ext/TradingAgents-astock
2
0
Fork 0
mirror of https://github.com/simonlin1212/TradingAgents-astock.git synced 2026-10-30 09:29:56 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix(web): stabilize Chinese PDF export (#56)

Browse Source 
Prefer PDF-compatible CJK fonts, reuse single-face fonts for bold fallback, and render Chinese text with compact spacing and character wrapping.

改进中文 PDF 导出：优先使用更稳的 CJK 字体，单字体 fallback 复用粗体，并使用紧凑空白与字符级换行改善排版。

Co-authored-by: Codex <codex@example.com>
This commit is contained in:
waterblue
2026-06-20 15:01:02 +08:00
committed by GitHub
co-authored by Codex
parent 13002631b9
commit 7f4431fdb1
3 changed files with 354 additions and 95 deletions
Download Patch File Download Diff File Expand all files Collapse all files
Dockerfile
+1 -1
View File
@@ -21,7 +21,7 @@ ENV PATH="/opt/venv/bin:$PATH"
# CJK fonts so PDF export (fpdf2) can render Chinese reports inside the container
# (issue #48 — the slim image ships no CJK font, so _find_cjk_font() returns None).
RUN apt-get update \
&& apt-get install -y --no-install-recommends fonts-noto-cjk \
&& apt-get install -y --no-install-recommends fonts-wqy-microhei fonts-noto-cjk \
&& rm -rf /var/lib/apt/lists/*
RUN useradd --create-home appuser
tests/test_pdf_export.py
+72
View File
@@ -0,0 +1,72 @@
"""Regression tests for PDF export."""
from pathlib import Path
import pytest
from fpdf.enums import WrapMode
from web.pdf_export import (
PDFExportError,
_ReportPDF,
_compact_inline_text,
_discover_cjk_fonts,
_find_cjk_fonts,
_format_table_cells,
generate_pdf,
)
def test_generate_pdf_with_chinese_markdown_when_cjk_font_available():
try:
_find_cjk_fonts()
except PDFExportError as exc:
pytest.skip(str(exc))
state = {
"market_report": "# 技术分析\n- 趋势: 偏强\n- 结论: 中文字体测试通过。",
"news_report": "| 项目 | 结论 |\n|---|---|\n| 中文 | 可嵌入 |",
"final_trade_decision": "最终建议: HOLD 观望。",
}
pdf_bytes = generate_pdf(state, "600519", "2026-05-30", "HOLD")
assert pdf_bytes.startswith(b"%PDF-")
assert len(pdf_bytes) > 1000
def test_pdf_text_helpers_compact_alignment_whitespace():
assert _compact_inline_text("Neutral Analyst: 我会把") == "Neutral Analyst: 我会把"
assert _format_table_cells(["表面看是放量反弹", "但对连续跌停后的", "ST"]) == "表面看是放量反弹 | 但对连续跌停后的 | ST"
def test_pdf_multicell_defaults_to_left_alignment():
pdf = object.__new__(_ReportPDF)
calls = []
pdf.l_margin = 10
pdf.r_margin = 10
pdf.w = 210
pdf.set_x = lambda x: None
pdf.multi_cell = lambda *args, **kwargs: calls.append((args, kwargs))
pdf._write_multicell(5.5, "近 5 日均量明显高于近 20 日均量。")
assert calls[0][1]["align"] == "L"
assert calls[0][1]["wrapmode"] == WrapMode.CHAR
def test_wqy_discovery_reuses_same_font_for_bold(monkeypatch, tmp_path):
wqy = tmp_path / "wqy-microhei.ttc"
noto_bold = tmp_path / "NotoSansCJK-Bold.ttc"
wqy.write_bytes(b"font")
noto_bold.write_bytes(b"font")
def fake_find_font_file(pattern: str) -> Path | None:
return {
"wqy-microhei.ttc": wqy,
"NotoSansCJK-Bold.ttc": noto_bold,
}.get(pattern)
monkeypatch.setattr("web.pdf_export._find_font_file", fake_find_font_file)
assert _discover_cjk_fonts() == (wqy, wqy)
