Repository navigation
PPTX: a hyperlink on a whole shape (p:cNvPr/a:hlinkClick) is dropped #4782
Description
Bug
A hyperlink attached to a whole shape, written as a:hlinkClick under the shape's p:cNvPr (python-pptx: shape.click_action.hyperlink.address), does not reach the document: the shape's text item has hyperlink=None.
#4442 covers the run-level form (a:rPr/a:hlinkClick) and PR #4443 fixes that form only; this is the other place PowerPoint stores a link, and it is lost on 2.138.0 and on main.
Steps to reproduce
"""Reproduce: a PPTX hyperlink on a whole shape (p:cNvPr/a:hlinkClick) is lost; so is one on text runs (#4442)."""
import sys
from io import BytesIO
import docling
import pptx
from docling.datamodel.base_models import InputFormat
from docling.document_converter import DocumentConverter
from docling_core.types.doc import ContentLayer, TextItem
from docling_core.types.io import DocumentStream
from lxml import etree
from pptx import Presentation
from pptx.util import Inches
def build() -> BytesIO:
prs = Presentation()
slide = prs.slides.add_slide(prs.slide_layouts[6])
paragraph = slide.shapes.add_textbox(Inches(1), Inches(1), Inches(6), Inches(1)).text_frame.paragraphs[0]
paragraph.add_run().text = "Read "
for word in ("the", " story"):
run = paragraph.add_run()
run.text = word
run.hyperlink.address = "https://example.com/story"
paragraph.add_run().text = " tonight."
button = slide.shapes.add_textbox(Inches(1), Inches(3), Inches(4), Inches(1))
button.text_frame.text = "Watch the talk"
button.click_action.hyperlink.address = "https://example.com/talk"
print(etree.tostring(button._element.nvSpPr.cNvPr, pretty_print=True).decode().strip())
buf = BytesIO()
prs.save(buf)
buf.seek(0)
return buf
result = DocumentConverter(allowed_formats=[InputFormat.PPTX]).convert(
DocumentStream(name="repro.pptx", stream=build())
)
print(f"docling {docling.__version__}, python-pptx {pptx.__version__}, python {sys.version.split()[0]}")
for item, _ in result.document.iterate_items(included_content_layers=set(ContentLayer)):
if isinstance(item, TextItem):
print(f"{item.label.value}: {item.text!r} hyperlink={item.hyperlink!r}")Output (the cNvPr print shows the link is in the file):
<p:cNvPr ... id="3" name="TextBox 2">
<a:hlinkClick r:id="rId3"/>
</p:cNvPr>
docling 2.138.0, python-pptx 1.0.2, python 3.14.8
paragraph: 'Read the story tonight.' hyperlink=None
paragraph: 'Watch the talk' hyperlink=None
(The first paragraph is the run-level loss of #4442; the second is this issue.)
Expected
The shape's text item carries the shape's link as its hyperlink, the way the DOCX backend carries a hyperlink run. Where a run inside the shape has its own link, the run's link wins for that run.
Docling version
2.138.0
Python version
3.14.8
Activity