microsoft/hve-core
Publicmirrored fromhttps://github.com/microsoft/hve-coreAvailable
.github/skills/experimental/powerpoint/scripts/extract_content.py
1101lines · modecode
| 1 | """Extract content from an existing PPTX into YAML content and style definitions. |
| 2 | |
| 3 | Usage:: |
| 4 | |
| 5 | python extract_content.py \ |
| 6 | --input existing-deck.pptx --output-dir content/ |
| 7 | |
| 8 | python extract_content.py \ |
| 9 | --input existing-deck.pptx --output-dir content/ \ |
| 10 | --slides 3,7,15 |
| 11 | """ |
| 12 | |
| 13 | import argparse |
| 14 | from collections import Counter |
| 15 | from pathlib import Path |
| 16 | |
| 17 | import yaml |
| 18 | from lxml import etree |
| 19 | from pptx import Presentation |
| 20 | from pptx.oxml.ns import qn |
| 21 | from pptx_charts import extract_chart |
| 22 | from pptx_colors import extract_color, hex_brightness |
| 23 | from pptx_fills import extract_effect_list, extract_fill, extract_line |
| 24 | from pptx_fonts import ( |
| 25 | extract_alignment, |
| 26 | extract_font_info, |
| 27 | extract_paragraph_font, |
| 28 | normalize_font_family, |
| 29 | ) |
| 30 | from pptx_shapes import AUTO_SHAPE_NAME_MAP, extract_rotation |
| 31 | from pptx_tables import extract_table |
| 32 | from pptx_text import ( |
| 33 | extract_bullet_properties, |
| 34 | extract_paragraph_properties, |
| 35 | extract_run_properties, |
| 36 | extract_text_frame_properties, |
| 37 | ) |
| 38 | from pptx_utils import emu_to_inches |
| 39 | |
| 40 | |
| 41 | def extract_connector(shape) -> dict: |
| 42 | """Extract a connector element definition.""" |
| 43 | elem = { |
| 44 | "type": "connector", |
| 45 | "begin_x": emu_to_inches(shape.begin_x), |
| 46 | "begin_y": emu_to_inches(shape.begin_y), |
| 47 | "end_x": emu_to_inches(shape.end_x), |
| 48 | "end_y": emu_to_inches(shape.end_y), |
| 49 | "name": shape.name, |
| 50 | } |
| 51 | line_props = extract_line(shape) |
| 52 | if line_props: |
| 53 | elem.update(line_props) |
| 54 | return elem |
| 55 | |
| 56 | |
| 57 | def _is_freeform(shape) -> bool: |
| 58 | """Check whether a shape is a freeform with custom geometry.""" |
| 59 | nsmap = {"a": "http://schemas.openxmlformats.org/drawingml/2006/main"} |
| 60 | return shape._element.find(".//a:custGeom", nsmap) is not None |
| 61 | |
| 62 | |
| 63 | def _is_background_image(shape, slide_w: float, slide_h: float) -> bool: |
| 64 | """Detect whether a PICTURE shape covers the full slide as a background. |
| 65 | |
| 66 | A shape qualifies if it covers at least 95% of slide dimensions. |
| 67 | """ |
| 68 | w = emu_to_inches(shape.width) |
| 69 | h = emu_to_inches(shape.height) |
| 70 | return (w >= slide_w * 0.95) and (h >= slide_h * 0.95) |
| 71 | |
| 72 | |
| 73 | def _save_image_blob(shape, output_dir: Path, slide_num: int, img_count: int) -> dict: |
| 74 | """Save an image shape's blob to disk and return a path dict.""" |
| 75 | try: |
| 76 | img = shape.image |
| 77 | except ValueError: |
| 78 | return {"path": "LINKED_IMAGE_NOT_EMBEDDED"} |
| 79 | |
| 80 | ext = img.content_type.split("/")[-1] |
| 81 | if ext == "jpeg": |
| 82 | ext = "jpg" |
| 83 | img_name = f"image-{img_count:02d}.{ext}" |
| 84 | img_path = output_dir / "images" / img_name |
| 85 | img_path.parent.mkdir(parents=True, exist_ok=True) |
| 86 | |
| 87 | with open(img_path, "wb") as f: |
| 88 | f.write(img.blob) |
| 89 | |
| 90 | return {"path": f"images/{img_name}"} |
| 91 | |
| 92 | |
| 93 | def extract_freeform(shape) -> dict: |
| 94 | """Extract a freeform shape with its path vertices.""" |
| 95 | elem = { |
| 96 | "type": "freeform", |
| 97 | "left": emu_to_inches(shape.left), |
| 98 | "top": emu_to_inches(shape.top), |
| 99 | "width": emu_to_inches(shape.width), |
| 100 | "height": emu_to_inches(shape.height), |
| 101 | "name": shape.name, |
| 102 | } |
| 103 | |
| 104 | rot = extract_rotation(shape) |
| 105 | if rot is not None: |
| 106 | elem["rotation"] = rot |
| 107 | |
| 108 | # Extract fill and line properties |
| 109 | try: |
| 110 | fill_result = extract_fill(shape.fill) |
| 111 | if fill_result is not None: |
| 112 | elem["fill"] = fill_result |
| 113 | except (AttributeError, TypeError): |
| 114 | pass |
| 115 | |
| 116 | line_props = extract_line(shape) |
| 117 | if line_props: |
| 118 | elem.update(line_props) |
| 119 | |
| 120 | # Extract path vertices from custGeom XML |
| 121 | nsmap = {"a": "http://schemas.openxmlformats.org/drawingml/2006/main"} |
| 122 | paths = [] |
| 123 | for path_el in shape._element.findall(".//a:custGeom/a:pathLst/a:path", nsmap): |
| 124 | commands = [] |
| 125 | for child in path_el: |
| 126 | tag = child.tag.split("}")[-1] if "}" in child.tag else child.tag |
| 127 | if tag == "moveTo": |
| 128 | pt = child.find("a:pt", nsmap) |
| 129 | if pt is not None: |
| 130 | commands.append( |
| 131 | { |
| 132 | "cmd": "moveTo", |
| 133 | "x": int(pt.get("x", 0)), |
| 134 | "y": int(pt.get("y", 0)), |
