support more corner cases in SAREF4LIFT and SAREF4WATR (bd022523) · Commits · SAREF / saref-pypeline

src/saref_pypeline/docgen/ts2md_extractor.py

+82 −32

Original line number	Diff line number	Diff line
		@@ -53,6 +53,8 @@ from docx.table import _Cell, Table
		from docx.text.paragraph import Paragraph
		from docx.text.hyperlink import Hyperlink
		from docx.shape import InlineShape
		from docx.oxml.ns import nsmap


		from saref_pypeline._logging import TRACE_LEVEL
		from saref_pypeline.docgen.utils import OWL_GRAPH, print_admonition, with_flags, pprint_xml
		@@ -74,6 +76,7 @@ if TYPE_CHECKING:
		from saref_pypeline.docgen import SiteManager

		logger = logging.getLogger(__name__)
		nsmap["v"] = "urn:schemas-microsoft-com:vml"


		class ExtractFormat(Enum):
		@@ -234,6 +237,41 @@ def process_substitutions_md(md):

		return md

		def parse_dml_crop(el) -> dict \| None:
		"""Parse DrawingML cropping (<pic:blipFill><a:srcRect>)."""
		crop_attrs = el.xpath(".//pic:blipFill/a:srcRect")
		if not crop_attrs:
		return None
		attrs = crop_attrs[0].attrib
		crop = {}
		for key in ("l", "r", "t", "b"):
		if key in attrs:
		# DrawingML: 1/1000 %
		crop[key] = int(attrs[key]) / 100000.0
		return crop or None


		def parse_vml_crop(vml) -> dict \| None:
		"""Parse VML cropping (<v:imagedata croptop=... cropleft=...>).
		TODO: non urgent, VML cropping seems to shrink image vertically.
		"""
		crop = {}
		mapping = {
		"croptop": "t",
		"cropleft": "l",
		"cropright": "r",
		"cropbottom": "b",
		}
		for attr, key in mapping.items():
		if attr in vml.attrib:
		raw = vml.attrib[attr]
		if raw.endswith("f"):
		# VML "f" units: 1/65536
		val = int(raw[:-1]) / 65536.0
		else:
		val = float(raw)
		crop[key] = val
		return crop or None

		def cell_align_style(cell):
		"""
		@@ -575,7 +613,8 @@ class TS2MDExtractor:
		try:
		style = P_STYLE(paragraph.style.name)
		except:
		style = P_STYLE.Normal
		logger.warning(f"Unknown style {paragraph.style.name}")
		style = paragraph.style

		fname = f"extract_{style.name}"
		if hasattr(self, fname) and callable(getattr(self, fname)):
		@@ -896,6 +935,10 @@ class TS2MDExtractor:
		self, paragraph: Paragraph, extract_format: ExtractFormat = ExtractFormat.MD
		):
		"""Table title"""
		if paragraph.text.startswith("Figure"):
		logger.warning(f"Wrong style for {paragraph.text}")
		return self.extract_TF(paragraph, extract_format)
		else:
		return ""

		def extract_TAH(
		@@ -993,44 +1036,41 @@ class TS2MDExtractor:
		return None, 0, None
		if isinstance(paragraph, Paragraph):
		for run in paragraph.runs:
		# An image inserted via add_picture() is stored as an inline shape
		if not run._element.xpath(".//pic:pic"):
		continue
		el = run._element

		try:
		inline_shape: InlineShape = run._element.xpath(".//a:blip/@r:embed")
		if not inline_shape:
		continue

		# Get the relationship id
		# --- CASE 1: DrawingML ---
		inline_shape = el.xpath(".//a:blip/@r:embed")
		if inline_shape:
		rId = inline_shape[0]

		# Retrieve the image part from the relationship
		image_part = paragraph.part.related_parts[rId]

		# Width from the drawing element (in EMU)
		cx_attr = run._element.xpath(".//a:xfrm/a:ext/@cx")
		# width in EMU
		cx_attr = el.xpath(".//a:xfrm/a:ext/@cx")
		width_emu = float(cx_attr[0]) if cx_attr else None

		# Original filename (as stored in the package, not necessarily original filesystem name)
		filename = image_part.partname # e.g. '/word/media/image1.png'
		filename = image_part.partname
		crop = parse_dml_crop(el)
		return filename, width_emu, crop

		# Cropping info (fractions of 1.0)
		crop = {}
		crop_attrs = run._element.xpath(".//pic:blipFill/a:srcRect")
		if crop_attrs:
		attrs = crop_attrs[0].attrib
		for key in ("l", "r", "t", "b"):
		if key in attrs:
		# Convert from 1/1000 percent to fraction
		crop[key] = int(attrs[key]) / 100000.0
		else:
		crop = None
		# --- CASE 2: VML ---
		vml_shape = el.xpath(".//v:imagedata")
		if vml_shape:
		vml = vml_shape[0]
		rId = vml.attrib.get("{http://schemas.openxmlformats.org/officeDocument/2006/relationships}id")
		image_part = paragraph.part.related_parts[rId]

		# width in EMU (pt → EMU)
		style = el.xpath(".//v:shape/@style")
		width_emu = None
		if style:
		m = re.search(r"width:(\d+(?:\.\d+)?)pt", style[0])
		if m:
		pt = float(m.group(1))
		width_emu = int(pt * 12700)

		filename = image_part.partname
		crop = parse_vml_crop(vml)
		return filename, width_emu, crop
		except Exception as ex:
		print(str(ex))
		continue

		image_paragraph = get_prev_block(self.document, paragraph)
		return self.get_paragraph_image_info(image_paragraph, i + 1)
		@@ -1145,3 +1185,13 @@ class TS2MDExtractor:

		return level4" "+ "* " + self.extract_Normal(paragraph, extract_format).strip()


		def extract_Caption(
		self, paragraph: Paragraph, extract_format: ExtractFormat = ExtractFormat.MD
		):
		"""Table title"""
		if paragraph.text.startswith("Figure"):
		logger.warning(f"Wrong style for {paragraph.text}")
		return self.extract_TF(paragraph, extract_format)
		else:
		return ""