from __future__ import annotations

from collections.abc import Iterable, Mapping
from difflib import SequenceMatcher

from .formatting import h16
from .model import Instruction
from .rom import Rom


DISPLAY_PRINTABLE = set(range(0x20, 0x7F))
DISPLAY_PUNCTUATION = set(b" ./:-+,%()[]")
DEFAULT_SEARCH_TERMS = ("CONNECT",)
MOV_IW_FIRST_OPCODE = 0x58
MOV_IW_LAST_OPCODE = 0x5F


def analyze_lcd_text(
    rom: Rom,
    instructions: Mapping[int, Instruction] | Iterable[Instruction] | None = None,
    *,
    start: int | None = None,
    end: int | None = None,
    search_terms: Iterable[str] = DEFAULT_SEARCH_TERMS,
    max_candidates: int = 240,
) -> dict[str, object]:
    """Find likely fixed-width LCD/menu strings and their nearby raw xrefs.

    The firmware stores some display text as inline menu/script records rather
    than as plain null-terminated strings. This pass scans the ROM bytes
    directly, then correlates likely text fields with decoded and raw immediate
    address loads such as ``MOV:I.W #H'63D4, R0``.
    """

    lower = 0 if start is None else max(0, start)
    upper = rom.end if end is None else min(rom.end, end)
    candidates = _unique_candidates(
        [
            *_ff_terminated_candidates(rom, lower, upper),
            *_printable_run_candidates(rom, lower, upper),
        ],
    )
    candidates = sorted(candidates, key=lambda item: (-float(item["score"]), int(item["address"])))[:max_candidates]
    candidates.sort(key=lambda item: int(item["address"]))

    instruction_list = _instruction_sequence(instructions)
    xrefs_by_address = _xref_map(rom, candidates, instruction_list)
    for candidate in candidates:
        xrefs = xrefs_by_address.get(int(candidate["address"]), [])
        if xrefs:
            candidate["xrefs"] = xrefs
            candidate["xref_count"] = len(xrefs)

    regions = _group_regions(candidates)
    searches = [_search_term(rom, candidates, term) for term in search_terms]
    return {
        "strings": candidates,
        "regions": regions,
        "searches": searches,
        "notes": [
            "LCD text scan is byte-oriented and conservative; strings may be inline script fields.",
            "Raw xrefs include MOV:I.W immediates to the string address and nearby record prefixes.",
        ],
    }


def lcd_text_comment_for_instruction(analysis: Mapping[str, object] | None, address: int) -> str:
    if not analysis:
        return ""
    for candidate in analysis.get("strings", []):
        if not isinstance(candidate, Mapping):
            continue
        for xref in candidate.get("xrefs", []):
            if isinstance(xref, Mapping) and int(xref.get("address", -1)) == address:
                text = str(candidate.get("trimmed") or candidate.get("text") or "").strip()
                return f"LCD text xref {h16(int(candidate['address']))} {text!r}"
    return ""


def _instruction_sequence(
    instructions: Mapping[int, Instruction] | Iterable[Instruction] | None,
) -> list[Instruction]:
    if instructions is None:
        return []
    values = instructions.values() if isinstance(instructions, Mapping) else instructions
    return sorted(values, key=lambda ins: ins.address)


def _ff_terminated_candidates(rom: Rom, start: int, end: int) -> list[dict[str, object]]:
    candidates: list[dict[str, object]] = []
    address = start
    while address < end:
        if rom.u8(address) not in DISPLAY_PRINTABLE:
            address += 1
            continue
        text_start = address
        raw = bytearray()
        while address < end and rom.u8(address) in DISPLAY_PRINTABLE:
            raw.append(rom.u8(address))
            address += 1
        ff_count = 0
        cursor = address
        while cursor < end and rom.u8(cursor) == 0xFF:
            ff_count += 1
            cursor += 1
        if ff_count and _looks_like_lcd_text(raw):
            candidates.append(_candidate(text_start, raw, "ff_terminated", ff_count=ff_count))
        address = max(cursor, address + 1)
    return candidates


def _printable_run_candidates(rom: Rom, start: int, end: int) -> list[dict[str, object]]:
    candidates: list[dict[str, object]] = []
    address = start
    while address < end:
        if rom.u8(address) not in DISPLAY_PRINTABLE:
            address += 1
            continue
        text_start = address
        raw = bytearray()
        while address < end and rom.u8(address) in DISPLAY_PRINTABLE:
            raw.append(rom.u8(address))
            address += 1
        if _looks_like_lcd_text(raw, allow_long=True):
            candidates.append(_candidate(text_start, raw, "printable_run", ff_count=0))
        address += 1
    return candidates


