Skip to content

fix(convert): preserve PDF→PPTX line layout and font names #40

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
nelsonduarte merged 1 commit into from
May 7, 2026
Merged

fix(convert): preserve PDF→PPTX line layout and font names #40

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
125 changes: 71 additions & 54 deletions app/tools/convert.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -458,38 +458,35 @@ def do_work(worker):
return None
slide = prs.slides.add_slide(blank)

# Extract text blocks and image blocks separately so
# the result is editable in PowerPoint instead of
# one rasterised image per slide. Each text block
# becomes a textbox at its bbox; each line in the
# block becomes a paragraph; each span becomes a
# run with its own size/bold/italic/color.
# Editable extraction: image blocks become picture
# shapes; each *line* of a text block becomes its
# own textbox at the line's bbox so PowerPoint
# doesn't re-flow lines (one textbox per block was
# wrong — multi-line blocks lost their original
# vertical positions when PowerPoint re-laid out
# the text). Spans within the same line stay as
# runs in a single paragraph so inline formatting
# (bold/italic mid-sentence) is preserved.
blocks = page.get_text("dict").get("blocks", [])
for block in blocks:
bbox = block.get("bbox")
if not bbox:
continue
x0, y0, x1, y1 = bbox
# Clamp to slide bounds; fitz can emit blocks
# whose bbox slightly overflows the page rect
# for italic/decorated glyphs.
x0 = max(0, x0); y0 = max(0, y0)
x1 = min(slide_w_pt, x1); y1 = min(slide_h_pt, y1)
w = x1 - x0; h = y1 - y0
if w <= 0 or h <= 0:
continue

if block.get("type") == 1: # image block
bbox = block.get("bbox")
img_data = block.get("image")
if not img_data:
if not (bbox and img_data):
continue
ix0, iy0, ix1, iy1 = bbox
ix0 = max(0, ix0); iy0 = max(0, iy0)
ix1 = min(slide_w_pt, ix1); iy1 = min(slide_h_pt, iy1)
iw = ix1 - ix0; ih = iy1 - iy0
if iw <= 0 or ih <= 0:
continue
try:
slide.shapes.add_picture(
io.BytesIO(img_data),
Emu(int(x0 * 12700)),
Emu(int(y0 * 12700)),
Emu(int(w * 12700)),
Emu(int(h * 12700)))
Emu(int(ix0 * 12700)),
Emu(int(iy0 * 12700)),
Emu(int(iw * 12700)),
Emu(int(ih * 12700)))
except Exception:
# Image format not supported by python-pptx
# (rare formats like JBIG2). Skip the
Expand All @@ -498,40 +495,49 @@ def do_work(worker):
pass
continue

lines = block.get("lines", [])
if not lines:
continue
try:
tb = slide.shapes.add_textbox(
Emu(int(x0 * 12700)),
Emu(int(y0 * 12700)),
Emu(int(w * 12700)),
Emu(int(h * 12700)))
except Exception:
continue
tf = tb.text_frame
tf.word_wrap = True
# Zero internal padding so the textbox bbox
# matches the PDF span bbox more faithfully.
# Older python-pptx versions reject Emu(0) for
# margins; tolerate that and leave the default.
for attr in ("margin_left", "margin_right",
"margin_top", "margin_bottom"):
try: setattr(tf, attr, 0)
except Exception: pass # noqa: S110

first_line = True
for line in lines:
spans = line.get("spans", [])
for line in block.get("lines", []):
spans = [s for s in line.get("spans", [])
if s.get("text", "")]
if not spans:
continue
para = tf.paragraphs[0] if first_line else tf.add_paragraph()
first_line = False
lbb = line.get("bbox")
if not lbb:
continue
lx0, ly0, lx1, ly1 = lbb
# Pad horizontally so PowerPoint's slightly
# different glyph metrics don't clip the
# last character; vertically use the line
# bbox as-is to keep baselines aligned.
lx0 = max(0, lx0 - 1)
ly0 = max(0, ly0)
lx1 = min(slide_w_pt, lx1 + 4)
ly1 = min(slide_h_pt, ly1)
lw = lx1 - lx0; lh = ly1 - ly0
if lw <= 0 or lh <= 0:
continue
try:
tb = slide.shapes.add_textbox(
Emu(int(lx0 * 12700)),
Emu(int(ly0 * 12700)),
Emu(int(lw * 12700)),
Emu(int(lh * 12700)))
except Exception:
continue
tf = tb.text_frame
tf.word_wrap = False # one line — no wrap
# Zero internal padding so the textbox bbox
# matches the PDF line bbox more faithfully.
# Older python-pptx versions reject Emu(0)
# for margins; tolerate that.
for attr in ("margin_left", "margin_right",
"margin_top", "margin_bottom"):
try: setattr(tf, attr, 0)
except Exception: pass # noqa: S110

para = tf.paragraphs[0]
first_run = True
for span in spans:
text = span.get("text", "")
if not text:
continue
text = span["text"]
# Reuse the auto-created empty run for
# the first span; add new runs after.
if first_run and len(para.runs) > 0:
Expand All @@ -549,6 +555,17 @@ def do_work(worker):
flags = span.get("flags", 0)
run.font.bold = bool(flags & 16)
run.font.italic = bool(flags & 2)
# PDF subset fonts come prefixed with
# 6 random caps + "+", strip them so
# PowerPoint can substitute the
# canonical face: e.g. "ABCDEF+Arial"
# → "Arial".
font_name = span.get("font", "") or ""
if len(font_name) > 7 and font_name[6] == "+":
font_name = font_name[7:]
if font_name:
try: run.font.name = font_name
except Exception: pass # noqa: S110
color = span.get("color", 0)
if color:
try:
Expand Down