t.bin> <out_v.bin>", file=sys.stderr)
    return 2

View on GitHub (pinned to 516ec1e682)

Solutions

Put an expression or a function body into the code file, e.g. 'return a * 2' or a bare expression using env variables
Check the file actually being read (code_path) - it may not be the file you edited
If the metric is intentionally unused, remove its config entry instead of leaving an empty code file

Example fix

# before (code file contents)

# after
return src_v * 3.28084  # m/s to ft/s

Defensive patterns

Strategy: validation

Validate before calling

code = open(code_path, encoding='utf-8').read()
if not code.strip():
    raise SystemExit(f"metric code file {code_path} is empty - write an expression or remove the metric")

Try / catch

try:
    result = _evaluate_user_code(user_code, env)
except ValueError as e:
    if 'empty' in str(e):
        raise SystemExit(f'fill in or delete {code_path}')
    raise

Prevention

Reject empty code files at metric-creation time (in scaffolding/tooling), not at eval time
Add a trivial smoke test per metric (eval once with sample env) to CI so blank templates fail early

When it happens

Trigger: A metric's code file is empty or contains only whitespace/newlines; template metric created but never filled in; file path resolution returned an empty file.

Common situations: Scaffolding a new custom series and forgetting to write the expression; editor saved an empty buffer; CI fixture generating blank code files.

Related errors

AI-assisted analysis of commaai/openpilot@516ec1e682 (2026-08-15). Data as JSON: /api/errors/5947c04e3b628eae. Report an issue: GitHub.