r"""Orchestrate parsing an X12 835 ERA file into a :class:`ParseResult835`.

Single-pass walker over the tokenized segment list:

- ISA / GS / ST / BPR / TRN — envelope + financial + trace
- Loop 1000A (N1*PR + N3/N4/PER/REF) — payer
- Loop 1000B (N1*PE + N3/N4) — payee
- Loop 2100 (LX + CLP + CAS + NM1 + REF + DTM + QTY) — claim payment
- Loop 2110 (SVC + CAS + REF + DTM + LQ) — service payment

Errors at the file level raise :class:`CycloneParseError`. Per-segment
deficiencies are recorded in :class:`ValidationReport` and never abort
the whole batch.
"""

from __future__ import annotations

import logging
from datetime import date
from decimal import Decimal
from typing import Any

from cyclone.parsers.exceptions import CycloneParseError
from cyclone.parsers.models import Address, BatchSummary, Envelope
from cyclone.parsers.models_835 import (
    ClaimAdjustment,
    ClaimPayment,
    FinancialInfo,
    ParseResult835,
    Payer835,
    Payee835,
    ReassociationTrace,
    ServicePayment,
    claim_status_label,
)
from cyclone.parsers.payer import PayerConfig835
from cyclone.parsers.segments import tokenize

log = logging.getLogger(__name__)


# --------------------------------------------------------------------------- #
# Date parsing
# --------------------------------------------------------------------------- #


def _parse_ccyymmdd(yymmdd: str) -> date:
    """Parse a CCYYMMDD (8-char) or YY?? (6-char) date string."""
    if not yymmdd:
        raise ValueError("empty date")
    if len(yymmdd) == 8:
        return date(int(yymmdd[0:4]), int(yymmdd[4:6]), int(yymmdd[6:8]))
    if len(yymmdd) == 6:
        yy = int(yymmdd[0:2])
        mm = int(yymmdd[2:4])
        dd = int(yymmdd[4:6])
        # Pivot year 30: 00-29 = 2000s, 30-99 = 1900s.
        year = 2000 + yy if yy < 30 else 1900 + yy
        return date(year, mm, dd)
    raise ValueError(f"unexpected date length: {yymmdd!r}")


def _safe_date(s: str) -> date | None:
    try:
        return _parse_ccyymmdd(s) if s else None
    except (ValueError, IndexError):
        return None


# --------------------------------------------------------------------------- #
# Envelope / BPR / TRN
# --------------------------------------------------------------------------- #


def _build_envelope(segments: list[list[str]], input_file: str) -> tuple[Envelope, BatchSummary]:
    """Build the 835 envelope from ISA/GS/ST. Raises on missing/malformed ISA."""
    summary = BatchSummary(input_file=input_file)
    envelope: Envelope | None = None
    for seg in segments:
        if seg[0] == "ISA":
            try:
                envelope = Envelope(
                    sender_id=seg[6].strip(),
                    receiver_id=seg[8].strip(),
                    control_number=seg[13].strip(),
                    transaction_date=date(2026, 1, 1),  # overwritten by BPR
                    implementation_guide=None,
                )
            except (IndexError, ValueError) as exc:
                raise CycloneParseError(f"Bad ISA: {exc}") from exc
        elif seg[0] == "ST" and envelope is not None:
            if len(seg) > 3:
                envelope = envelope.model_copy(update={"implementation_guide": seg[3]})
    if envelope is None:
        raise CycloneParseError("No ISA envelope found")
    return envelope, summary


def _consume_bpr(segments: list[list[str]], idx: int) -> tuple[FinancialInfo, int]:
    seg = segments[idx]
    try:
        handling = seg[1] if len(seg) > 1 else ""
        paid = Decimal(seg[2]) if len(seg) > 2 and seg[2] else Decimal("0")
        credit_debit = seg[3] if len(seg) > 3 else ""
        method = seg[4] if len(seg) > 4 and seg[4] else None
        # BPR10 = Payer Identifier (Origin Company ID), per the CO Medicaid
        # guide: "81-1725341" (TXIX) or "84-0644739" (BHA).
        payer_tax = seg[10] if len(seg) > 10 and seg[10] else None
        # BPR16 = Effective Entry Date (CCYYMMDD). The segment is sparse —
        # positions 11..15 (BPR11..BPR15) are usually empty, so we look for
        # any 8-digit CCYYMMDD token from index 11 onward.
        payment_date = None
        for k in range(11, len(seg)):
            if seg[k] and len(seg[k]) == 8 and seg[k].isdigit():
                payment_date = _safe_date(seg[k])
                if payment_date is not None:
                    break
    except (IndexError, ValueError) as exc:
        raise CycloneParseError(f"Bad BPR: {exc}", segment_index=idx) from exc
    return (
        FinancialInfo(
            handling_code=handling,
            paid_amount=paid,
            credit_debit_flag=credit_debit,
            payment_method=method,
            payer_tax_id=payer_tax,
            payment_date=payment_date,
        ),
        idx + 1,
    )