web/pdf_export.py
+284 -97
View File
@@ -2,14 +2,16 @@
from __future__ import annotations
import platform
import os
import re
import subprocess
from datetime import datetime
from pathlib import Path
from typing import Any
import fpdf as _fpdf_mod
from fpdf import FPDF
from fpdf.enums import WrapMode
# fpdf2 (maintained fork) and the abandoned pyfpdf 1.x BOTH import as `fpdf`, and
@@ -20,6 +22,112 @@ from fpdf import FPDF
# fix it, instead of letting the PDF blow up mid-render.
_FPDF_VERSION = getattr(_fpdf_mod, "__version__", None) or getattr(_fpdf_mod, "FPDF_VERSION", "0")
_PDF_FONT_ENV = "TRADINGAGENTS_PDF_FONT"
_PDF_BOLD_FONT_ENV = "TRADINGAGENTS_PDF_BOLD_FONT"
_FONT_CANDIDATES = [
(
"/usr/share/fonts/truetype/wqy/wqy-microhei.ttc",
"/usr/share/fonts/truetype/wqy/wqy-microhei.ttc",
),
(
"/usr/share/fonts/truetype/wqy/wqy-zenhei.ttc",
"/usr/share/fonts/truetype/wqy/wqy-zenhei.ttc",
),
(
"/usr/share/fonts/opentype/noto/NotoSansCJK-Regular.ttc",
"/usr/share/fonts/opentype/noto/NotoSansCJK-Bold.ttc",
),
(
"/usr/share/fonts/noto-cjk/NotoSansCJK-Regular.ttc",
"/usr/share/fonts/noto-cjk/NotoSansCJK-Bold.ttc",
),
(
"/usr/share/fonts/noto-cjk/NotoSansCJKsc-Regular.otf",
"/usr/share/fonts/noto-cjk/NotoSansCJKsc-Bold.otf",
),
(
"/usr/share/fonts/truetype/noto/NotoSansSC-Regular.ttf",
"/usr/share/fonts/truetype/noto/NotoSansSC-Bold.ttf",
),
(
"/usr/share/fonts/truetype/droid/DroidSansFallbackFull.ttf",
"/usr/share/fonts/truetype/droid/DroidSansFallbackFull.ttf",
),
(
"/System/Library/Fonts/PingFang.ttc",
"/System/Library/Fonts/PingFang.ttc",
),
(
"/System/Library/Fonts/STHeiti Light.ttc",
"/System/Library/Fonts/STHeiti Medium.ttc",
),
(
"C:/Windows/Fonts/msyh.ttc",
"C:/Windows/Fonts/msyhbd.ttc",
),
(
"C:/Windows/Fonts/simhei.ttf",
"C:/Windows/Fonts/simhei.ttf",
),
]
_FONT_NAME_CANDIDATES = [
("WenQuanYi Micro Hei", "Regular"),
("WenQuanYi Zen Hei", "Regular"),
("Noto Sans CJK SC", "Regular"),
("Noto Sans CJK SC", "Bold"),
("Noto Sans SC", "Regular"),
("Noto Sans SC", "Bold"),
("Source Han Sans SC", "Regular"),
("Source Han Sans SC", "Bold"),
]
_FONT_FILE_PATTERNS = (
"wqy-microhei.ttc",
"wqy-zenhei.ttc",
"NotoSansCJK-Regular.ttc",
"NotoSansCJK-Bold.ttc",
"NotoSansCJKsc-Regular.otf",
"NotoSansCJKsc-Bold.otf",
"NotoSansSC-Regular.ttf",
"NotoSansSC-Bold.ttf",
"SourceHanSansSC-Regular.otf",
"SourceHanSansSC-Bold.otf",
"DroidSansFallbackFull.ttf",
)
_CJK_FONT_MARKERS = (
"NotoSansCJK",
"NotoSansSC",
"NotoSerifCJK",
"SourceHanSans",
"SourceHanSerif",
"wqy-",
"DroidSansFallback",
"PingFang",
"STHeiti",
"msyh",
"simhei",
)
_TTC_SC_FACE_INDEXES = {
"NotoSansCJK-Regular.ttc": 2,
"NotoSansCJK-Bold.ttc": 2,
"NotoSerifCJK-Regular.ttc": 2,
"NotoSerifCJK-Bold.ttc": 2,
}
_SINGLE_FACE_BOLD_FALLBACKS = {
"wqy-microhei.ttc",
"wqy-zenhei.ttc",
"DroidSansFallbackFull.ttf",
}
class PDFExportError(RuntimeError):
"""Raised when the PDF report cannot be exported."""
