mapping.py

"""Normalize structured JSON into the checker's flat schema.

Real systems export noon data with their own field names, often wrapped inside
a query result like {"select * from ...": [ { ...record... } ]}. This module:

  1. unwraps nested wrappers down to the actual record, and
  2. maps known source field names (case-insensitive) onto our schema keys,
     including summing grade-level fuel ROB into total FO / DO ROB.

Already-canonical dicts (from the LLM/regex parsers) pass through unchanged.
"""

# canonical schema key -> accepted source keys (lower-cased on lookup)
ALIASES = {
    "vessel_name":    ["vessel_name"],
    "report_date":    ["report_date", "utc_date_time"],
    "position":       ["position"],
    "course_deg":     ["course_deg", "course", "heading", "course_made_good"],
    "steaming_hours": ["steaming_hours", "steaming_time_hrs"],
    "distance_run":   ["distance_run", "observerd_distance", "observed_distance",
                       "distancetravelled", "distance_travelled_actual"],
    "avg_speed":      ["avg_speed", "speed", "avg_speed_kn"],
    "rpm_avg":        ["rpm_avg", "merpm", "me_rpm"],
    "slip_pct":       ["slip_pct", "slip"],
    "me_power_kw":    ["me_power_kw", "me_shaft_power_kw"],
    "me_load_pct":    ["me_load_pct", "me_load"],
    "me_fo_cons":     ["me_fo_cons", "me_consumption"],
    "ae_fo_cons":     ["ae_fo_cons", "ae_consumption"],
    "boiler_fo_cons": ["boiler_fo_cons", "boiler_consumption"],
    "cyl_oil_cons":   ["cyl_oil_cons"],
    "fo_rob":         ["fo_rob"],
    "fo_rob_prev":    ["fo_rob_prev"],
    "fo_bunkered":    ["fo_bunkered"],
    "do_rob":         ["do_rob"],
    "fw_rob":         ["fw_rob"],
    "fw_produced":    ["fw_produced"],
    "draft_fwd":      ["draft_fwd", "draftfwd"],
    "draft_aft":      ["draft_aft", "draftaft"],
    "wind_force":     ["wind_force", "windforce"],
    "sea_state":      ["sea_state"],
    "swell_m":        ["swell_m"],
    "air_temp":       ["air_temp", "ambient_air_temperature"],
    "sea_temp":       ["sea_temp"],
}

# When only per-grade ROB is present, total FO = residual grades, DO = distillates.
FO_ROB_PARTS = ["rob_vlsfo", "rob_hsfo", "rob_ulsfo", "rob_lfo"]
DO_ROB_PARTS = ["rob_lsmgo", "rob_mdo", "rob_go_do"]

# every source key we recognise (used to spot the real record while unwrapping)
_KNOWN = set()
for _v in ALIASES.values():
    _KNOWN.update(_v)
_KNOWN.update(FO_ROB_PARTS)
_KNOWN.update(DO_ROB_PARTS)


def _looks_like_record(d) -> bool:
    return isinstance(d, dict) and bool({k.lower() for k in d} & _KNOWN)


def _unwrap(obj, depth=0):
    """Descend through wrapper dicts/lists to the first object that looks like
    an actual noon record."""
    if depth > 6:
        return obj
    if isinstance(obj, list):
        return _unwrap(obj[0], depth + 1) if obj else {}
    if isinstance(obj, dict):
        if _looks_like_record(obj):
            return obj
        for v in obj.values():
            if isinstance(v, (dict, list)):
                inner = _unwrap(v, depth + 1)
                if _looks_like_record(inner):
                    return inner
    return obj


def _num_or_none(v):
    try:
        return float(v)
    except (TypeError, ValueError):
        return None


def normalize_report(obj) -> dict:
    """Return a dict using the checker's schema keys. Safe on already-canonical
    input (it simply passes the recognised keys through)."""
    rec = _unwrap(obj)
    if not isinstance(rec, dict):
        return {}
    low = {k.lower(): v for k, v in rec.items()}

    out = {}
    for canon, sources in ALIASES.items():
        for s in sources:
            if s in low and low[s] not in (None, ""):
                out[canon] = low[s]
                break

    # Derive total FO / DO ROB from per-grade fields if not already mapped.
    if "fo_rob" not in out:
        parts = [_num_or_none(low.get(p)) for p in FO_ROB_PARTS]
        parts = [p for p in parts if p is not None]
        if parts:
            out["fo_rob"] = round(sum(parts), 2)
    if "do_rob" not in out:
        parts = [_num_or_none(low.get(p)) for p in DO_ROB_PARTS]
        parts = [p for p in parts if p is not None]
        if parts:
            out["do_rob"] = round(sum(parts), 2)

    return out