"""S5.4 — verdirapport (målbilde §7): read-only savings roll-up over SavingsLedger.

MAF-bound via ``ledger`` (which imports the MAF-bound ``verdicts``), so this module CANNOT live in
the framework-neutral ``shared/`` subtree and is deliberately NOT registered in
``tests/test_okf.py``'s ``_MAF_FREE_MODULES``. It is a pure OFFLINE read: it turns the ledger's
already-deduped integer-øre totals into a legible statement of value (per-project + portfolio
totals, flagged cross-dimension overlaps, per-entry provenance). Zero model calls, zero writes.

Correctness rests on CALLING the ledger's roll-up accessors (``per_project_total`` /
``portfolio_total`` / ``overlaps``) — never re-summing ``.entries``, which would double-count a
saving realized under two dimensions (the exact bug the ledger's dimension-free dedup prevents).
"""

from __future__ import annotations

import json

from pydantic import BaseModel

from portfolio_optimiser.costsim import _ore_to_kr_str
from portfolio_optimiser.ledger import SavingsLedger


class ProvenanceLine(BaseModel):
    """One realized ledger entry surfaced at the report's ENTRY level: verbatim ``verdict_id`` +
    ``provenance`` string kept in association, so provenance is legible per-entry in the JSON
    output. Reads ONLY the entry's own fields — never joins the MAF verdict store (Non-Goal)."""

    project_id: str
    dimension: str
    candidate_identity: str
    verdict_id: str
    amount_ore: int
    provenance: str


class ValueReport(BaseModel):
    """The read-only value roll-up over a ``SavingsLedger``: per-project + portfolio totals (integer
    øre, dimension-free-deduped by the ledger accessors), the flagged cross-dimension ``overlaps``,
    and one ``ProvenanceLine`` per ledger entry (deterministically ordered)."""

    per_project: dict[str, int]
    portfolio_total_ore: int
    overlaps: list[tuple[str, str]]
    provenance: list[ProvenanceLine]


def build_value_report(ledger: SavingsLedger) -> ValueReport:
    """Build the value report by CALLING the ledger's deduped roll-up accessors — never summing
    ``.entries`` directly (double-count risk on cross-dimension overlaps).

    The provenance sort key is TOTAL: the leading ``(project_id, dimension, candidate_identity)``
    triple is the ledger's ``_storage_key``, unique for in-process-built ledgers but NOT for a
    ``load()``-ed / hand-authored file (load does no dedup), so two rows sharing the triple would tie
    and their order would follow file insertion order (byte-instability). Appending
    ``verdict_id, provenance, amount_ore`` makes the order total and byte-stable (SC6).
    """
    project_ids = sorted({e.project_id for e in ledger.entries})
    per_project = {pid: ledger.per_project_total(pid) for pid in project_ids}
    ordered_entries = sorted(
        ledger.entries,
        key=lambda e: (
            e.project_id,
            e.dimension,
            e.candidate_identity,
            e.verdict_id,
            e.provenance,
            e.amount_ore,
        ),
    )
    provenance = [
        ProvenanceLine(
            project_id=e.project_id,
            dimension=e.dimension,
            candidate_identity=e.candidate_identity,
            verdict_id=e.verdict_id,
            amount_ore=e.amount_ore,
            provenance=e.provenance,
        )
        for e in ordered_entries
    ]
    return ValueReport(
        per_project=per_project,
        portfolio_total_ore=ledger.portfolio_total(),
        overlaps=ledger.overlaps(),
        provenance=provenance,
    )


def format_report_text(report: ValueReport) -> str:
    """Render the value report as a column-aligned human table: one per-project row (id + realized
    NOK), a portfolio-total footer, and an overlaps section listing the flagged
    ``(project_id, candidate_identity)`` keys (or a ``ingen overlapp`` line when empty). Every øre
    value is formatted through the single byte-pinned money edge ``costsim._ore_to_kr_str``
    (float-free integer ``divmod``), and ``per_project`` is iterated in sorted key order for
    byte-stability (SC6)."""
    lines = [
        "Verdirapport — realiserte besparelser (kun-lesing, ingen modellkall)",
        "",
        f"{'prosjekt':<24} {'realisert':>16}",
    ]
    for pid in sorted(report.per_project):
        lines.append(f"{pid:<24} {_ore_to_kr_str(report.per_project[pid]):>16}")
    lines += [
        "",
        f"{'portefølje totalt':<24} {_ore_to_kr_str(report.portfolio_total_ore):>16}",
        "",
        "Kryss-dimensjon overlapp (talt én gang, flagget):",
    ]
    if report.overlaps:
        for project_id, candidate_identity in report.overlaps:
            lines.append(f"  {project_id} / {candidate_identity}")
    else:
        lines.append("  ingen overlapp")
    return "\n".join(lines)


def dump_report_json(report: ValueReport) -> str:
    """Byte-deterministic JSON serialization of the value report (the repo-wide idiom
    ``sort_keys=True, indent=2`` + trailing LF; mirrors ``costsim.dump_estimate_table`` /
    ``outbox._dump``). ``overlaps`` tuples serialize to JSON arrays."""
    return json.dumps(report.model_dump(), sort_keys=True, indent=2) + "\n"