"""Validacion estatica de dataset_query MBQL antes de enviarlo a Metabase."""

from __future__ import annotations

import collections
import re
from typing import Any


def metabase_mbql_validate(dataset_query: dict) -> list[str]:
    """Valida la estructura de un dataset_query MBQL sin hacer I/O.

    Detecta los errores mas comunes que causan respuestas 400/500 de la API de
    Metabase, permitiendo corregirlos antes del round-trip.

    Checks realizados:
    1. UUIDs duplicados: cualquier ``lib/uuid`` que aparezca mas de una vez en
       el arbol MBQL. Metabase los requiere unicos globalmente por query.
    2. Stage mixing: stages que tienen tanto ``aggregation`` como ``expressions``
       donde las expressions referencian slot names (``sum``, ``sum_N``, etc.)
       generados por aggregations. Esas expressions deben ir en la stage siguiente.
    3. Slot refs rotos: expressions que referencian ``sum_X`` deben tener X menor
       que la cantidad de sums en la stage previa (o misma).
    4. Case structure: nodos ``["case", meta, cases]`` deben tener ``cases``
       como lista de pares ``[cond, result]``.
    5. Name collision: dos expressions con el mismo ``lib/expression-name`` en
       la misma stage.

    Args:
        dataset_query: Dict con la estructura completa del dataset_query MBQL
            tal como lo devuelve GET /api/card/:id o lo construye el caller.
            Debe tener clave ``stages`` (lista de stage dicts). Si no tiene
            ``stages``, se devuelve error de estructura.

    Returns:
        Lista de strings describiendo errores encontrados. Lista vacia si el
        dataset_query es valido segun todos los checks.

    Example:
        >>> errors = metabase_mbql_validate(card["dataset_query"])
        >>> if errors:
        ...     for e in errors:
        ...         print(e)
        ... else:
        ...     print("Query valida")
    """
    errors: list[str] = []

    stages = dataset_query.get("stages")
    if not isinstance(stages, list):
        errors.append("dataset_query.stages ausente o no es lista")
        return errors

    # ---- Check 1: UUIDs duplicados ------------------------------------------
    uuid_locations: list[tuple[str, str]] = []
    _collect_uuids(dataset_query, root="", out=uuid_locations)
    uuid_counter: dict[str, list[str]] = collections.defaultdict(list)
    for uid, path in uuid_locations:
        uuid_counter[uid].append(path)
    for uid, paths in uuid_counter.items():
        if len(paths) > 1:
            errors.append(
                f"Duplicate lib/uuid '{uid}' aparece {len(paths)} veces: "
                + ", ".join(paths[:3])
                + ("..." if len(paths) > 3 else "")
            )

    # ---- Checks por stage ---------------------------------------------------
    for si, stage in enumerate(stages):
        if not isinstance(stage, dict):
            continue
        tag = f"stage[{si}]"

        expressions: list[Any] = stage.get("expressions") or []
        aggregations: list[Any] = stage.get("aggregation") or []

        # Check 5: name collision en expressions
        expr_names: list[str] = []
        for expr in expressions:
            name = _expr_name(expr)
            if name:
                if name in expr_names:
                    errors.append(
                        f"{tag} tiene dos expressions con mismo "
                        f"lib/expression-name '{name}'"
                    )
                else:
                    expr_names.append(name)

        # Check 2: stage mixing
        if aggregations and expressions:
            for expr in expressions:
                slot_refs = _find_slot_refs(expr)
                if slot_refs:
                    ename = _expr_name(expr) or "?"
                    errors.append(
                        f"{tag} mezcla aggregations con expressions "
                        f"post-agg que referencian slot names "
                        f"({', '.join(repr(s) for s in slot_refs)}) "
                        f"en expression '{ename}'. "
                        f"Mover esas expressions a la stage siguiente."
                    )

        # Check 3: slot refs rotos
        # Contar sums en aggregations de esta stage
        sum_count = sum(1 for agg in aggregations if _agg_is_sum(agg))
        for expr in expressions:
            for slot in _find_slot_refs(expr):
                m = re.match(r"sum(?:_(\d+))?$", slot, re.IGNORECASE)
                if m:
                    idx = int(m.group(1)) if m.group(1) else 0
                    if idx >= sum_count:
                        ename = _expr_name(expr) or "?"
                        errors.append(
                            f"{tag} expression '{ename}' referencia "
                            f"'{slot}' que no existe "
                            f"(solo hay {sum_count} sum(s) en aggregation)"
                        )

