| 1 | #!/usr/bin/env python3 |
| 2 | """ |
| 3 | PPT Master - Source to Markdown Dispatcher |
| 4 | |
| 5 | Shared routing and backend command construction for source-to-Markdown tools. |
| 6 | |
| 7 | Usage: |
| 8 | Imported by scripts/source_to_md.py and scripts/project_manager.py |
| 9 | |
| 10 | Examples: |
| 11 | build_conversion_command("report.pdf", "report.md") |
| 12 | |
| 13 | Dependencies: |
| 14 | None |
| 15 | """ |
| 16 | |
| 17 | from __future__ import annotations |
| 18 | |
| 19 | import shutil |
| 20 | import sys |
| 21 | from dataclasses import dataclass |
| 22 | from pathlib import Path |
| 23 | from urllib.parse import urlparse |
| 24 | |
| 25 | |
| 26 | SOURCE_TO_MD_DIR = Path(__file__).resolve().parent |
| 27 | WECHAT_HOST_KEYWORDS = ("mp.weixin.qq.com", "weixin.qq.com") |
| 28 | |
| 29 | DOC_SUFFIXES = { |
| 30 | ".docx", ".doc", ".odt", ".rtf", # Office documents |
| 31 | ".epub", # eBooks |
| 32 | ".html", ".htm", # Web pages |
| 33 | ".tex", ".latex", ".rst", ".org", # Academic / technical |
| 34 | ".ipynb", ".typ", # Notebooks / Typst |
| 35 | } |
| 36 | EXCEL_SUFFIXES = {".xlsx", ".xlsm"} |
| 37 | LEGACY_EXCEL_SUFFIXES = {".xls"} |
| 38 | MARKDOWN_SUFFIXES = {".md", ".markdown"} |
| 39 | PDF_SUFFIXES = {".pdf"} |
| 40 | PRESENTATION_SUFFIXES = {".pptx", ".pptm", ".ppsx", ".ppsm", ".potx", ".potm"} |
| 41 | TEXT_SUFFIXES = {".txt", ".text"} |
| 42 | |
| 43 | |
| 44 | @dataclass |
| 45 | class ConversionCommand: |
| 46 | """Backend command plus metadata for one conversion route.""" |
| 47 | |
| 48 | command: list[str] |
| 49 | script_name: str |
| 50 | conversion_type: str |
| 51 | output_path: Path | None |
| 52 | |
| 53 | |
| 54 | def is_url(value: str) -> bool: |
| 55 | """Return whether a string looks like an HTTP(S) URL.""" |
| 56 | parsed = urlparse(value) |
| 57 | return parsed.scheme in {"http", "https"} and bool(parsed.netloc) |
| 58 | |
| 59 | |
| 60 | def detect_source_type(input_arg: str) -> str: |
| 61 | """Detect a conversion type from a URL, file, or directory.""" |
| 62 | if is_url(input_arg): |
| 63 | return "web" |
| 64 | |
| 65 | path = Path(input_arg) |
| 66 | if not path.exists(): |
| 67 | return "unknown" |
| 68 | if path.is_dir(): |
| 69 | return "directory" |
| 70 | |
| 71 | suffix = path.suffix.lower() |
| 72 | if suffix in PDF_SUFFIXES: |
| 73 | return "pdf" |
| 74 | if suffix in DOC_SUFFIXES: |
| 75 | return "doc" |
| 76 | if suffix in EXCEL_SUFFIXES or suffix in LEGACY_EXCEL_SUFFIXES: |
| 77 | return "excel" |
| 78 | if suffix in PRESENTATION_SUFFIXES: |
| 79 | return "pptx" |
| 80 | if suffix in MARKDOWN_SUFFIXES: |
| 81 | return "markdown" |
| 82 | if suffix in TEXT_SUFFIXES: |
| 83 | return "text" |
| 84 | return "unknown" |
| 85 | |
| 86 | |
| 87 | def default_markdown_path(input_arg: str) -> Path: |
| 88 | """Return the conventional Markdown output path for a local input.""" |
| 89 | path = Path(input_arg) |
| 90 | return path.parent / f"{path.stem}.md" |
| 91 | |
| 92 | |
| 93 | def _curl_cffi_available() -> bool: |
| 94 | """Return whether curl_cffi is importable.""" |
| 95 | try: |
| 96 | import curl_cffi # noqa: F401 |
| 97 | return True |
| 98 | except ImportError: |
| 99 | return False |
| 100 | |
| 101 | |
| 102 | def _web_script_command( |
| 103 | url: str, |
| 104 | output_path: Path | None, |
