US2KR/us2kr/render.py

"""Rendering: MarketPayload (+ optional LLM Report) -> Korean Markdown, and Markdown -> Telegram HTML chunks.

Every number in the output comes from the payload. The Report is used for prose only.
"""

from __future__ import annotations

import re
import unicodedata
from datetime import UTC, datetime
from urllib.parse import quote
from zoneinfo import ZoneInfo

from us2kr.models import EarningsItem, EconEvent, MarketPayload, NewsItem, Quote, Report
from us2kr.symbols import GROUP_TITLES_KO, POINT_CHANGE_GROUPS, SYMBOL_GROUPS

KST = ZoneInfo("Asia/Seoul")
MAX_NEWS_ITEMS = 12
DASH = "—"

_TIMING_KO = {"amc": "장마감후", "bmo": "장전", "unknown": ""}


# ------------------------------------------------------------------ formatting helpers


def _kst(dt: datetime) -> datetime:
    if dt.tzinfo is None:
        dt = dt.replace(tzinfo=UTC)
    return dt.astimezone(KST)


def _fmt_num(v: float | None, decimals: int = 2) -> str:
    return DASH if v is None else f"{v:,.{decimals}f}"


def _fmt_pct(v: float | None) -> str:
    if v is None:
        return DASH
    if abs(v) < 0.005:  # avoid printing "-0.00%"
        v = 0.0
    return f"{v:+.2f}%"


def _fmt_bp(change_abs: float | None) -> str:
    return DASH if change_abs is None else f"{change_abs * 100:+.1f}bp"


def _fmt_money(v: float | None) -> str:
    if v is None:
        return DASH
    if abs(v) >= 1e9:
        return f"{v / 1e9:,.2f}B"
    if abs(v) >= 1e6:
        return f"{v / 1e6:,.1f}M"
    return f"{v:,.2f}"


def _cell(text: str) -> str:
    """Escape a Markdown table cell (pipes, newlines)."""
    return text.replace("|", "\\|").replace("\n", " ").strip()


def _text(s: object) -> str:
    """Normalise a prose string from the Report: strip, collapse whitespace/newlines."""
    return re.sub(r"\s*\n\s*", " ", str(s or "")).strip()


def _link_text(s: str) -> str:
    return _text(s).replace("[", "(").replace("]", ")")


# ------------------------------------------------------------------ Markdown sections


def _header(payload: MarketPayload, report: Report | None) -> list[str]:
    lines = [
        f"# 🇺🇸→🇰🇷 미국 마감 브리핑 — {payload.krx_date.isoformat()} (미 {payload.session_date.isoformat()} 세션)"
    ]
    close_kst = _kst(payload.us_close_time).strftime("%Y-%m-%d %H:%M")
    gen_kst = _kst(payload.generated_at).strftime("%Y-%m-%d %H:%M")
    lines.append(f"미국 마감: {close_kst} KST · 생성: {gen_kst} KST")
    if not payload.krx_open_today:
        lines.append("⚠️ 오늘 한국 휴장")
    if report is None:
        lines.append("🧩 축소 모드(LLM 요약 없음)")
    return lines


def _bullets(items: list[str]) -> list[str]:
    return [f"- {_text(it)}" for it in items if _text(it)]


def _section(title: str, body: list[str]) -> list[str]:
    if not body:
        return []
    return [f"## {title}", *body, ""]


def _report_sections(report: Report) -> list[str]:
    out: list[str] = []
    if _text(report.headline):
        out += _section("총평", [f"**{_text(report.headline)}**"])
    if _text(report.market_overview):
        out += _section("시장 개요", [_text(report.market_overview)])
    out += _section("주요 요인", _bullets(report.key_drivers))
    if _text(report.rates_fx_commodities):
        out += _section("금리·환율·원자재", [_text(report.rates_fx_commodities)])
    out += _section("한국 장 시사점", _bullets(report.korea_implications))
    out += _section("오늘 볼 것", _bullets(report.watchlist_today))
    out += _section("리스크", _bullets(report.risk_flags))
    return out