| 135 | } |
| 136 | ) |
| 137 | elif tag == "lnTo": |
| 138 | pt = child.find("a:pt", nsmap) |
| 139 | if pt is not None: |
| 140 | commands.append( |
| 141 | { |
| 142 | "cmd": "lineTo", |
| 143 | "x": int(pt.get("x", 0)), |
| 144 | "y": int(pt.get("y", 0)), |
| 145 | } |
| 146 | ) |
| 147 | elif tag == "cubicBezTo": |
| 148 | pts = child.findall("a:pt", nsmap) |
| 149 | commands.append( |
| 150 | { |
| 151 | "cmd": "cubicBezTo", |
| 152 | "pts": [ |
| 153 | {"x": int(p.get("x", 0)), "y": int(p.get("y", 0))} |
| 154 | for p in pts |
| 155 | ], |
| 156 | } |
| 157 | ) |
| 158 | elif tag == "close": |
| 159 | commands.append({"cmd": "close"}) |
| 160 | if commands: |
| 161 | paths.append(commands) |
| 162 | |
| 163 | if paths: |
| 164 | elem["paths"] = paths |
| 165 | |
| 166 | return elem |
| 167 | |
| 168 | |
| 169 | def extract_group(shape, slide_num: int, output_dir, img_count: int) -> dict: |
| 170 | """Extract a group shape and its nested child elements.""" |
| 171 | elem = { |
| 172 | "type": "group", |
| 173 | "left": emu_to_inches(shape.left), |
| 174 | "top": emu_to_inches(shape.top), |
| 175 | "width": emu_to_inches(shape.width), |
| 176 | "height": emu_to_inches(shape.height), |
| 177 | "name": shape.name, |
| 178 | "elements": [], |
| 179 | } |
| 180 | for child in shape.shapes: |
| 181 | child_elem = extract_child_shape(child, slide_num, output_dir, img_count) |
| 182 | if child_elem: |
| 183 | elem["elements"].append(child_elem) |
| 184 | return elem |
| 185 | |
| 186 | |
| 187 | def _extract_shape_by_type( |
| 188 | shape, slide_num: int, output_dir, img_count: int |
| 189 | ) -> dict | None: |
| 190 | """Dispatch extraction based on shape_type, table/chart, or freeform.""" |
| 191 | shape_type = shape.shape_type |
| 192 | |
| 193 | # Simple shape_type dispatch (these extractors need no extra context) |
| 194 | _SIMPLE_EXTRACTORS = { |
| 195 | 17: extract_textbox, # TEXT_BOX |
| 196 | 1: extract_shape, # AUTO_SHAPE |
| 197 | 9: extract_connector, # LINE / CONNECTOR |
| 198 | } |
| 199 | extractor = _SIMPLE_EXTRACTORS.get(shape_type) |
| 200 | if extractor: |
| 201 | return extractor(shape) |
| 202 | |
| 203 | if shape_type == 13: # PICTURE |
| 204 | return extract_image(shape, output_dir, slide_num, img_count) |
| 205 | if shape_type == 6: # GROUP |
| 206 | return extract_group(shape, slide_num, output_dir, img_count) |
| 207 | |
| 208 | # Table and chart detection via attribute check |
| 209 | if hasattr(shape, "has_table") and shape.has_table: |
| 210 | return extract_table(shape) |
| 211 | if hasattr(shape, "has_chart") and shape.has_chart: |
| 212 | return extract_chart(shape) |
| 213 | if _is_freeform(shape): |
| 214 | return extract_freeform(shape) |
| 215 | |
| 216 | return None |
| 217 | |
| 218 | |
| 219 | def extract_child_shape( |
| 220 | shape, slide_num: int, output_dir, img_count: int |
| 221 | ) -> dict | None: |
| 222 | """Extract a single child shape within a group.""" |
| 223 | result = _extract_shape_by_type(shape, slide_num, output_dir, img_count) |
| 224 | if result is not None: |
| 225 | return result |
| 226 | |
| 227 | # Fallback for unrecognized shape types |
| 228 | elem = { |
| 229 | "type": "shape", |
| 230 | "shape": "rectangle", |
| 231 | "left": emu_to_inches(shape.left), |
| 232 | "top": emu_to_inches(shape.top), |
| 233 | "width": emu_to_inches(shape.width), |
| 234 | "height": emu_to_inches(shape.height), |
| 235 | "name": shape.name, |
| 236 | } |
| 237 | if shape.shape_type is not None: |
| 238 | elem["_unrecognized_shape_type"] = int(shape.shape_type) |
| 239 | return elem |
| 240 | |
| 241 | |
| 242 | def _has_formatting_variation(runs: list) -> bool: |
| 243 | """Check if multiple runs have different formatting properties.""" |
| 244 | if len(runs) <= 1: |
| 245 | return False |
| 246 | fonts = {r.get("font") for r in runs if "font" in r} |
| 247 | sizes = {r.get("size") for r in runs if "size" in r} |
| 248 | colors = {r.get("color") for r in runs if "color" in r} |
| 249 | bolds = {r.get("bold", False) for r in runs} |
| 250 | italics = {r.get("italic", False) for r in runs} |
| 251 | underlines = {r.get("underline", False) for r in runs} |
| 252 | return ( |
| 253 | len(fonts) > 1 |
| 254 | or len(sizes) > 1 |
| 255 | or len(colors) > 1 |
| 256 | or len(bolds) > 1 |
| 257 | or len(italics) > 1 |
| 258 | or len(underlines) > 1 |
| 259 | ) |
| 260 | |
| 261 | |
| 262 | # Key-mapping for extraction: maps canonical keys to output YAML key names |
| 263 | _SHAPE_EXTRACT_KEYS = { |
| 264 | "font": "text_font", |
| 265 | "size": "text_size", |
| 266 | "color": "text_color", |
| 267 | "bold": "text_bold", |
| 268 | } |
| 269 | _TEXTBOX_EXTRACT_KEYS = { |
| 270 | "font": "font", |
| 271 | "size": "font_size", |
| 272 | "color": "font_color", |
| 273 | "bold": "font_bold", |
| 274 | } |
| 275 | |
| 276 | # Keys to promote from first paragraph to element level |
| 277 | _SHAPE_PROMOTE_KEYS = ( |