def _ensure_fpdf2() -> None:
try:
@@ -36,88 +144,138 @@ def _ensure_fpdf2() -> None:
)
# Per-OS CJK font candidates. The current OS's fonts are tried first so a
# user on Windows/Linux/macOS all get a working PDF without manual config.
_WIN_FONTS = [
"C:/Windows/Fonts/msyh.ttc", # 微软雅黑
"C:/Windows/Fonts/msyhbd.ttc", # 微软雅黑 Bold
"C:/Windows/Fonts/simhei.ttf", # 黑体
"C:/Windows/Fonts/simsun.ttc", # 宋体
"C:/Windows/Fonts/simfang.ttf", # 仿宋
]
_MAC_FONTS = [
"/System/Library/Fonts/PingFang.ttc",
"/System/Library/Fonts/STHeiti Light.ttc",
"/System/Library/Fonts/Hiragino Sans GB.ttc",
"/Library/Fonts/Arial Unicode.ttf",
]
_LINUX_FONTS = [
"/usr/share/fonts/truetype/noto/NotoSansSC-Regular.ttf",
"/usr/share/fonts/noto-cjk/NotoSansCJKsc-Regular.otf",
"/usr/share/fonts/opentype/noto/NotoSansCJK-Regular.ttc",
"/usr/share/fonts/opentype/noto/NotoSansCJK-Regular.otf",
"/usr/share/fonts/truetype/wqy/wqy-zenhei.ttc",
"/usr/share/fonts/truetype/wqy/wqy-microhei.ttc",
"/usr/share/fonts/truetype/arphic/uming.ttc",
]
# Substrings that reliably indicate a CJK-capable font during the recursive
# fallback scan (deliberately excludes bare "noto", which also matches the
# Latin-only Noto family).
_CJK_FONT_KEYWORDS = (
"msyh", "simhei", "simsun", "simfang", "yahei", "fangsong",
"pingfang", "heiti", "stheiti", "stsong", "songti", "kaiti",
"hiragino sans gb", "arial unicode",
"notosanscjk", "notoserifcjk", "notosanssc", "notoserifsc",
"sourcehansans", "sourcehanserif", "wqy", "uming", "ukai",
)
def _font_missing_message() -> str:
candidates = ", ".join(regular for regular, _ in _FONT_CANDIDATES[:4])
return (
"PDF 导出需要可嵌入的 Unicode 中文字体。请优先安装 fonts-wqy-microhei，"
f"或设置 {_PDF_FONT_ENV}=/path/to/wqy-microhei.ttc。"
f"已检查的常见路径包括: {candidates}"
)
def _font_candidates() -> list[str]:
"""Return CJK font paths ordered with the current OS's fonts first."""
system = platform.system()
if system == "Windows":
return _WIN_FONTS + _MAC_FONTS + _LINUX_FONTS
if system == "Darwin":
return _MAC_FONTS + _WIN_FONTS + _LINUX_FONTS
return _LINUX_FONTS + _MAC_FONTS + _WIN_FONTS
def _env_font_path(env_name: str) -> Path | None:
configured = os.getenv(env_name)
if not configured:
return None
def _search_dirs() -> list[str]:
home = Path.home()
system = platform.system()
if system == "Windows":
return ["C:/Windows/Fonts"]
if system == "Darwin":
return ["/System/Library/Fonts", "/Library/Fonts", str(home / "Library/Fonts")]
return ["/usr/share/fonts", "/usr/local/share/fonts", str(home / ".fonts")]
def _find_cjk_font() -> str | None:
"""Locate a CJK-capable TTF/TTC/OTF font, cross-platform.
1. Try the known per-OS candidate paths.
2. Fall back to recursively scanning common font directories for any
file whose name looks CJK-capable.