def _ordered_groups(payload: MarketPayload) -> list[tuple[str, str, list[Quote]]]:
    """(group key, Korean title, quotes) in GROUP_TITLES_KO order; unknown groups appended."""
    by_group = payload.quotes_by_group()
    ordered: list[tuple[str, str, list[Quote]]] = []
    for g, title in GROUP_TITLES_KO.items():
        if g in by_group:
            ordered.append((g, title, _order_quotes(g, by_group[g])))
    for g, qs in by_group.items():
        if g not in GROUP_TITLES_KO:
            ordered.append((g, g, qs))
    return ordered


def _order_quotes(group: str, quotes: list[Quote]) -> list[Quote]:
    order = list(SYMBOL_GROUPS.get(group, {}))
    rank = {t: i for i, t in enumerate(order)}
    return sorted(quotes, key=lambda q: (rank.get(q.symbol, len(rank)), q.symbol))


def _quote_table(group: str, quotes: list[Quote]) -> list[str]:
    points = group in POINT_CHANGE_GROUPS
    has_after = any(q.after_hours_price is not None for q in quotes)
    header = ["종목", "종가", "변동"] + (["애프터"] if has_after else [])
    lines = ["| " + " | ".join(header) + " |", "|" + "|".join("---" for _ in header) + "|"]
    for q in quotes:
        name = _cell(q.name or q.symbol)
        if points:
            close = _fmt_num(q.close, 3)
            change = _fmt_bp(q.change_abs)
        else:
            close = _fmt_num(q.close, 2)
            change = _fmt_pct(q.change_pct)
        cells = [name, close, change]
        if has_after:
            if q.after_hours_price is None:
                cells.append(DASH)
            else:
                ah = _fmt_num(q.after_hours_price, 3 if points else 2)
                if q.after_hours_change_pct is not None:
                    ah += f" ({_fmt_pct(q.after_hours_change_pct)})"
                cells.append(ah)
        lines.append("| " + " | ".join(cells) + " |")
    return lines


# Normal URL punctuation and existing %XX escapes stay as-is; '(' ')' and whitespace become %28 %29 %20 so
# they cannot terminate the Markdown ``(url)`` destination early (see _LINK_RE).
_URL_SAFE = ":/?#[]@!$&'*+,;=%"


def _link_url(url: str) -> str:
    """Percent-encode characters that would break a Markdown / Telegram link destination."""
    return quote(url.strip(), safe=_URL_SAFE)


def _news_lines(news: list[NewsItem]) -> list[str]:
    lines = []
    for n in news[:MAX_NEWS_ITEMS]:
        when = _kst(n.published).strftime("%H:%M")
        lines.append(f"- [{_link_text(n.title)}]({_link_url(n.link)}) — {_text(n.source)}, {when} KST")
    return lines


def _event_line(e: EconEvent) -> str:
    star = "★ " if e.importance == "high" else ""
    when = f" {_kst(e.time).strftime('%m/%d %H:%M')} KST" if e.time else ""
    if e.kind == "released":
        return (
            f"- [발표{when}] {star}{_text(e.name)}: 실제 {e.actual or DASH} / 예상 {e.forecast or DASH}"
            f" / 이전 {e.previous or DASH}"
        )
    tail = f" (예상 {e.forecast})" if e.forecast else ""
    return f"- [예정{when}] {star}{_text(e.name)}{tail}"


def _earnings_line(it: EarningsItem) -> str:
    timing = _TIMING_KO.get(it.timing, "")
    tag = f"실적·{timing}" if timing else "실적"
    parts = []
    if it.eps_actual is not None or it.eps_estimate is not None:
        eps = f"EPS {_fmt_num(it.eps_actual)}"
        if it.eps_estimate is not None:
            eps += f" (예상 {_fmt_num(it.eps_estimate)})"
        parts.append(eps)
    if it.revenue_actual is not None or it.revenue_estimate is not None:
        rev = f"매출 {_fmt_money(it.revenue_actual)}"
        if it.revenue_estimate is not None:
            rev += f" (예상 {_fmt_money(it.revenue_estimate)})"
        parts.append(rev)
    detail = " · ".join(parts) if parts else "발표 예정"
    return f"- [{tag}] {_text(it.name)}({it.symbol}) {it.report_date.isoformat()}: {detail}"


