cleanup-slides.py

#!/bin/python3
from PyPDF2 import PdfFileWriter, PdfFileReader
from Levenshtein import distance
import sys
import re

# Requires PyPDF2 and python-Levenshtein.	
# Install them with `pip install PyPDF2 python-Levenshtein

# Maximum distance between the same content located in
# two different slides (accounts for some lib/pdf inconsistencies)
MAX_DISTANCE = 10


infile, outfile = None, None

if len(sys.argv) != 3:
	print('usage: cleanup-slides [input.pdf] [output.pdf]')
	quit()

try:
	infile = open(sys.argv[1], 'rb')
	outfile = open(sys.argv[2], 'wb')
except:
	print("Failed to open specified files")
	quit()

srcpdf = PdfFileReader(infile)
dstpdf = PdfFileWriter()


def detectPageIndex(txt):
	index = 1
	while txt[-index:].isnumeric():
		index += 1
	if index == 1 or index > 6:
		return False
	return txt[-index+1:]

# Process all the pages in the pdf
for i in range(srcpdf.numPages):
	pageComplete = True
	
	curpg = srcpdf.getPage(i)
	curtxt = re.sub('[^a-zA-Z0-9]+', '', curpg.extractText())
	curidx = detectPageIndex(curtxt)

	# Unless this is not the last page, check if the page
	# is complete and not an intermediate slide.
	#
	# A page is complete when the next page contains
	# completly different text.
	# A page is intermediate when the next page contains
	# similar starting text to the current slide
	if i != srcpdf.numPages - 1:
		nextpg = srcpdf.getPage(i + 1)
		nexttxt = re.sub('[^a-zA-Z0-9]+', '', nextpg.extractText())
		nextidx = detectPageIndex(nexttxt)

		if (nextidx != False) and (curidx != False):
			pageComplete = curidx != nextidx
		else:
			# if the starting text of the two slides is close,
			# they are probably the same slide
			if distance(curtxt, nexttxt[:len(curtxt)]) < MAX_DISTANCE:
				pageComplete = False

	if pageComplete:
		dstpdf.addPage(curpg)

# Write filtered pdf
dstpdf.write(outfile)

infile.close()
outfile.close()