def _consume_trn(segments: list[list[str]], idx: int) -> tuple[ReassociationTrace, int]:
    seg = segments[idx]
    trace_type = seg[1] if len(seg) > 1 else ""
    trace_number = seg[2] if len(seg) > 2 else ""
    originating = seg[3] if len(seg) > 3 else ""
    payer_id = seg[4] if len(seg) > 4 and seg[4] else None
    return (
        ReassociationTrace(
            trace_type_code=trace_type,
            trace_number=trace_number,
            originating_company_id=originating,
            payer_id=payer_id,
        ),
        idx + 1,
    )


# --------------------------------------------------------------------------- #
# Loop 1000A (payer) / 1000B (payee)
# --------------------------------------------------------------------------- #


def _consume_address(segments: list[list[str]], idx: int) -> tuple[Address | None, int]:
    """Walk N3/N4 (if present) and return the assembled Address, plus new idx."""
    addr: Address | None = None
    while idx < len(segments) and segments[idx][0] in {"N3", "N4"}:
        seg = segments[idx]
        if seg[0] == "N3":
            line1 = seg[1] if len(seg) > 1 else ""
            line2 = seg[2] if len(seg) > 2 else None
            addr = (addr or Address(line1=line1, city="", state="", zip="")).model_copy(
                update={"line1": line1, "line2": line2}
            )
        elif seg[0] == "N4":
            city = seg[1] if len(seg) > 1 else ""
            state = seg[2] if len(seg) > 2 else ""
            zip_ = seg[3] if len(seg) > 3 else ""
            addr = (addr or Address(line1="", city=city, state=state, zip=zip_)).model_copy(
                update={"city": city, "state": state, "zip": zip_}
            )
        idx += 1
    return addr, idx


def _consume_1000a(segments: list[list[str]], idx: int) -> tuple[Payer835, int]:
    """Read N1*PR + N3 + N4 + PER* + REF* up to next N1 or CLP/LX/SE."""
    name = ""
    pid: str | None = None
    contact_url: str | None = None
    # First segment is N1*PR
    seg = segments[idx]
    if seg[0] == "N1" and len(seg) > 1 and seg[1] == "PR":
        # NM103 is the name (or organization name for non-person entities).
        name = seg[3] if len(seg) > 3 and seg[3] else (seg[2] if len(seg) > 2 else "")
        # NM108/NM109 carry the ID code qualifier and value. The CO guide
        # documents XV for the Health Plan ID, but other payers may use
        # PI / 24 / etc., so we accept any qualifier.
        for j, qual in enumerate(seg):
            if qual in {"XV", "PI", "24"} and j + 1 < len(seg):
                pid = seg[j + 1]
                break
    idx += 1
    addr: Address | None = None
    while idx < len(segments) and segments[idx][0] not in {"N1", "LX", "CLP", "SE"}:
        seg = segments[idx]
        if seg[0] in {"N3", "N4"}:
            addr, idx = _consume_address(segments, idx)
            continue
        if seg[0] == "PER":
            # PER03 is the communication number qualifier; PER04 is the
            # value. The CO guide documents PER04 as the URL but in real
            # 835s the URL is tagged with PER03=UR. We scan for the UR
            # qualifier to be safe.
            for j in range(1, len(seg) - 1, 2):
                if seg[j] == "UR" and j + 1 < len(seg):
                    contact_url = seg[j + 1]
                    break
            # Fall back to PER04 directly if it looks like a URL (matches
            # the literal CO guide wording).
            if contact_url is None and len(seg) > 4 and seg[4]:
                value = seg[4]
                if "@" in value or "http" in value or "." in value:
                    contact_url = value
        idx += 1
    return Payer835(name=name, id=pid, address=addr, contact_url=contact_url), idx