def render_markdown(payload: MarketPayload, report: Report | None, model_used: str | None) -> str:
    """Render the Korean Markdown briefing. Works with ``report=None`` (degraded: tables only)."""
    lines: list[str] = _header(payload, report)
    lines.append("")
    if report is not None:
        lines += _report_sections(report)
    for g, title, qs in _ordered_groups(payload):
        lines += _section(f"📊 {title}", _quote_table(g, qs))
    lines += _section("히스토리 노트", _bullets(payload.history_notes))
    lines += _section("주요 뉴스", _news_lines(payload.news))
    ev_lines = [_event_line(e) for e in payload.events] + [_earnings_line(e) for e in payload.earnings]
    lines += _section("경제지표·실적", ev_lines)
    lines += _section("수집 누락", _bullets(payload.data_gaps))
    return "\n".join(lines).rstrip() + "\n"


# ------------------------------------------------------------------ Telegram HTML

_LINK_RE = re.compile(r"\[([^\]]+)\]\(([^)\s]+)\)")
_BOLD_RE = re.compile(r"\*\*(.+?)\*\*")
_HEADER_RE = re.compile(r"^(#{1,6})\s+(.*)$")
_BULLET_RE = re.compile(r"^\s*[-*]\s+")
_HR_RE = re.compile(r"^\s*(-{3,}|\*{3,}|_{3,})\s*$")
_SEP_CELL_RE = re.compile(r":?-{2,}:?")
_PRE_OPEN, _PRE_CLOSE = "<pre>", "</pre>"

_Unit = tuple[str, list[str]]  # ("text", [html_line]) | ("pre", [escaped_lines])


def _esc(s: str) -> str:
    return s.replace("&", "&amp;").replace("<", "&lt;").replace(">", "&gt;")


def _inline(md: str) -> str:
    """Escape HTML and convert inline Markdown (links, bold) to Telegram HTML."""
    out: list[str] = []
    pos = 0
    for m in _LINK_RE.finditer(md):
        out.append(_esc(md[pos : m.start()]))
        url = _esc(m.group(2)).replace('"', "&quot;")
        out.append(f'<a href="{url}">{_esc(m.group(1))}</a>')
        pos = m.end()
    out.append(_esc(md[pos:]))
    return _BOLD_RE.sub(r"<b>\1</b>", "".join(out))


def _width(s: str) -> int:
    """Approximate display width (CJK / emoji count as 2 columns)."""
    w = 0
    for ch in s:
        if unicodedata.combining(ch):
            continue
        if unicodedata.east_asian_width(ch) in ("W", "F") or ord(ch) >= 0x1F000:
            w += 2
        else:
            w += 1
    return w


def _split_row(line: str) -> list[str]:
    inner = line.strip().removeprefix("|")
    if inner.endswith("|") and not inner.endswith("\\|"):
        inner = inner[:-1]
    return [c.strip().replace("\\|", "|") for c in re.split(r"(?<!\\)\|", inner)]


def _format_table(md_rows: list[str]) -> list[str]:
    rows: list[list[str]] = []
    for r in md_rows:
        cells = _split_row(r)
        if cells and all(_SEP_CELL_RE.fullmatch(c) for c in cells if c):
            continue
        rows.append(cells)
    if not rows:
        return []
    ncol = max(len(r) for r in rows)
    rows = [r + [""] * (ncol - len(r)) for r in rows]
    widths = [max(_width(r[i]) for r in rows) for i in range(ncol)]
    out = []
    for r in rows:
        cells = []
        for i, c in enumerate(r):
            pad = " " * (widths[i] - _width(c))
            cells.append(c + pad if i == 0 else pad + c)
        out.append(_esc("  ".join(cells).rstrip()))
    return out