| 278 | "text_font", |
| 279 | "text_size", |
| 280 | "text_color", |
| 281 | "text_bold", |
| 282 | "italic", |
| 283 | "alignment", |
| 284 | "char_spacing", |
| 285 | ) |
| 286 | _TEXTBOX_PROMOTE_KEYS = ( |
| 287 | "font", |
| 288 | "font_size", |
| 289 | "font_color", |
| 290 | "font_bold", |
| 291 | "italic", |
| 292 | "alignment", |
| 293 | "char_spacing", |
| 294 | ) |
| 295 | |
| 296 | |
| 297 | def _extract_text_content(text_frame, keys: dict, promote_keys: tuple) -> dict: |
| 298 | """Extract text content from a text frame into an element dict fragment. |
| 299 | |
| 300 | Handles paragraph iteration, run extraction, rich-text detection, and |
| 301 | paragraph/element-level key promotion. |
| 302 | |
| 303 | Args: |
| 304 | text_frame: python-pptx TextFrame object. |
| 305 | keys: Key-mapping dict for font/size/color/bold output names. |
| 306 | promote_keys: Tuple of keys to promote from first paragraph to element level. |
| 307 | |
| 308 | Returns: |
| 309 | Dict with text, text frame properties, paragraph data, and promoted defaults. |
| 310 | """ |
| 311 | result = {} |
| 312 | text = text_frame.text.strip() |
| 313 | if not text: |
| 314 | return result |
| 315 | |
| 316 | result["text"] = text |
| 317 | |
| 318 | tf_props = extract_text_frame_properties(text_frame) |
| 319 | if tf_props: |
| 320 | result.update(tf_props) |
| 321 | |
| 322 | para_dicts = [] |
| 323 | for para in text_frame.paragraphs: |
| 324 | run_info = {} |
| 325 | para_runs = [] |
| 326 | for run in para.runs: |
| 327 | font_info = extract_font_info(run.font) |
| 328 | run_extra = extract_run_properties(run) |
| 329 | para_runs.append({"text": run.text, **font_info, **run_extra}) |
| 330 | if not run_info: |
| 331 | run_info = {**font_info, **run_extra} |
| 332 | |
| 333 | para_info = extract_paragraph_font(para) |
| 334 | para_spacing = extract_paragraph_properties(para) |
| 335 | bullet_props = extract_bullet_properties(para) |
| 336 | alignment = extract_alignment(para) |
| 337 | merged = {**para_info, **run_info} |
| 338 | |
| 339 | p_dict = {"text": para.text} |
| 340 | if "font" in merged: |
| 341 | p_dict[keys["font"]] = merged["font"] |
| 342 | if "size" in merged: |
| 343 | p_dict[keys["size"]] = merged["size"] |
| 344 | if "color" in merged: |
| 345 | p_dict[keys["color"]] = merged["color"] |
| 346 | if merged.get("bold"): |
| 347 | p_dict[keys["bold"]] = True |
| 348 | if merged.get("italic"): |
| 349 | p_dict["italic"] = True |
| 350 | if merged.get("underline"): |
| 351 | p_dict["underline"] = True |
| 352 | if merged.get("hyperlink"): |
| 353 | p_dict["hyperlink"] = merged["hyperlink"] |
| 354 | if "char_spacing" in merged: |
| 355 | p_dict["char_spacing"] = merged["char_spacing"] |
| 356 | if "effect" in merged: |
| 357 | p_dict["text_effect"] = merged["effect"] |
| 358 | if alignment: |
| 359 | p_dict["alignment"] = alignment |
| 360 | if para_spacing: |
| 361 | p_dict.update(para_spacing) |
| 362 | if bullet_props: |
| 363 | p_dict.update(bullet_props) |
| 364 | if _has_formatting_variation(para_runs): |
| 365 | p_dict["runs"] = para_runs |
| 366 | para_dicts.append(p_dict) |
| 367 | |
| 368 | non_empty = [p for p in para_dicts if p["text"].strip()] |
| 369 | any_has_runs = any("runs" in p for p in para_dicts) |
| 370 | if len(para_dicts) > 1 or any_has_runs: |
| 371 | result["paragraphs"] = para_dicts |
| 372 | if non_empty: |
| 373 | first = non_empty[0] |
| 374 | for key in promote_keys: |
| 375 | if key in first: |
| 376 | result[key] = first[key] |
| 377 | elif non_empty: |
| 378 | first = non_empty[0] |
| 379 | for key, val in first.items(): |
| 380 | if key != "text": |
| 381 | result[key] = val |
| 382 | |
| 383 | return result |
| 384 | |
| 385 | |
| 386 | def extract_shape(shape) -> dict: |
| 387 | """Extract a shape element definition.""" |
| 388 | elem = { |
| 389 | "type": "shape", |
| 390 | "shape": "rectangle", |
| 391 | "left": emu_to_inches(shape.left), |
| 392 | "top": emu_to_inches(shape.top), |
| 393 | "width": emu_to_inches(shape.width), |
| 394 | "height": emu_to_inches(shape.height), |
| 395 | "name": shape.name, |
| 396 | } |
| 397 | |
| 398 | rot = extract_rotation(shape) |
| 399 | if rot is not None: |
| 400 | elem["rotation"] = rot |
| 401 | |
| 402 | # Detect shape type from auto_shape_type enum |
| 403 | try: |
| 404 | elem["shape"] = AUTO_SHAPE_NAME_MAP.get(shape.auto_shape_type, "rectangle") |
| 405 | except (AttributeError, TypeError): |
| 406 | elem["shape"] = "rectangle" |
| 407 | |
| 408 | # Extract corner radius (adjustment values) for rounded rectangles |
| 409 | try: |
| 410 | if shape.adjustments and len(shape.adjustments) > 0: |
| 411 | elem["corner_radius"] = round(shape.adjustments[0], 5) |
| 412 | except (AttributeError, TypeError, IndexError): |
| 413 | pass |
| 414 | |
| 415 | # Extract fill |
| 416 | try: |
| 417 | fill_result = extract_fill(shape.fill) |
| 418 | if fill_result is not None: |