def _consume_1000b(segments: list[list[str]], idx: int) -> tuple[Payee835, int]:
    """Read N1*PE + N3 + N4 up to next N1 or CLP/LX/SE."""
    name = ""
    npi = ""
    seg = segments[idx]
    if seg[0] == "N1" and len(seg) > 1 and seg[1] == "PE":
        name = seg[3] if len(seg) > 3 and seg[3] else (seg[2] if len(seg) > 2 else "")
        for j, qual in enumerate(seg):
            if qual == "XX" and j + 1 < len(seg):
                npi = seg[j + 1]
                break
    idx += 1
    addr: Address | None = None
    while idx < len(segments) and segments[idx][0] not in {"N1", "LX", "CLP", "SE"}:
        seg = segments[idx]
        if seg[0] in {"N3", "N4"}:
            addr, idx = _consume_address(segments, idx)
            continue
        idx += 1
    return Payee835(name=name, npi=npi, address=addr), idx


# --------------------------------------------------------------------------- #
# Loop 2100 (claim payment) and 2110 (service payment)
# --------------------------------------------------------------------------- #


def _cas_segments(segments: list[list[str]], idx: int) -> tuple[list[ClaimAdjustment], int]:
    """Read consecutive CAS segments starting at ``idx`` and return aggregated adjustments."""
    adjustments: list[ClaimAdjustment] = []
    while idx < len(segments) and segments[idx][0] == "CAS":
        seg = segments[idx]
        try:
            group = seg[1] if len(seg) > 1 else ""
            # Each CAS carries up to 6 reason/amount/quantity triplets
            # (elements 2..19). We expand each one into its own row.
            # Indexes 2,3,4 | 5,6,7 | 8,9,10 | 11,12,13 | 14,15,16 | 17,18,19
            triplet_starts = [2, 5, 8, 11, 14, 17]
            for start in triplet_starts:
                if len(seg) <= start:
                    break
                reason = seg[start]
                if not reason:
                    continue
                amount_s = seg[start + 1] if len(seg) > start + 1 else ""
                qty_s = seg[start + 2] if len(seg) > start + 2 else ""
                amount = Decimal(amount_s) if amount_s else Decimal("0")
                quantity = Decimal(qty_s) if qty_s else None
                adjustments.append(
                    ClaimAdjustment(
                        group_code=group,
                        reason_code=reason,
                        amount=amount,
                        quantity=quantity,
                    )
                )
        except (IndexError, ValueError) as exc:
            log.warning("Bad CAS at segment %d: %s", idx, exc)
        idx += 1
    return adjustments, idx


def _consume_service_payment(segments: list[list[str]], idx: int, line_no: int) -> tuple[ServicePayment, int, list[list[str]]]:
    """Read SVC + its child segments. Returns (ServicePayment, new_idx, raw_segments).

    The third return value is the list of raw segments consumed (SVC itself
    plus any DTM/REF/CAS children) so the caller can fold them into the
    claim's raw_segments. :class:`ServicePayment` does not carry a
    ``raw_segments`` field per the 835 spec, so we surface them at the
    claim level instead.
    """
    seg = segments[idx]
    raw: list[list[str]] = [seg]
    qualifier = ""
    code = ""
    modifiers: list[str] = []
    if len(seg) > 1 and seg[1]:
        svc01 = seg[1].split(":")
        if len(svc01) > 0:
            qualifier = svc01[0]
        if len(svc01) > 1:
            code = svc01[1]
        modifiers = [m for m in svc01[2:] if m]
    charge = Decimal(seg[2]) if len(seg) > 2 and seg[2] else Decimal("0")
    payment = Decimal(seg[3]) if len(seg) > 3 and seg[3] else Decimal("0")
    units: Decimal | None = None
    unit_type: str | None = None
    # X12 005010X221A1: SVC04 = Unit or Basis for Measurement Code
    # (UN, MJ, DA, ...); SVC05 = Service Unit Count.
    if len(seg) > 4 and seg[4]:
        unit_type = seg[4]
    if len(seg) > 5 and seg[5]:
        try:
            units = Decimal(seg[5])
        except Exception:
            units = None
    # Some payers omit SVC04 but still mean "UN" when SVC05 carries a
    # count. Be conservative: only default to UN if no other value was
    # set above.
    if unit_type is None and units is not None:
        unit_type = "UN"
    service_date: date | None = None
    ref_benefit_plan: str | None = None
    adjustments: list[ClaimAdjustment] = []

