Dump and rebuild books from Calameo website

build_calameo_pdf.sh

#!/bin/bash
URL="https://p.calameoassets.com/200424144000-11e93cd5206820f83e75e58a9b1e4652/p%page_number%.svgz"


function get_images {
	# download svg files
	for i in {1..196}
	do
		echo "page $i..."
		VAR="${URL/\%page_number\%/$i}"
		# Or printf -v VAR "$URL" $i
		# with as placeholder "%s"
		wget --quiet --user-agent="Mozilla/5.0 (X11; Linux x86_64; rv:80.0) Gecko/20100101 Firefox/80.0" $VAR --directory-prefix="download/"

		sleep 0.5
	done
}

function merge_images {

	# convert svg files into pdfs
	for i in *.svgz
   	do
		inkscape $i --export-pdf $i.pdf
	done

	# merge pdfs (sort their numbers by natural order)
	pdfunite $(ls -v *.pdf) output.pdf
}

mkdir -p download
get_images
cd download
# Extract fonts
python3 extract_fonts.py
# Install fonts
mkdir -p ~/.fonts/custom/
cp *.woff ~/.fonts/custom/ && fc-cache
# Make book
merge_images
# Purge fonts
rm -rf ~/.fonts/custom/ && fc-cache

extract_fonts.py

#!/usr/bin/env python3
# coding: utf-8
# Standard imports
from lxml import etree
import re
import base64
import glob
from collections import defaultdict
# Custom imports
import cssutils


def write_font(content, title):
    """Dump binary font into woff font file"""
    with open(title + ".woff", "wb+") as f_d:
        f_d.write(base64.b64decode(content))


def extract_fonts(svg_file, font_db):
    """Extract fonts stored in base64 in the stylesheet of the given svg file"""

    tree = etree.parse(svg_file)
    root = tree.getroot()

    ns = {'s': 'http://www.w3.org/2000/svg'}

    # Get only the stylesheet (1 per document)
    nodes = root.xpath('//s:style', namespaces=ns)
    assert len(nodes) == 1
    style_node = nodes[0]

    #print(style_node.text)
    #print(style_node.tag)

    # Parse stylesheet
    sheet = cssutils.parseString(style_node.text)

    g = (rule for rule in sheet if isinstance(rule, cssutils.css.CSSFontFaceRule))

    for rule in g:
        #print(rule.style.cssText)
        #print(rule.style.src)
        font_family = rule.style.fontFamily.replace("\"", "")

        # Get the src attribute and extract font in base64
        for property in rule.style:
            #print(property.name)
            #print(property.value)

            if property.name != "src":
                continue

            # url(data:application/font-woff;charset=utf-8;base64,XXX) format("woff")
            m = re.search('base64(.*)"\) format\("woff"\)', property.value)
            if m:
                # base64 code found => dump it
                # PS: 1 url per CSS rule in practical...
                #print(m.group(1))

                write_font(m.group(1), font_family)

                # Keep the font and its name in memory (test)
                # font_db[m.group(1)].add(font_family)


def replace_private_chars(svgz_file):
    """Replace "" (not in fonts) by non-breaking character in the given svg file"""

    with open(svgz_file, "r+", encoding="utf8") as f_d:
        data = f_d.read().replace(u"\ue01f", u"\u00a0")
        #data = f_d.read().replace(u"\u00a0", " ")
        f_d.seek(0)
        f_d.write(data)


font_database = defaultdict(set)

# test
#extract_fonts("p126.svg", font_database)
#replace_private_chars("p194.svgz")

for svgz_file in glob.glob("*.svgz"):

    print(svgz_file)

    extract_fonts(svgz_file, font_database)
    replace_private_chars(svgz_file)


#print(font_database.values())
#print(len(font_database))

#for subset in font_database.values():
    #assert len(subset) == 1

#=> no font duplication...