| 419 | elem["fill"] = fill_result |
| 420 | except (AttributeError, TypeError): |
| 421 | pass |
| 422 | |
| 423 | # Extract line properties |
| 424 | line_props = extract_line(shape) |
| 425 | if line_props: |
| 426 | elem.update(line_props) |
| 427 | |
| 428 | # Extract effect list (outer shadow) |
| 429 | effect = extract_effect_list(shape) |
| 430 | if effect: |
| 431 | elem["effect"] = effect |
| 432 | |
| 433 | # Extract text if present |
| 434 | if shape.has_text_frame: |
| 435 | text_data = _extract_text_content( |
| 436 | shape.text_frame, _SHAPE_EXTRACT_KEYS, _SHAPE_PROMOTE_KEYS |
| 437 | ) |
| 438 | elem.update(text_data) |
| 439 | |
| 440 | return elem |
| 441 | |
| 442 | |
| 443 | def extract_textbox(shape) -> dict: |
| 444 | """Extract a text box element definition.""" |
| 445 | elem = { |
| 446 | "type": "textbox", |
| 447 | "left": emu_to_inches(shape.left), |
| 448 | "top": emu_to_inches(shape.top), |
| 449 | "width": emu_to_inches(shape.width), |
| 450 | "height": emu_to_inches(shape.height), |
| 451 | "text": shape.text_frame.text.strip() if shape.has_text_frame else "", |
| 452 | "name": shape.name, |
| 453 | } |
| 454 | |
| 455 | rot = extract_rotation(shape) |
| 456 | if rot is not None: |
| 457 | elem["rotation"] = rot |
| 458 | |
| 459 | if shape.has_text_frame: |
| 460 | text_data = _extract_text_content( |
| 461 | shape.text_frame, _TEXTBOX_EXTRACT_KEYS, _TEXTBOX_PROMOTE_KEYS |
| 462 | ) |
| 463 | elem.update(text_data) |
| 464 | |
| 465 | return elem |
| 466 | |
| 467 | |
| 468 | def extract_image(shape, output_dir: Path, slide_num: int, img_count: int) -> dict: |
| 469 | """Extract an image element and save the image file.""" |
| 470 | try: |
| 471 | img = shape.image |
| 472 | except ValueError: |
| 473 | # Linked images have no embedded blob |
| 474 | elem = { |
| 475 | "type": "image", |
| 476 | "path": "LINKED_IMAGE_NOT_EMBEDDED", |
| 477 | "left": emu_to_inches(shape.left), |
| 478 | "top": emu_to_inches(shape.top), |
| 479 | "width": emu_to_inches(shape.width), |
| 480 | "height": emu_to_inches(shape.height), |
| 481 | "name": shape.name, |
| 482 | "_note": "Image was linked, not embedded in the PPTX", |
| 483 | } |
| 484 | rot = extract_rotation(shape) |
| 485 | if rot is not None: |
| 486 | elem["rotation"] = rot |
| 487 | return elem |
| 488 | |
| 489 | ext = img.content_type.split("/")[-1] |
| 490 | if ext == "jpeg": |
| 491 | ext = "jpg" |
| 492 | |
| 493 | img_name = f"image-{img_count:02d}.{ext}" |
| 494 | img_path = output_dir / "images" / img_name |
| 495 | img_path.parent.mkdir(parents=True, exist_ok=True) |
| 496 | |
| 497 | with open(img_path, "wb") as f: |
| 498 | f.write(img.blob) |
| 499 | |
| 500 | elem = { |
| 501 | "type": "image", |
| 502 | "path": f"images/{img_name}", |
| 503 | "left": emu_to_inches(shape.left), |
| 504 | "top": emu_to_inches(shape.top), |
| 505 | "width": emu_to_inches(shape.width), |
| 506 | "height": emu_to_inches(shape.height), |
| 507 | "name": shape.name, |
| 508 | } |
| 509 | rot = extract_rotation(shape) |
| 510 | if rot is not None: |
| 511 | elem["rotation"] = rot |
| 512 | |
| 513 | # Extract image crop from srcRect on blipFill |
| 514 | blipFill = shape._element.find(qn("p:blipFill")) |
| 515 | if blipFill is not None: |
| 516 | # Preserve blipFill attributes (rotWithShape, dpi, etc.) |
| 517 | blip_fill_attrs = {} |
| 518 | for attr_name in ("rotWithShape", "dpi"): |
| 519 | val = blipFill.get(attr_name) |
| 520 | if val is not None: |
| 521 | blip_fill_attrs[attr_name] = val |
| 522 | if blip_fill_attrs: |
| 523 | elem["blip_fill_attrs"] = blip_fill_attrs |
| 524 | |
| 525 | srcRect = blipFill.find(qn("a:srcRect")) |
| 526 | if srcRect is not None and srcRect.attrib: |
| 527 | crop = {} |
| 528 | for side in ("l", "t", "r", "b"): |
| 529 | val = srcRect.get(side) |
| 530 | if val is not None: |
| 531 | crop[side] = int(val) |
| 532 | if crop: |
| 533 | elem["crop"] = crop |
| 534 | |
| 535 | # Extract image opacity from alphaModFix on the blip element |
| 536 | blip = shape._element.find(".//" + qn("a:blip")) |
| 537 | if blip is not None: |
| 538 | amf = blip.find(qn("a:alphaModFix")) |
| 539 | if amf is not None: |
| 540 | amt = int(amf.get("amt", "100000")) |
| 541 | elem["opacity"] = round(amt / 1000, 1) |
| 542 | |
| 543 | return elem |
| 544 | |
| 545 | |
| 546 | def detect_global_style(prs) -> dict: |
| 547 | """Analyze the presentation to detect common styling patterns. |
| 548 | |
| 549 | Detects multiple theme zones (e.g., light and dark slides) by clustering |
| 550 | slides based on background brightness and dominant text colors. |
| 551 | """ |
| 552 | bg_colors = Counter() |
| 553 | text_colors = Counter() |
| 554 | accent_colors = Counter() |
| 555 | fill_colors = Counter() |
| 556 | font_names = Counter() |
| 557 | font_sizes = Counter() |
| 558 | |
| 559 | # Per-slide analysis for theme clustering |
| 560 | slide_profiles = [] |