"""
for path in _font_candidates():
if Path(path).exists():
path = Path(configured).expanduser()
if not path.exists():
raise PDFExportError(f"{env_name} 指向的字体文件不存在: {path}")
return path
for directory in _search_dirs():
dpath = Path(directory)
if not dpath.exists():
continue
for ext in ("*.ttc", "*.ttf", "*.otf"):
try:
for font_path in sorted(dpath.rglob(ext)):
if any(k in font_path.name.lower() for k in _CJK_FONT_KEYWORDS):
return str(font_path)
except OSError:
def _is_likely_cjk_font(path: Path) -> bool:
return any(marker in path.name for marker in _CJK_FONT_MARKERS)
def _font_search_roots() -> list[Path]:
roots = [
Path("/usr/share/fonts"),
Path("/usr/local/share/fonts"),
Path("~/.local/share/fonts").expanduser(),
Path("~/.fonts").expanduser(),
]
xdg_data_home = os.getenv("XDG_DATA_HOME")
if xdg_data_home:
roots.append(Path(xdg_data_home).expanduser() / "fonts")
return roots
def _find_font_file(pattern: str) -> Path | None:
for root in _font_search_roots():
if not root.exists():
continue
matches = sorted(root.rglob(pattern))
if matches:
return matches[0]
return None
def _font_from_fontconfig(family: str, style: str) -> Path | None:
try:
output = subprocess.check_output(
["fc-match", "-f", "%{file}", f"{family}:style={style}"],
stderr=subprocess.DEVNULL,
text=True,
timeout=2,
).strip()
except (OSError, subprocess.SubprocessError):
return None
if not output:
return None
path = Path(output)
if path.exists() and _is_likely_cjk_font(path):
return path
return None
def _discover_cjk_fonts() -> tuple[Path, Path] | None:
discovered: dict[str, Path] = {}
for pattern in _FONT_FILE_PATTERNS:
path = _find_font_file(pattern)
if path:
discovered[pattern] = path
regular = (
discovered.get("wqy-microhei.ttc")
or discovered.get("wqy-zenhei.ttc")
or discovered.get("NotoSansCJK-Regular.ttc")
or discovered.get("NotoSansCJKsc-Regular.otf")
or discovered.get("NotoSansSC-Regular.ttf")
or discovered.get("SourceHanSansSC-Regular.otf")
or discovered.get("DroidSansFallbackFull.ttf")
)
if regular and regular.name in _SINGLE_FACE_BOLD_FALLBACKS:
return regular, regular
bold = (
discovered.get("NotoSansCJK-Bold.ttc")
or discovered.get("NotoSansCJKsc-Bold.otf")
or discovered.get("NotoSansSC-Bold.ttf")
or discovered.get("SourceHanSansSC-Bold.otf")
or regular
)
if regular and bold:
return regular, bold
for family, style in _FONT_NAME_CANDIDATES:
font_path = _font_from_fontconfig(family, style)
if not font_path:
continue
if style == "Bold" and regular:
return regular, font_path
if style != "Bold":
return font_path, font_path
return None
def _find_cjk_fonts() -> tuple[Path, Path]:
env_regular = _env_font_path(_PDF_FONT_ENV)
if env_regular:
return env_regular, _env_font_path(_PDF_BOLD_FONT_ENV) or env_regular
for regular_path, bold_path in _FONT_CANDIDATES:
regular = Path(regular_path)
if regular.exists():
bold = Path(bold_path)
return regular, bold if bold.exists() else regular
discovered = _discover_cjk_fonts()
if discovered:
return discovered
raise PDFExportError(_font_missing_message())
def _collection_font_number(path: Path) -> int:
"""Select the Simplified Chinese face from known CJK font collections."""
return _TTC_SC_FACE_INDEXES.get(path.name, 0)
def _strip_think(text: str) -> str:
return re.sub(r"<think>.*?</think>\s*", "", text, flags=re.DOTALL).strip()
@@ -131,6 +289,17 @@ def _strip_md_inline(text: str) -> str:
return text
def _compact_inline_text(text: str) -> str:
"""Collapse table/alignment whitespace that would create wide PDF gaps."""