def _candidate(address: int, raw: bytearray, kind: str, *, ff_count: int) -> dict[str, object]:
    text = raw.decode("ascii", errors="replace")
    trimmed = text.strip()
    score = _display_score(raw, kind, ff_count)
    payload: dict[str, object] = {
        "address": address,
        "length": len(raw),
        "text": text,
        "trimmed": trimmed,
        "kind": kind,
        "score": round(score, 3),
        "confidence": _confidence(score),
    }
    if ff_count:
        payload["ff_terminators"] = ff_count
    if len(raw) > 32:
        payload["segments"] = _fixed_width_segments(text)
    return payload


def _unique_candidates(candidates: list[dict[str, object]]) -> list[dict[str, object]]:
    by_key: dict[tuple[int, int], dict[str, object]] = {}
    for candidate in candidates:
        key = (int(candidate["address"]), int(candidate["length"]))
        existing = by_key.get(key)
        if existing is None or float(candidate["score"]) > float(existing["score"]):
            by_key[key] = candidate
    return list(by_key.values())


def _looks_like_lcd_text(raw: bytearray, *, allow_long: bool = False) -> bool:
    if len(raw) < 4:
        return False
    if len(raw) > 80 and not allow_long:
        return False
    trimmed = bytes(raw).strip()
    if len(trimmed) < 2:
        return False
    good = sum(1 for value in trimmed if _display_char_score(value) > 0)
    return good / max(len(trimmed), 1) >= 0.78


def _display_char_score(value: int) -> float:
    if 0x41 <= value <= 0x5A or 0x30 <= value <= 0x39:
        return 1.0
    if value == 0x20:
        return 0.8
    if value in DISPLAY_PUNCTUATION:
        return 0.7
    if 0x61 <= value <= 0x7A:
        return 0.35
    return 0.0


def _display_score(raw: bytearray, kind: str, ff_count: int) -> float:
    trimmed = bytes(raw).strip()
    if not trimmed:
        return 0.0
    char_score = sum(_display_char_score(value) for value in trimmed) / len(trimmed)
    length_bonus = 0.15 if len(raw) in {8, 10, 16, 17, 18, 19, 20} else 0.0
    terminator_bonus = min(ff_count, 3) * 0.08
    kind_bonus = 0.08 if kind == "ff_terminated" else 0.0
    long_penalty = 0.2 if len(raw) > 40 else 0.0
    return max(0.0, char_score + length_bonus + terminator_bonus + kind_bonus - long_penalty)


def _confidence(score: float) -> str:
    if score >= 1.05:
        return "high"
    if score >= 0.82:
        return "medium"
    return "low"


def _fixed_width_segments(text: str) -> list[dict[str, object]]:
    segments: list[dict[str, object]] = []
    for width in (10, 16, 18, 20):
        if len(text) < width * 2:
            continue
        chunks = [text[index : index + width] for index in range(0, len(text), width)]
        useful = [chunk.strip() for chunk in chunks if len(chunk.strip()) >= 2]
        if len(useful) >= 2:
            segments.append({"width": width, "chunks": chunks})
    return segments[:3]


def _xref_map(
    rom: Rom,
    candidates: list[dict[str, object]],
    instructions: list[Instruction],
) -> dict[int, list[dict[str, object]]]:
    addresses = [int(candidate["address"]) for candidate in candidates]
    target_to_address: dict[int, int] = {}
    for address in addresses:
        for delta in range(-4, 5):
            target = address + delta
            if 0 <= target <= 0xFFFF:
                target_to_address.setdefault(target, address)

    xrefs: dict[int, list[dict[str, object]]] = {address: [] for address in addresses}
    _add_decoded_xrefs(xrefs, target_to_address, instructions)
    _add_raw_mov_iw_xrefs(xrefs, target_to_address, rom)
    return {address: refs for address, refs in xrefs.items() if refs}


def _add_decoded_xrefs(
    xrefs: dict[int, list[dict[str, object]]],
    target_to_address: Mapping[int, int],
    instructions: list[Instruction],
) -> None:
    for ins in instructions:
        for target, candidate_address in target_to_address.items():
            needle = f"H'{target:04X}"
            if needle in ins.operands:
                xrefs[candidate_address].append(
                    {
                        "address": ins.address,
                        "kind": "decoded_operand",
                        "target": target,
                        "delta": target - candidate_address,
                        "instruction": ins.text,
                    },
                )