| 561 | |
| 562 | slide_w = emu_to_inches(prs.slide_width) |
| 563 | slide_h = emu_to_inches(prs.slide_height) |
| 564 | |
| 565 | for slide_idx, slide in enumerate(prs.slides): |
| 566 | slide_num = slide_idx + 1 |
| 567 | slide_bg = None |
| 568 | slide_text_colors = Counter() |
| 569 | slide_fill_colors = Counter() |
| 570 | has_bg_image = False |
| 571 | |
| 572 | # Detect background colors |
| 573 | try: |
| 574 | fill_result = extract_fill(slide.background.fill) |
| 575 | if isinstance(fill_result, str): |
| 576 | bg_colors[fill_result] += 1 |
| 577 | slide_bg = fill_result |
| 578 | except (AttributeError, TypeError): |
| 579 | pass |
| 580 | |
| 581 | for i, shape in enumerate(slide.shapes): |
| 582 | # Detect full-slide background images |
| 583 | if ( |
| 584 | i == 0 |
| 585 | and shape.shape_type == 13 |
| 586 | and _is_background_image(shape, slide_w, slide_h) |
| 587 | ): |
| 588 | has_bg_image = True |
| 589 | continue |
| 590 | |
| 591 | # Collect fill colors |
| 592 | try: |
| 593 | fill_result = extract_fill(shape.fill) |
| 594 | if isinstance(fill_result, str): |
| 595 | h = emu_to_inches(shape.height) |
| 596 | if h < 0.1: |
| 597 | accent_colors[fill_result] += 1 |
| 598 | else: |
| 599 | fill_colors[fill_result] += 1 |
| 600 | slide_fill_colors[fill_result] += 1 |
| 601 | except (AttributeError, TypeError): |
| 602 | pass |
| 603 | |
| 604 | # Collect font information |
| 605 | if shape.has_text_frame: |
| 606 | for para in shape.text_frame.paragraphs: |
| 607 | for run in para.runs: |
| 608 | if run.font.name: |
| 609 | base_name = normalize_font_family(run.font.name) |
| 610 | font_names[base_name] += 1 |
| 611 | if run.font.size: |
| 612 | font_sizes[int(run.font.size.pt)] += 1 |
| 613 | try: |
| 614 | color = extract_color(run.font.color) |
| 615 | if isinstance(color, str) and color.startswith("#"): |
| 616 | text_colors[color] += 1 |
| 617 | slide_text_colors[color] += 1 |
| 618 | except (AttributeError, TypeError): |
| 619 | pass |
| 620 | |
| 621 | # Classify slide brightness |
| 622 | bg_brightness = _classify_slide_brightness( |
| 623 | slide_bg, slide_text_colors, has_bg_image |
| 624 | ) |
| 625 | slide_profiles.append( |
| 626 | { |
| 627 | "slide": slide_num, |
| 628 | "bg_color": slide_bg, |
| 629 | "bg_brightness": bg_brightness, |
| 630 | "has_bg_image": has_bg_image, |
| 631 | "text_colors": dict(slide_text_colors), |
| 632 | "fill_colors": dict(slide_fill_colors), |
| 633 | } |
| 634 | ) |
| 635 | |
| 636 | # Build global color map from frequency analysis |
| 637 | colors = _build_color_map(bg_colors, fill_colors, text_colors, accent_colors) |
| 638 | |
| 639 | # Detect themes by clustering slides into light/dark groups |
| 640 | themes = _cluster_themes(slide_profiles, text_colors, fill_colors, accent_colors) |
| 641 | |
| 642 | # Determine primary fonts |
| 643 | body_font = "Segoe UI" |
| 644 | code_font = "Cascadia Code" |
| 645 | for f, _count in font_names.most_common(): |
| 646 | if any(kw in f.lower() for kw in ("cascadia", "consolas", "mono", "courier")): |
| 647 | code_font = f |
| 648 | else: |
| 649 | body_font = f |
| 650 | break |
| 651 | |
| 652 | # Determine font sizes |
| 653 | heading_size = 28 |
| 654 | body_size = 16 |
| 655 | if font_sizes: |
| 656 | filtered = {s: c for s, c in font_sizes.items() if 8 < s < 60} |
| 657 | if filtered: |
| 658 | sorted_sizes = sorted(filtered.keys()) |
| 659 | body_size = sorted_sizes[len(sorted_sizes) // 2] |
| 660 | heading_size = sorted_sizes[int(len(sorted_sizes) * 0.85)] |
| 661 | |
| 662 | style = { |
| 663 | "dimensions": { |
| 664 | "width_inches": emu_to_inches(prs.slide_width), |
| 665 | "height_inches": emu_to_inches(prs.slide_height), |
| 666 | "format": "16:9", |
| 667 | }, |
| 668 | "defaults": { |
| 669 | "speaker_notes_required": True, |
| 670 | }, |
| 671 | "typography": { |
| 672 | "body_font": body_font, |
| 673 | "code_font": code_font, |
| 674 | "heading_size": heading_size, |
| 675 | "body_size": body_size, |
| 676 | }, |
| 677 | } |
| 678 | |
| 679 | if colors: |
| 680 | style["colors"] = colors |
| 681 | |
| 682 | if themes: |
| 683 | style["themes"] = themes |
| 684 | |
| 685 | # Extract presentation metadata |
| 686 | metadata = {} |
| 687 | props = prs.core_properties |
| 688 | for attr in ("title", "author", "subject", "keywords", "description", "category"): |
| 689 | val = getattr(props, attr, None) |
| 690 | if val: |
| 691 | metadata[attr] = val |
| 692 | if metadata: |
| 693 | style["metadata"] = metadata |
| 694 | |
| 695 | return style |
| 696 | |
| 697 | |
| 698 | def _classify_slide_brightness( |
| 699 | bg_color: str | None, text_colors: Counter, has_bg_image: bool |
| 700 | ) -> str: |
| 701 | """Classify a slide as 'light' or 'dark' based on background and text colors.""" |
| 702 | if has_bg_image and bg_color is None: |
| 703 | # Slides with background images and no solid bg — infer from text colors |