return re.sub(r"[ \t\u3000]{2,}", " ", text).strip()
def _format_table_cells(cells: list[str]) -> str:
cleaned = [_compact_inline_text(_strip_md_inline(cell)) for cell in cells]
cleaned = [cell for cell in cleaned if cell]
return " | ".join(cleaned)
def _signal_color(signal: str) -> tuple[int, int, int]:
s = signal.upper()
if "BUY" in s:
@@ -157,19 +326,39 @@ class _ReportPDF(FPDF):
self.ticker = ticker
self.trade_date = trade_date
self.signal = signal
font_path = _find_cjk_font()
if not font_path:
raise RuntimeError(
"未找到可用的中文字体，无法生成 PDF。请安装一款中文字体后重试"
"（Windows 自带微软雅黑/黑体，macOS 自带苹方，Linux 可 "
"`apt install fonts-noto-cjk`），或改用「下载 Markdown」导出。"
regular_font, bold_font = _find_cjk_fonts()
try:
self.add_font(
"CJK",
"",
str(regular_font),
collection_font_number=_collection_font_number(regular_font),
)
self.add_font("CJK", "", font_path)
self.add_font("CJK", "B", font_path)
self.add_font(
"CJK",
"B",
str(bold_font),
collection_font_number=_collection_font_number(bold_font),
)
except Exception as exc:
raise PDFExportError(
f"无法加载 PDF 中文字体: {regular_font}。"
f"请换一个 TTF/OTF/TTC 字体并通过 {_PDF_FONT_ENV} 指定。"
) from exc
def _use_font(self, style: str = "", size: int = 10) -> None:
self.set_font("CJK", style, size)
def _text_block_width(self) -> float:
return self.w - self.l_margin - self.r_margin
def _write_multicell(self, height: float, text: str, **kwargs: Any) -> None:
self.set_x(self.l_margin)
kwargs.setdefault("align", "L")
kwargs.setdefault("wrapmode", WrapMode.CHAR)
self.multi_cell(self._text_block_width(), height, text, **kwargs)
def header(self) -> None:
self._use_font("", 8)
self.set_text_color(150, 150, 150)
@@ -218,8 +407,8 @@ class _ReportPDF(FPDF):
self._use_font("", 9)
self.set_text_color(120, 120, 120)
self.multi_cell(
0, 5,
self._write_multicell(
5,
"免责声明: 本报告由 AI 多 Agent 系统自动生成, 仅供学习研究与技术演示, "
"不构成任何投资建议。投资决策请咨询持牌专业机构。"
"使用本报告所产生的任何损失由使用者自行承担。",
@@ -254,21 +443,21 @@ class _ReportPDF(FPDF):
if stripped.startswith("###"):
self._use_font("B", 11)
self.set_text_color(50, 50, 50)
self.cell(0, 7, stripped.lstrip("#").strip())
self.cell(0, 7, _compact_inline_text(stripped.lstrip("#")))
self.ln(8)
i += 1
continue
if stripped.startswith("##"):
self._use_font("B", 13)
self.set_text_color(40, 40, 40)
self.cell(0, 8, stripped.lstrip("#").strip())
self.cell(0, 8, _compact_inline_text(stripped.lstrip("#")))
self.ln(9)
i += 1
continue
if stripped.startswith("#"):
self._use_font("B", 14)
self.set_text_color(255, 90, 31)
self.cell(0, 9, stripped.lstrip("#").strip())
self.cell(0, 9, _compact_inline_text(stripped.lstrip("#")))
self.ln(10)
i += 1
continue
@@ -293,13 +482,13 @@ class _ReportPDF(FPDF):
m = re.match(r"^(\d+[.)])\s*(.*)", stripped)
bullet = f" {m.group(1)} "
body = m.group(2)
body = _strip_md_inline(body)
self.set_x(self.l_margin)
self.multi_cell(0, 5.5, bullet + body, wrapmode="CHAR")
body = _compact_inline_text(_strip_md_inline(body))
self._write_multicell(5.5, bullet + body)
i += 1
continue
# Table rows (|col|col|) → render as plain text with spacing
# Table rows (|col|col|) → render compactly; fixed-width spacing
# creates large visual gaps in proportional PDF fonts.
if stripped.startswith("|") and stripped.endswith("|"):
# Skip separator rows like |---|---|
if re.match(r"^\|[-:\s|]+\|$", stripped):
@@ -308,9 +497,8 @@ class _ReportPDF(FPDF):
self._use_font("", 9)
self.set_text_color(60, 60, 60)
cells = [c.strip() for c in stripped.strip("|").split("|")]
row_text = " ".join(_strip_md_inline(c) for c in cells)
self.set_x(self.l_margin)
self.multi_cell(0, 5, row_text, wrapmode="CHAR")
row_text = _format_table_cells(cells)
self._write_multicell(5, row_text)
i += 1
continue
@@ -327,9 +515,8 @@ class _ReportPDF(FPDF):
self._use_font("", 10)
self.set_text_color(40, 40, 40)
para = " ".join(para_lines)
para = _strip_md_inline(para)
self.set_x(self.l_margin)
self.multi_cell(0, 5.5, para, wrapmode="CHAR")
para = _compact_inline_text(_strip_md_inline(para))
self._write_multicell(5.5, para)
self.ln(2)
continue