    idx += 1
    while idx < len(segments) and segments[idx][0] not in {"SVC", "LX", "CLP", "SE"}:
        s = segments[idx]
        raw.append(s)
        if s[0] == "DTM" and len(s) > 2 and s[1] == "472":
            service_date = _safe_date(s[2]) if s[2] else None
        elif s[0] == "REF" and len(s) > 2 and s[1] == "0K":
            ref_benefit_plan = s[2]
        elif s[0] == "CAS":
            cas, idx = _cas_segments(segments, idx)
            raw.extend(c[0] for c in cas if False)  # CAS rows are already in adjustments; nothing more to do
            adjustments.extend(cas)
            continue
        idx += 1

    return (
        ServicePayment(
            line_number=line_no,
            procedure_qualifier=qualifier,
            procedure_code=code,
            modifiers=modifiers,
            charge=charge,
            payment=payment,
            units=units,
            unit_type=unit_type,
            service_date=service_date,
            adjustments=adjustments,
            ref_benefit_plan=ref_benefit_plan,
        ),
        idx,
        raw,
    )


def _consume_claim_payment(segments: list[list[str]], idx: int) -> tuple[ClaimPayment, int]:
    """Read one CLP and all its child segments (CAS / NM1 / REF / DTM / QTY / SVC)."""
    clp = segments[idx]
    raw: list[list[str]] = [clp]
    try:
        pcn = clp[1] if len(clp) > 1 else ""
        status = clp[2] if len(clp) > 2 else ""
        total_charge = Decimal(clp[3]) if len(clp) > 3 and clp[3] else Decimal("0")
        total_paid = Decimal(clp[4]) if len(clp) > 4 and clp[4] else Decimal("0")
        patient_resp = Decimal(clp[5]) if len(clp) > 5 and clp[5] else None
        claim_filing = clp[6] if len(clp) > 6 and clp[6] else None
        original_claim_id = clp[7] if len(clp) > 7 and clp[7] else None
        facility_type = clp[8] if len(clp) > 8 and clp[8] else None
        # CLP09 is a single element (no composite qualifier on the 835 side).
        frequency = clp[9] if len(clp) > 9 and clp[9] else None
    except (IndexError, ValueError) as exc:
        raise CycloneParseError(f"Bad CLP: {exc}", segment_index=idx) from exc

    service_payments: list[ServicePayment] = []
    ref_benefit_plan: str | None = None
    per_diem: Decimal | None = None
    status_label = claim_status_label(status)

    idx += 1
    while idx < len(segments) and segments[idx][0] not in {"CLP", "SE"}:
        s = segments[idx]
        # Only consume this CLP's children; an LX starts a new 2100 loop.
        if s[0] == "LX":
            # Skip the LX and expect an SVC; if absent, stop.
            raw.append(s)
            idx += 1
            if idx < len(segments) and segments[idx][0] == "SVC":
                line_no = (
                    int(s[1]) if len(s) > 1 and s[1].isdigit() else len(service_payments) + 1
                )
                sp, idx, svc_raw = _consume_service_payment(segments, idx, line_no)
                raw.extend(svc_raw)
                service_payments.append(sp)
            continue
        if s[0] == "SVC":
            sp, idx, svc_raw = _consume_service_payment(segments, idx, len(service_payments) + 1)
            raw.extend(svc_raw)
            service_payments.append(sp)
            continue
        raw.append(s)
        if s[0] == "CAS":
            cas, idx = _cas_segments(segments, idx)
            # Top-level claim adjustments are aggregated into a synthesized
            # ServicePayment with line_number=0 if there are no SVC rows.
            # Per the spec, ClaimPayment doesn't carry adjustments directly,
            # so we record them on the first service payment if any,
            # otherwise drop them on the floor (rare in practice).
            if service_payments:
                first = service_payments[0]
                service_payments[0] = first.model_copy(
                    update={"adjustments": first.adjustments + cas}
                )
            continue
        if s[0] == "REF" and len(s) > 2 and s[1] == "CE":
            ref_benefit_plan = s[2]
        elif s[0] == "QTY" and len(s) > 2 and s[1] == "CA":
            try:
                per_diem = Decimal(s[2]) if s[2] else None
            except ValueError:
                per_diem = None
        elif s[0] == "NM1":
            # Patient (QC) / service-provider (1P) — captured in raw_segments.
            # The 835 spec doesn't require a structured patient model in v1.
            pass
        elif s[0] == "DTM":
            # Claim-level dates — captured in raw_segments.
            pass
        idx += 1

