fancier options for PDF strategy

This commit is contained in:
xenofem 2024-03-12 15:50:12 -04:00
parent 35f0c1fffe
commit 18fbc7f8dc

View file

@ -382,7 +382,7 @@ class Collator:
def link_pdf(self, src):
with fitz.open(src) as pdf:
images = pdf_images(pdf, self.args.force_convert_pdf)
images = pdf_images(pdf, self.args.pdf_strategy)
if images is None:
print(f'Failed to enumerate page images in PDF, skipping {src}')
return None
@ -569,7 +569,7 @@ def display_sixel_page(page):
finally:
sixel_output_unref(output)
def pdf_images(pdf, force=False):
def pdf_images(pdf, strategy):
images_by_page = [(page.get_images(), is_single_image(page)) for page in pdf]
if all(len(images) == 1 and single for (images, single) in images_by_page):
return (extract_image(pdf, images[0][0]) for (images, _) in images_by_page)
@ -587,24 +587,29 @@ def pdf_images(pdf, force=False):
image_extractors.append(lambda p=pdf, x=xref: extract_image(p, x))
else:
print(f'\nPage {idx+1}: {len(page_images)} images, {len([img for img in page_images if img["xref"] == 0])} non-xref images, {len(relevant_blocks(page))} total relevant objects')
if force:
print(f'Converting page {idx+1}')
choice = 'c'
else:
shown = False
while True:
choice = input(f'[N]ope out of this PDF / [c]onvert page to image{"" if xref is None else " / e[x]tract embedded image without text"}{"" if shown else " / [s]how page before deciding"}? [N/c{"" if xref is None else "/x"}{"" if shown else "/s"}] ')
if not shown and choice != '' and choice[0].lower() == 's':
display_sixel_page(page)
shown = True
else:
break
if xref is not None and choice != '' and choice[0].lower() == 'x':
image_extractors.append(lambda p=pdf, x=xref: extract_image(p, x))
elif choice != '' and choice[0].lower() == 'c':
image_extractors.append(lambda p=page: { 'ext': 'png', 'image': p.get_pixmap(dpi=PDF_CONVERSION_DPI).tobytes('png') })
else:
return None
choice = strategy
while True:
if choice.lower().startswith('n'):
return None
if choice.lower().startswith('c'):
if choice == strategy:
print(f'Converting page {idx+1}')
image_extractors.append(lambda p=page: { 'ext': 'png', 'image': p.get_pixmap(dpi=PDF_CONVERSION_DPI).tobytes('png') })
break
if xref is not None and (choice.lower().startswith('x') or choice.lower() == 'extract'):
if choice == strategy:
print(f'Extracting image from page {idx+1} without text')
image_extractors.append(lambda p=pdf, x=xref: extract_image(p, x))
break
if choice.lower().startswith('d'):
if choice == strategy:
print(f'Dropping page {idx+1}')
break
if choice.lower().startswith('s'):
display_sixel_page(page)
choice = input(f'[N]ope out / [c]onvert page{"" if xref is None else " / e[x]tract image"} / [d]rop page / [s]how page? [n/c{"" if xref is None else "/x"}/d/s] ')
print(f'\x1b[2K\r{idx+1}/{pdf.page_count} pages processed...', end=('' if idx+1 < pdf.page_count else '\n'))
return (extractor() for extractor in image_extractors)
@ -1182,9 +1187,17 @@ parser_collate.add_argument(
help="only collate works that are explicitly specified",
)
parser_collate.add_argument(
'--force-convert-pdf',
action='store_true',
help="convert a PDF page to a 300dpi image if there isn't a single image we can extract directly",
'-p', '--pdf-strategy',
choices=[
'ask', '?',
'show-ask', 's',
'convert', 'c',
'extract', 'x',
'drop', 'd',
'nope', 'n'
],
default='show-ask',
help="how to handle PDF pages that aren't a single image with no text",
)
parser_collate.add_argument(
'expression',