def _to_units(lines: list[str]) -> list[_Unit]:
    units: list[_Unit] = []
    i = 0
    while i < len(lines):
        line = lines[i]
        if line.lstrip().startswith("|"):
            table = []
            while i < len(lines) and lines[i].lstrip().startswith("|"):
                table.append(lines[i])
                i += 1
            pre = _format_table(table)
            if pre:
                units.append(("pre", pre))
            continue
        i += 1
        if _HR_RE.match(line):
            continue
        if not line.strip():
            if units and units[-1] == ("text", [""]):
                continue  # collapse blank runs
            units.append(("text", [""]))
            continue
        m = _HEADER_RE.match(line)
        if m:
            units.append(("text", [f"<b>{_inline(m.group(2).strip())}</b>"]))
        elif _BULLET_RE.match(line):
            units.append(("text", ["• " + _inline(_BULLET_RE.sub("", line, count=1))]))
        else:
            units.append(("text", [_inline(line)]))
    while units and units[0] == ("text", [""]):
        units.pop(0)
    while units and units[-1] == ("text", [""]):
        units.pop()
    return units


def _split_sections(markdown: str) -> list[list[str]]:
    sections: list[list[str]] = []
    cur: list[str] = []
    for line in markdown.splitlines():
        if line.startswith("## ") and cur:
            sections.append(cur)
            cur = []
        cur.append(line)
    if cur:
        sections.append(cur)
    return sections


def _units_html(units: list[_Unit]) -> str:
    parts: list[str] = []
    for kind, lines in units:
        if kind == "pre":
            parts.append("\n".join([_PRE_OPEN, *lines, _PRE_CLOSE]))
        else:
            parts.extend(lines)
    return "\n".join(parts)


def _joined_len(lines: list[str]) -> int:
    return sum(len(x) for x in lines) + max(len(lines) - 1, 0)


def _hard_split(line: str, limit: int) -> list[str]:
    return [line[i : i + limit] for i in range(0, len(line), limit)] or [""]


def _pack_units(units: list[_Unit], limit: int) -> list[str]:
    """Pack one oversized section line by line; <pre> blocks are closed/reopened across chunks."""
    chunks: list[str] = []
    buf: list[str] = []

    def flush() -> None:
        while buf and not buf[-1]:
            buf.pop()
        if buf:
            chunks.append("\n".join(buf))
        buf.clear()

    for kind, lines in units:
        if kind == "text":
            for line in lines:
                if not line and not buf:
                    continue
                if _joined_len([*buf, line]) > limit:
                    flush()
                if len(line) > limit:
                    for piece in _hard_split(line, limit):
                        flush()
                        buf.append(piece)
                    continue
                buf.append(line)
            continue
        pre: list[str] = []
        for line in lines:
            if _joined_len([*buf, _PRE_OPEN, *pre, line, _PRE_CLOSE]) <= limit:
                pre.append(line)
                continue
            if pre:
                buf.extend([_PRE_OPEN, *pre, _PRE_CLOSE])
                pre = []
            flush()
            pre.append(line)  # may exceed limit only if a single line is longer than limit
        if pre:
            buf.extend([_PRE_OPEN, *pre, _PRE_CLOSE])
    flush()
    return chunks


def render_telegram_chunks(markdown: str, limit: int = 4000) -> list[str]:
    """Convert the briefing Markdown to Telegram HTML and split it into chunks of at most ``limit`` chars.

    Sections (``## `` headers) are packed greedily; an oversized section is split on lines, and
    ``<pre>`` blocks are closed and reopened so every chunk is balanced HTML.
    """
    chunks: list[str] = []
    cur = ""
    for section in _split_sections(markdown):
        units = _to_units(section)
        if not units:
            continue
        html = _units_html(units)
        candidate = html if not cur else f"{cur}\n\n{html}"
        if len(candidate) <= limit:
            cur = candidate
            continue
        if cur:
            chunks.append(cur)
            cur = ""
        if len(html) <= limit:
            cur = html
            continue
        parts = _pack_units(units, limit)
        if parts:
            chunks.extend(parts[:-1])
            cur = parts[-1]
    if cur:
        chunks.append(cur)
    return chunks