    return (
        ClaimPayment(
            payer_claim_control_number=pcn,
            status_code=status,
            status_label=status_label,
            total_charge=total_charge,
            total_paid=total_paid,
            patient_responsibility=patient_resp,
            claim_filing_indicator=claim_filing,
            original_claim_id=original_claim_id,
            facility_type=facility_type,
            frequency_code=frequency,
            per_diem_covered_days=per_diem,
            ref_benefit_plan=ref_benefit_plan,
            service_payments=service_payments,
            raw_segments=raw,
        ),
        idx,
    )


# --------------------------------------------------------------------------- #
# Top-level orchestrator
# --------------------------------------------------------------------------- #


def parse(text: str, payer_config: PayerConfig835, input_file: str = "") -> ParseResult835:  # noqa: ARG001 — payer_config reserved for future use
    """Parse a complete 835 ERA document and return a :class:`ParseResult835`.

    ``payer_config`` is accepted for symmetry with the 837P parser and
    for future payer-aware heuristics; in v1 the structural validation
    runs against the full result via :func:`cyclone.parsers.validator_835.validate`.
    """
    segments = tokenize(text)
    envelope, summary = _build_envelope(segments, input_file=input_file)

    financial: FinancialInfo | None = None
    trace: ReassociationTrace | None = None
    payer = Payer835(name="", id=None, address=None, contact_url=None)
    payee = Payee835(name="", npi="", address=None)
    claims: list[ClaimPayment] = []

    i = 0
    n_bprs = 0  # CO Medicaid sends multiple BPRs in one 835; we sum them.
    # Pre-envelope segments (ISA/GS) are already consumed by _build_envelope.
    # Walk forward looking for the header (BPR/TRN), then loops.
    while i < len(segments):
        seg = segments[i]
        if seg[0] == "ST":
            # Skip the ST control segment.
            i += 1
            continue
        if seg[0] == "BPR":
            new_fin, i = _consume_bpr(segments, i)
            if financial is None:
                financial = new_fin
            else:
                # CO Medicaid "split payment" pattern: multiple BPRs whose
                # BPR02 amounts sum to the total paid. X12 835 only allows
                # one BPR per transaction set; we accept the data and warn.
                # Keep the first BPR's handling_code, tax_id, payment_date;
                # only the paid_amount is additive.
                financial = financial.model_copy(update={
                    "paid_amount": financial.paid_amount + new_fin.paid_amount,
                })
            n_bprs += 1
            continue
        if seg[0] == "TRN":
            trace, i = _consume_trn(segments, i)
            continue
        if seg[0] == "N1" and len(seg) > 1 and seg[1] == "PR":
            payer, i = _consume_1000a(segments, i)
            continue
        if seg[0] == "N1" and len(seg) > 1 and seg[1] == "PE":
            payee, i = _consume_1000b(segments, i)
            continue
        if seg[0] == "CLP":
            try:
                claim, i = _consume_claim_payment(segments, i)
                claims.append(claim)
            except CycloneParseError:
                raise
            except Exception as exc:  # pragma: no cover - safety net
                log.warning("Claim parse failed at segment %d: %s", i, exc)
                i += 1
            continue
        if seg[0] == "LX" and not claims and not payee.npi:
            # Defensive: a stray LX before any CLP (no 1000B seen) is ignored.
            i += 1
            continue
        i += 1

    if financial is None:
        raise CycloneParseError("No BPR (Financial Information) segment found")
    if trace is None:
        raise CycloneParseError("No TRN (Reassociation Trace) segment found")

    # Update the envelope's transaction date from the BPR payment date if available.
    if financial.payment_date is not None:
        envelope = envelope.model_copy(update={"transaction_date": financial.payment_date})

    summary = summary.model_copy(update={
        "control_number": envelope.control_number,
        "transaction_date": envelope.transaction_date,
        "total_claims": len(claims),
    })

    return ParseResult835(
        envelope=envelope,
        financial_info=financial,
        trace=trace,
        payer=payer,
        payee=payee,
        claims=claims,
        multi_bpr=(n_bprs > 1),
        summary=summary,
        validation=None,
    )


__all__ = ["parse"]