| 704 | dark_text = sum( |
| 705 | c for hex_c, c in text_colors.items() if hex_brightness(hex_c) < 100 |
| 706 | ) |
| 707 | light_text = sum( |
| 708 | c for hex_c, c in text_colors.items() if hex_brightness(hex_c) > 150 |
| 709 | ) |
| 710 | return "light" if dark_text >= light_text else "dark" |
| 711 | |
| 712 | if bg_color and isinstance(bg_color, str) and bg_color.startswith("#"): |
| 713 | return "light" if hex_brightness(bg_color) > 128 else "dark" |
| 714 | |
| 715 | # Default: infer from text colors |
| 716 | dark_text = sum( |
| 717 | c for hex_c, c in text_colors.items() if hex_brightness(hex_c) < 100 |
| 718 | ) |
| 719 | light_text = sum( |
| 720 | c for hex_c, c in text_colors.items() if hex_brightness(hex_c) > 150 |
| 721 | ) |
| 722 | if dark_text > light_text: |
| 723 | return "light" |
| 724 | if light_text > dark_text: |
| 725 | return "dark" |
| 726 | return "dark" |
| 727 | |
| 728 | |
| 729 | def _build_color_map( |
| 730 | bg_colors: Counter, |
| 731 | fill_colors: Counter, |
| 732 | text_colors: Counter, |
| 733 | accent_colors: Counter, |
| 734 | ) -> dict: |
| 735 | """Build the global color map from frequency analysis.""" |
| 736 | colors = {} |
| 737 | if bg_colors: |
| 738 | colors["bg_dark"] = bg_colors.most_common(1)[0][0] |
| 739 | if fill_colors: |
| 740 | colors["bg_card"] = fill_colors.most_common(1)[0][0] |
| 741 | |
| 742 | for color_hex, _count in text_colors.most_common(5): |
| 743 | brightness = hex_brightness(color_hex) |
| 744 | if brightness > 200 and "text_white" not in colors: |
| 745 | colors["text_white"] = color_hex |
| 746 | elif brightness < 80 and "text_dark" not in colors: |
| 747 | colors["text_dark"] = color_hex |
| 748 | elif 80 <= brightness <= 200 and "text_gray" not in colors: |
| 749 | colors["text_gray"] = color_hex |
| 750 | |
| 751 | accent_names = ["accent_blue", "accent_teal", "accent_green"] |
| 752 | for i, (color_hex, _count) in enumerate(accent_colors.most_common(3)): |
| 753 | if i < len(accent_names): |
| 754 | colors[accent_names[i]] = color_hex |
| 755 | |
| 756 | return colors |
| 757 | |
| 758 | |
| 759 | def _cluster_themes( |
| 760 | slide_profiles: list[dict], |
| 761 | text_colors: Counter, |
| 762 | fill_colors: Counter, |
| 763 | accent_colors: Counter, |
| 764 | ) -> list[dict]: |
| 765 | """Cluster slides into theme groups based on brightness classification.""" |
| 766 | light_slides = [p for p in slide_profiles if p["bg_brightness"] == "light"] |
| 767 | dark_slides = [p for p in slide_profiles if p["bg_brightness"] == "dark"] |
| 768 | |
| 769 | # Only produce themes when both light and dark groups exist |
| 770 | if not light_slides or not dark_slides: |
| 771 | return [] |
| 772 | |
| 773 | themes = [] |
| 774 | |
| 775 | # Light theme |
| 776 | light_text = Counter() |
| 777 | light_fills = Counter() |
| 778 | for p in light_slides: |
| 779 | light_text.update(p["text_colors"]) |
| 780 | light_fills.update(p["fill_colors"]) |
| 781 | |
| 782 | light_colors = {} |
| 783 | for color_hex, _count in light_text.most_common(5): |
| 784 | brightness = hex_brightness(color_hex) |
| 785 | if brightness < 80 and "text_primary" not in light_colors: |
| 786 | light_colors["text_primary"] = color_hex |
| 787 | elif 80 <= brightness <= 200 and "text_secondary" not in light_colors: |
| 788 | light_colors["text_secondary"] = color_hex |
| 789 | if light_fills: |
| 790 | light_colors["bg_card"] = light_fills.most_common(1)[0][0] |
| 791 | |
| 792 | themes.append( |
| 793 | { |
| 794 | "name": "light", |
| 795 | "slides": sorted(p["slide"] for p in light_slides), |
| 796 | "colors": light_colors, |
| 797 | } |
| 798 | ) |
| 799 | |
| 800 | # Dark theme |
| 801 | dark_text = Counter() |
| 802 | dark_fills = Counter() |
| 803 | dark_bgs = Counter() |
| 804 | for p in dark_slides: |
| 805 | dark_text.update(p["text_colors"]) |
| 806 | dark_fills.update(p["fill_colors"]) |
| 807 | if p["bg_color"]: |
| 808 | dark_bgs[p["bg_color"]] += 1 |
| 809 | |
| 810 | dark_colors = {} |
| 811 | if dark_bgs: |
| 812 | dark_colors["bg_dark"] = dark_bgs.most_common(1)[0][0] |
| 813 | for color_hex, _count in dark_text.most_common(5): |
| 814 | brightness = hex_brightness(color_hex) |
| 815 | if brightness > 200 and "text_primary" not in dark_colors: |
| 816 | dark_colors["text_primary"] = color_hex |
| 817 | elif 80 <= brightness <= 200 and "text_secondary" not in dark_colors: |
| 818 | dark_colors["text_secondary"] = color_hex |
| 819 | if dark_fills: |
| 820 | dark_colors["bg_card"] = dark_fills.most_common(1)[0][0] |
| 821 | |
| 822 | themes.append( |
| 823 | { |
| 824 | "name": "dark", |
| 825 | "slides": sorted(p["slide"] for p in dark_slides), |
| 826 | "colors": dark_colors, |
| 827 | } |
| 828 | ) |
| 829 | |
| 830 | return themes |
| 831 | |
| 832 | |
| 833 | def extract_slide( |
| 834 | slide, |
| 835 | slide_num: int, |
| 836 | output_dir: Path, |