| 105 | python_executable: str, |
| 106 | allow_node_fallback: bool, |
| 107 | ) -> tuple[str, list[str]]: |
| 108 | """Return the web backend script name and base command.""" |
| 109 | host = urlparse(url).netloc.lower() |
| 110 | is_tls_sensitive = any(keyword in host for keyword in WECHAT_HOST_KEYWORDS) |
| 111 | node_backend = SOURCE_TO_MD_DIR / "web_to_md.cjs" |
| 112 | |
| 113 | if ( |
| 114 | allow_node_fallback |
| 115 | and is_tls_sensitive |
| 116 | and not _curl_cffi_available() |
| 117 | and node_backend.is_file() |
| 118 | and shutil.which("node") |
| 119 | ): |
| 120 | command = ["node", str(node_backend), url] |
| 121 | script_name = "web_to_md.cjs" |
| 122 | else: |
| 123 | command = [python_executable, str(SOURCE_TO_MD_DIR / "web_to_md.py"), url] |
| 124 | script_name = "web_to_md.py" |
| 125 | |
| 126 | if output_path is not None: |
| 127 | command.extend(["-o", str(output_path)]) |
| 128 | return script_name, command |
| 129 | |
| 130 | |
| 131 | def build_conversion_command( |
| 132 | input_arg: str, |
| 133 | output_path: str | Path | None, |
| 134 | *, |
| 135 | forced_type: str | None = None, |
| 136 | extra_args: list[str] | None = None, |
| 137 | pdf_image_mode: str | None = None, |
| 138 | render_vector_figures: bool = False, |
| 139 | python_executable: str | None = None, |
| 140 | allow_node_web_fallback: bool = True, |
| 141 | ) -> ConversionCommand: |
| 142 | """Build the backend CLI command for one source-to-Markdown conversion.""" |
| 143 | conversion_type = forced_type or detect_source_type(input_arg) |
| 144 | output = Path(output_path) if output_path is not None else None |
| 145 | extra = extra_args or [] |
| 146 | python = python_executable or sys.executable |
| 147 | |
| 148 | if conversion_type == "web": |
| 149 | if not is_url(input_arg): |
| 150 | raise ValueError("web conversion requires an http:// or https:// URL") |
| 151 | script_name, command = _web_script_command( |
| 152 | input_arg, |
| 153 | output, |
| 154 | python, |
| 155 | allow_node_web_fallback, |
| 156 | ) |
| 157 | command.extend(extra) |
| 158 | return ConversionCommand(command, script_name, conversion_type, output) |
| 159 | |
| 160 | if conversion_type in {"markdown", "text", "directory", "unknown"}: |
| 161 | raise ValueError(f"conversion type {conversion_type!r} has no backend command") |
| 162 | |
| 163 | script_by_type = { |
| 164 | "pdf": "pdf_to_md.py", |
| 165 | "doc": "doc_to_md.py", |
| 166 | "excel": "excel_to_md.py", |
| 167 | "pptx": "ppt_to_md.py", |
| 168 | } |
| 169 | script_name = script_by_type.get(conversion_type) |
| 170 | if script_name is None: |
| 171 | raise ValueError(f"unsupported conversion type: {conversion_type}") |
| 172 | if output is None: |
| 173 | output = default_markdown_path(input_arg) |
| 174 | |
| 175 | command = [ |
| 176 | python, |
| 177 | str(SOURCE_TO_MD_DIR / script_name), |
| 178 | input_arg, |
| 179 | "-o", |
| 180 | str(output), |
| 181 | ] |
| 182 | if conversion_type == "pdf" and pdf_image_mode: |
| 183 | command.extend(["--images", pdf_image_mode]) |
| 184 | if conversion_type == "pdf" and render_vector_figures: |
| 185 | command.append("--render-vector-figures") |
| 186 | command.extend(extra) |
| 187 | return ConversionCommand(command, script_name, conversion_type, output) |
| 188 |