def _add_raw_mov_iw_xrefs(
    xrefs: dict[int, list[dict[str, object]]],
    target_to_address: Mapping[int, int],
    rom: Rom,
) -> None:
    data = rom.data
    for address in range(0, max(len(data) - 2, 0)):
        opcode = data[address]
        if not MOV_IW_FIRST_OPCODE <= opcode <= MOV_IW_LAST_OPCODE:
            continue
        target = (data[address + 1] << 8) | data[address + 2]
        candidate_address = target_to_address.get(target)
        if candidate_address is None:
            continue
        register = f"R{opcode - MOV_IW_FIRST_OPCODE}"
        xref: dict[str, object] = {
            "address": address,
            "kind": "raw_mov_iw",
            "target": target,
            "delta": target - candidate_address,
            "register": register,
            "instruction": f"MOV:I.W #{h16(target)}, {register}",
        }
        bsr = _following_bsr(data, address + 3)
        if bsr:
            xref["following_bsr"] = bsr
        xrefs[candidate_address].append(xref)


def _following_bsr(data: bytes, address: int) -> dict[str, object] | None:
    if address + 2 >= len(data) or data[address] != 0x1E:
        return None
    displacement = (data[address + 1] << 8) | data[address + 2]
    if displacement & 0x8000:
        displacement -= 0x10000
    target = (address + 3 + displacement) & 0xFFFF
    return {"address": address, "target": target, "instruction": f"BSR {h16(target)}"}


def _group_regions(candidates: list[dict[str, object]]) -> list[dict[str, object]]:
    regions: list[dict[str, object]] = []
    current: list[dict[str, object]] = []
    previous_end: int | None = None
    for candidate in candidates:
        address = int(candidate["address"])
        length = int(candidate["length"])
        if current and previous_end is not None and address - previous_end > 0x80:
            _append_region(regions, current)
            current = []
        current.append(candidate)
        previous_end = address + length
    if current:
        _append_region(regions, current)
    return regions


def _append_region(regions: list[dict[str, object]], candidates: list[dict[str, object]]) -> None:
    if len(candidates) < 2:
        return
    start = int(candidates[0]["address"])
    end = max(int(item["address"]) + int(item["length"]) for item in candidates)
    regions.append(
        {
            "start": start,
            "end": end,
            "count": len(candidates),
            "samples": [str(item.get("trimmed") or item.get("text")) for item in candidates[:8]],
        },
    )


def _search_term(rom: Rom, candidates: list[dict[str, object]], term: str) -> dict[str, object]:
    raw = term.encode("ascii", errors="ignore")
    literal_hits = _literal_hits(rom.data, raw)
    folded = term.upper()
    candidate_hits = [
        {
            "address": int(candidate["address"]),
            "text": candidate["text"],
            "trimmed": candidate["trimmed"],
        }
        for candidate in candidates
        if folded in str(candidate.get("text", "")).upper()
    ]
    near_matches = _near_matches(candidates, term)
    return {
        "term": term,
        "literal_hits": literal_hits,
        "candidate_hits": candidate_hits,
        "near_matches": near_matches,
        "status": "found" if literal_hits or candidate_hits else "not_found",
    }


def _literal_hits(data: bytes, needle: bytes) -> list[int]:
    if not needle:
        return []
    hits: list[int] = []
    start = 0
    upper_data = data.upper()
    upper_needle = needle.upper()
    while True:
        index = upper_data.find(upper_needle, start)
        if index < 0:
            return hits
        hits.append(index)
        start = index + 1


def _near_matches(candidates: list[dict[str, object]], term: str) -> list[dict[str, object]]:
    normalized_term = _normalize_text(term)
    matches: list[dict[str, object]] = []
    for candidate in candidates:
        normalized = _normalize_text(str(candidate.get("trimmed") or candidate.get("text") or ""))
        if not normalized:
            continue
        ratio = SequenceMatcher(None, normalized_term, normalized).ratio()
        if ratio >= 0.34:
            matches.append(
                {
                    "address": int(candidate["address"]),
                    "text": candidate["text"],
                    "trimmed": candidate["trimmed"],
                    "score": round(ratio, 3),
                },
            )
    matches.sort(key=lambda item: (-float(item["score"]), int(item["address"])))
    return matches[:12]


def _normalize_text(text: str) -> str:
    return "".join(char for char in text.upper() if "A" <= char <= "Z" or "0" <= char <= "9")