| 837 | slide_dims: tuple[float, float] | None = None, |
| 838 | ) -> dict: |
| 839 | """Extract all elements from a slide into a content.yaml structure.""" |
| 840 | slide_dir = output_dir / f"slide-{slide_num:03d}" |
| 841 | slide_dir.mkdir(parents=True, exist_ok=True) |
| 842 | |
| 843 | content = { |
| 844 | "slide": slide_num, |
| 845 | "title": "", |
| 846 | "elements": [], |
| 847 | } |
| 848 | |
| 849 | # Extract layout name |
| 850 | try: |
| 851 | layout_name = slide.slide_layout.name |
| 852 | if layout_name: |
| 853 | content["layout"] = layout_name |
| 854 | except (AttributeError, TypeError): |
| 855 | pass |
| 856 | |
| 857 | # Extract slide background |
| 858 | try: |
| 859 | if not slide.follow_master_background: |
| 860 | fill_result = extract_fill(slide.background.fill) |
| 861 | if fill_result is not None: |
| 862 | content["background"] = {"fill": fill_result} |
| 863 | except (AttributeError, TypeError): |
| 864 | pass |
| 865 | |
| 866 | # Extract speaker notes (include empty string when notes slide exists) |
| 867 | try: |
| 868 | if slide.has_notes_slide: |
| 869 | notes = slide.notes_slide.notes_text_frame.text.strip() |
| 870 | content["speaker_notes"] = notes |
| 871 | except (AttributeError, TypeError): |
| 872 | pass |
| 873 | |
| 874 | img_count = 0 |
| 875 | |
| 876 | for z_index, shape in enumerate(list(slide.shapes)): |
| 877 | shape_type = shape.shape_type |
| 878 | |
| 879 | # Track image count for filename generation |
| 880 | if shape_type == 13: |
| 881 | img_count += 1 |
| 882 | |
| 883 | # Handle placeholders specially (extract as textbox with marker) |
| 884 | if shape_type == 14: |
| 885 | if not shape.has_text_frame: |
| 886 | continue |
| 887 | elem = extract_textbox(shape) |
| 888 | elem["_placeholder"] = True |
| 889 | elem["z_order"] = z_index |
| 890 | content["elements"].append(elem) |
| 891 | continue |
| 892 | |
| 893 | # Use shared dispatcher for all other shape types |
| 894 | elem = _extract_shape_by_type(shape, slide_num, slide_dir, img_count) |
| 895 | if elem is not None: |
| 896 | elem["z_order"] = z_index |
| 897 | content["elements"].append(elem) |
| 898 | |
| 899 | # Detect title from textbox near top of slide |
| 900 | if ( |
| 901 | shape_type == 17 |
| 902 | and not content["title"] |
| 903 | and emu_to_inches(shape.top) < 1.5 |
| 904 | ): |
| 905 | text = shape.text_frame.text.strip() if shape.has_text_frame else "" |
| 906 | if text and len(text) < 100: |
| 907 | content["title"] = text |
| 908 | continue |
| 909 | |
| 910 | # Fallback for unrecognized shape types |
| 911 | elem_data = { |
| 912 | "type": "shape", |
| 913 | "shape": "rectangle", |
| 914 | "left": emu_to_inches(shape.left), |
| 915 | "top": emu_to_inches(shape.top), |
| 916 | "width": emu_to_inches(shape.width), |
| 917 | "height": emu_to_inches(shape.height), |
| 918 | "name": shape.name, |
| 919 | "z_order": z_index, |
| 920 | } |
| 921 | if shape_type is not None: |
| 922 | elem_data["_unrecognized_shape_type"] = int(shape_type) |
| 923 | content["elements"].append(elem_data) |
| 924 | |
| 925 | return content, slide_dir |
| 926 | |
| 927 | |
| 928 | def _resolve_theme_colors(prs) -> dict: |
| 929 | """Extract theme color name→hex mappings from the presentation's theme XML. |
| 930 | |
| 931 | Reads clrScheme from the slide master's theme and maps theme names |
| 932 | (background_1, text_1, accent_1, etc.) to their actual hex values. |
| 933 | """ |
| 934 | color_map = {} |
| 935 | scheme_names = { |
| 936 | "dk1": "dark_1", |
| 937 | "dk2": "dark_2", |
| 938 | "lt1": "light_1", |
| 939 | "lt2": "light_2", |
| 940 | "accent1": "accent_1", |
| 941 | "accent2": "accent_2", |
| 942 | "accent3": "accent_3", |
| 943 | "accent4": "accent_4", |
| 944 | "accent5": "accent_5", |
| 945 | "accent6": "accent_6", |
| 946 | "hlink": "hyperlink", |
| 947 | "folHlink": "followed_hyperlink", |
| 948 | } |
| 949 | # Map canonical aliases |
| 950 | aliases = { |
| 951 | "dark_1": "text_1", |
| 952 | "dark_2": "text_2", |
| 953 | "light_1": "background_1", |
| 954 | "light_2": "background_2", |
| 955 | } |
| 956 | try: |
| 957 | ns_a = "http://schemas.openxmlformats.org/drawingml/2006/main" |
| 958 | master = prs.slide_masters[0] |
| 959 | theme_el = None |
| 960 | # Theme is stored as a related part (generic Part, not XmlPart), |
| 961 | # so parse its blob directly with lxml. |
| 962 | for rel in master.part.rels.values(): |
| 963 | if "theme" in rel.reltype: |
| 964 | theme_el = etree.fromstring(rel.target_part.blob) |
| 965 | break |
| 966 | |
| 967 | if theme_el is not None: |
| 968 | clr_scheme = theme_el.find(f".//{{{ns_a}}}clrScheme") |
| 969 | if clr_scheme is not None: |
| 970 | for child in clr_scheme: |
| 971 | tag = child.tag.split("}")[-1] if "}" in child.tag else child.tag |
| 972 | theme_name = scheme_names.get(tag) |
| 973 | if theme_name is None: |
| 974 | continue |
| 975 | # Extract hex value from srgbClr or sysClr |