        # Check 4: case structure
        for expr in expressions:
            _check_case_structure(expr, tag, errors)

    return errors


# ---------------------------------------------------------------------------
# Helpers privados
# ---------------------------------------------------------------------------


def _collect_uuids(
    obj: Any,
    root: str,
    out: list[tuple[str, str]],
) -> None:
    """Recorre obj recursivamente y añade (uuid, path) a out."""
    if isinstance(obj, dict):
        if "lib/uuid" in obj:
            out.append((obj["lib/uuid"], root))
        for k, v in obj.items():
            _collect_uuids(v, f"{root}.{k}" if root else k, out)
    elif isinstance(obj, list):
        for i, item in enumerate(obj):
            _collect_uuids(item, f"{root}[{i}]", out)


def _expr_name(expr: Any) -> str | None:
    """Extrae lib/expression-name del segundo elemento de un nodo MBQL."""
    if isinstance(expr, list) and len(expr) >= 2 and isinstance(expr[1], dict):
        return expr[1].get("lib/expression-name")
    return None


# Patron de slot name: word chars, puede terminar en _N
_SLOT_RE = re.compile(r"^[a-zA-Z_][a-zA-Z0-9_]*(?:_\d+)?$")
# Slots que corresponden a aggregation functions conocidas
_AGG_SLOTS = {
    "sum", "avg", "count", "min", "max",
    "distinct", "cum-sum", "cum-count", "share", "stddev",
}


def _find_slot_refs(obj: Any) -> list[str]:
    """Devuelve lista de slot names encontrados en refs tipo ["field", meta, slot]."""
    slots: list[str] = []
    _collect_slot_refs(obj, slots)
    return slots


def _collect_slot_refs(obj: Any, out: list[str]) -> None:
    if isinstance(obj, list):
        if (
            len(obj) == 3
            and obj[0] == "field"
            and isinstance(obj[1], dict)
            and isinstance(obj[2], str)
            and not obj[1].get("base-type")  # field sin base-type = slot ref
            and _is_slot_name(obj[2])
        ):
            out.append(obj[2])
        else:
            for item in obj:
                _collect_slot_refs(item, out)
    elif isinstance(obj, dict):
        for v in obj.values():
            _collect_slot_refs(v, out)


def _is_slot_name(s: str) -> bool:
    """Devuelve True si s parece un slot name de aggregation."""
    # Slot: nombre sin espacio que es una funcion de agg o variant con sufijo _N
    base = re.sub(r"_\d+$", "", s)
    return base in _AGG_SLOTS


def _agg_is_sum(agg: Any) -> bool:
    """Retorna True si el nodo aggregation es de tipo sum."""
    if isinstance(agg, list) and len(agg) >= 1:
        return str(agg[0]).lower() == "sum"
    return False


def _check_case_structure(expr: Any, tag: str, errors: list[str]) -> None:
    """Valida recursivamente nodos case dentro de una expression."""
    if not isinstance(expr, list):
        return
    if expr and expr[0] == "case":
        ename = _expr_name(expr) or "?"
        # Esperado: ["case", meta, [[cond, result], ...]]
        if len(expr) < 3:
            errors.append(
                f"{tag} expression '{ename}': case con menos de 3 elementos"
            )
            return
        cases = expr[2]
        if not isinstance(cases, list):
            errors.append(
                f"{tag} expression '{ename}': tercer elemento de case "
                f"debe ser lista de pares, got {type(cases).__name__}"
            )
            return
        for i, pair in enumerate(cases):
            if not (isinstance(pair, list) and len(pair) == 2):
                errors.append(
                    f"{tag} expression '{ename}': case[{i}] no es par "
                    f"[cond, result], got {pair!r}"
                )
        # Recursar en ramas
        for pair in cases:
            if isinstance(pair, list):
                for node in pair:
                    _check_case_structure(node, tag, errors)
    else:
        for item in expr:
            _check_case_structure(item, tag, errors)