| 976 | srgb = child.find(f"{{{ns_a}}}srgbClr") |
| 977 | if srgb is not None: |
| 978 | color_map[theme_name] = f"#{srgb.get('val', '000000')}" |
| 979 | else: |
| 980 | sys_clr = child.find(f"{{{ns_a}}}sysClr") |
| 981 | if sys_clr is not None: |
| 982 | color_map[theme_name] = ( |
| 983 | f"#{sys_clr.get('lastClr', '000000')}" |
| 984 | ) |
| 985 | # Add alias mappings |
| 986 | if theme_name in aliases: |
| 987 | alias = aliases[theme_name] |
| 988 | if theme_name in color_map: |
| 989 | color_map[alias] = color_map[theme_name] |
| 990 | except (AttributeError, TypeError, IndexError): |
| 991 | pass |
| 992 | return color_map |
| 993 | |
| 994 | |
| 995 | def _resolve_theme_refs_in_content(content: dict, theme_colors: dict) -> dict: |
| 996 | """Replace @theme_name references with resolved hex values in content.""" |
| 997 | |
| 998 | def resolve_value(val): |
| 999 | if isinstance(val, str) and val.startswith("@"): |
| 1000 | theme_name = val[1:] |
| 1001 | return theme_colors.get(theme_name, val) |
| 1002 | if isinstance(val, dict): |
| 1003 | return {k: resolve_value(v) for k, v in val.items()} |
| 1004 | if isinstance(val, list): |
| 1005 | return [resolve_value(item) for item in val] |
| 1006 | return val |
| 1007 | |
| 1008 | return resolve_value(content) |
| 1009 | |
| 1010 | |
| 1011 | def main(): |
| 1012 | """CLI entry point for extracting PPTX content into YAML.""" |
| 1013 | parser = argparse.ArgumentParser( |
| 1014 | description="Extract content from a PPTX into YAML" |
| 1015 | ) |
| 1016 | parser.add_argument("--input", required=True, help="Input PPTX file path") |
| 1017 | parser.add_argument("--output-dir", required=True, help="Output content directory") |
| 1018 | parser.add_argument( |
| 1019 | "--slides", help="Comma-separated slide numbers to extract (default: all)" |
| 1020 | ) |
| 1021 | parser.add_argument( |
| 1022 | "--resolve-themes", |
| 1023 | action="store_true", |
| 1024 | help="Resolve @theme references to actual hex RGB values from the deck's theme", |
| 1025 | ) |
| 1026 | args = parser.parse_args() |
| 1027 | |
| 1028 | pptx_path = Path(args.input) |
| 1029 | output_dir = Path(args.output_dir) |
| 1030 | output_dir.mkdir(parents=True, exist_ok=True) |
| 1031 | |
| 1032 | slide_filter = None |
| 1033 | if args.slides: |
| 1034 | slide_filter = {int(s.strip()) for s in args.slides.split(",")} |
| 1035 | |
| 1036 | prs = Presentation(str(pptx_path)) |
| 1037 | print(f"Extracting from: {pptx_path}") |
| 1038 | print(f"Slides: {len(prs.slides)}") |
| 1039 | w = emu_to_inches(prs.slide_width) |
| 1040 | h = emu_to_inches(prs.slide_height) |
| 1041 | print(f'Dimensions: {w}" x {h}"') |
| 1042 | |
| 1043 | # Detect and save global style |
| 1044 | global_style = detect_global_style(prs) |
| 1045 | |
| 1046 | # Resolve theme colors when requested |
| 1047 | theme_colors = {} |
| 1048 | if args.resolve_themes: |
| 1049 | theme_colors = _resolve_theme_colors(prs) |
| 1050 | if theme_colors: |
| 1051 | global_style["theme_colors"] = theme_colors |
| 1052 | global_style = _resolve_theme_refs_in_content(global_style, theme_colors) |
| 1053 | print(f"Resolved {len(theme_colors)} theme colors") |
| 1054 | |
| 1055 | global_dir = output_dir / "global" |
| 1056 | global_dir.mkdir(parents=True, exist_ok=True) |
| 1057 | style_path = global_dir / "style.yaml" |
| 1058 | with open(style_path, "w", encoding="utf-8") as f: |
| 1059 | yaml.dump( |
| 1060 | global_style, |
| 1061 | f, |
| 1062 | default_flow_style=False, |
| 1063 | sort_keys=False, |
| 1064 | allow_unicode=True, |
| 1065 | ) |
| 1066 | print(f"Global style saved to {style_path}") |
| 1067 | |
| 1068 | # Extract slides (filtered or all) |
| 1069 | slide_dims = (emu_to_inches(prs.slide_width), emu_to_inches(prs.slide_height)) |
| 1070 | extracted = 0 |
| 1071 | for i, slide in enumerate(prs.slides): |
| 1072 | slide_num = i + 1 |
| 1073 | if slide_filter and slide_num not in slide_filter: |
| 1074 | continue |
| 1075 | content, slide_dir = extract_slide( |
| 1076 | slide, slide_num, output_dir, slide_dims=slide_dims |
| 1077 | ) |
| 1078 | |
| 1079 | # Resolve @theme references to hex values when --resolve-themes is set |
| 1080 | if args.resolve_themes and theme_colors: |
| 1081 | content = _resolve_theme_refs_in_content(content, theme_colors) |
| 1082 | |
| 1083 | content_path = slide_dir / "content.yaml" |
| 1084 | with open(content_path, "w", encoding="utf-8") as f: |
| 1085 | yaml.dump( |
| 1086 | content, |
| 1087 | f, |
| 1088 | default_flow_style=False, |
| 1089 | sort_keys=False, |
| 1090 | allow_unicode=True, |
| 1091 | ) |
| 1092 | print( |
| 1093 | f"Slide {slide_num}: {content.get('title', 'Untitled')} -> {content_path}" |
| 1094 | ) |
| 1095 | extracted += 1 |
| 1096 | |
| 1097 | print(f"\nExtraction complete. {extracted} slide(s) extracted to {output_dir}") |
| 1098 | |
| 1099 | |
| 1100 | if __name__ == "__main__": |
| 1101 | main() |
| 1102 | |