# -*- coding: utf-8 -*-
#
# Copyright (c) 2012-2014 Ciro Mattia Gonano <ciromattia@gmail.com>
# Copyright (c) 2013-2017 Pawel Jastrzebski <pawelj@iosphe.re>
#
# Permission to use, copy, modify, and/or distribute this software for
# any purpose with or without fee is hereby granted, provided that the
# above copyright notice and this permission notice appear in all
# copies.
#
# THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL
# WARRANTIES WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED
# WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE
# AUTHOR BE LIABLE FOR ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL
# DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA
# OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER
# TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR
# PERFORMANCE OF THIS SOFTWARE.
#
import os
import sys
from time import strftime, gmtime
from copy import copy
from glob import glob
from json import loads
from urllib.request import Request, urlopen
from re import sub
from stat import S_IWRITE, S_IREAD, S_IEXEC
from zipfile import ZipFile, ZIP_STORED, ZIP_DEFLATED
from tempfile import mkdtemp, gettempdir, TemporaryFile
from shutil import move, copytree, rmtree
from optparse import OptionParser, OptionGroup
from multiprocessing import Pool
from uuid import uuid4
from slugify import slugify as slugifyExt
from PIL import Image
from subprocess import STDOUT, PIPE
from psutil import Popen, virtual_memory
from html import escape
try:
from PyQt5 import QtCore
except ImportError:
QtCore = None
from .shared import md5Checksum, getImageFileName, walkSort, walkLevel, sanitizeTrace
from . import comic2panel
from . import image
from . import cbxarchive
from . import pdfjpgextract
from . import dualmetafix
from . import metadata
from . import kindle
from . import __version__
def main(argv=None):
global options
parser = makeParser()
optionstemplate, args = parser.parse_args(argv)
if len(args) == 0:
parser.print_help()
return 0
if sys.platform.startswith('win'):
sources = set([source for arg in args for source in glob(arg)])
else:
sources = set(args)
if len(sources) == 0:
print('No matching files found.')
return 1
for source in sources:
source = source.rstrip('\\').rstrip('/')
options = copy(optionstemplate)
checkOptions()
if len(sources) > 1:
print('Working on ' + source + '...')
makeBook(source)
return 0
def buildHTML(path, imgfile, imgfilepath):
imgfilepath = md5Checksum(imgfilepath)
filename = getImageFileName(imgfile)
deviceres = options.profileData[1]
if "Rotated" in options.imgMetadata[imgfilepath]:
rotatedPage = True
else:
rotatedPage = False
if "BlackBackground" in options.imgMetadata[imgfilepath]:
additionalStyle = 'background-color:#000000;'
else:
additionalStyle = 'background-color:#FFFFFF;'
postfix = ''
backref = 1
head = path
while True:
head, tail = os.path.split(head)
if tail == 'Images':
htmlpath = os.path.join(head, 'Text', postfix)
break
postfix = tail + "/" + postfix
backref += 1
if not os.path.exists(htmlpath):
os.makedirs(htmlpath)
htmlfile = os.path.join(htmlpath, filename[0] + '.xhtml')
imgsize = Image.open(os.path.join(head, "Images", postfix, imgfile)).size
f = open(htmlfile, "w", encoding='UTF-8')
f.writelines(["<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n",
"<!DOCTYPE html>\n",
"<html xmlns=\"http://www.w3.org/1999/xhtml\" xmlns:epub=\"http://www.idpf.org/2007/ops\">\n",
"<head>\n",
"<title>", escape(filename[0]), "</title>\n",
"<link href=\"", "../" * (backref - 1), "style.css\" type=\"text/css\" rel=\"stylesheet\"/>\n",
"<meta name=\"viewport\" "
"content=\"width=" + str(deviceres[0]) + ", height=" + str(deviceres[1]) + "\"/>\n"
"</head>\n",
"<body style=\"" + additionalStyle + "\">\n",
"<div style=\"text-align:center;top:" + getTopMargin(deviceres, imgsize) + "%;\">\n",
"<img width=\"" + str(imgsize[0]) + "\" height=\"" + str(imgsize[1]) + "\" ",
"src=\"", "../" * backref, "Images/", postfix, imgfile, "\"/>\n</div>\n"])
if options.iskindle and options.panelview:
if options.autoscale:
size = (getPanelViewResolution(imgsize, deviceres))
else:
size = (int(imgsize[0] * 1.5), int(imgsize[1] * 1.5))
if size[0] - deviceres[0] < deviceres[0] * 0.01:
noHorizontalPV = True
else:
noHorizontalPV = False
if size[1] - deviceres[1] < deviceres[1] * 0.01:
noVerticalPV = True
else:
noVerticalPV = False
x, y = getPanelViewSize(deviceres, size)
boxStyles = {"PV-TL": "position:absolute;left:0;top:0;",
"PV-TR": "position:absolute;right:0;top:0;",
"PV-BL": "position:absolute;left:0;bottom:0;",
"PV-BR": "position:absolute;right:0;bottom:0;",
"PV-T": "position:absolute;top:0;left:" + x + "%;",
"PV-B": "position:absolute;bottom:0;left:" + x + "%;",
"PV-L": "position:absolute;left:0;top:" + y + "%;",
"PV-R": "position:absolute;right:0;top:" + y + "%;"}
f.write("<div id=\"PV\">\n")
if not noHorizontalPV and not noVerticalPV:
if rotatedPage:
if options.righttoleft:
order = [1, 3, 2, 4]
else:
order = [2, 4, 1, 3]
else:
if options.righttoleft:
order = [2, 1, 4, 3]
else:
order = [1, 2, 3, 4]
boxes = ["PV-TL", "PV-TR", "PV-BL", "PV-BR"]
elif noHorizontalPV and not noVerticalPV:
if rotatedPage:
if options.righttoleft:
order = [1, 2]
else:
order = [2, 1]
else:
order = [1, 2]
boxes = ["PV-T", "PV-B"]
elif not noHorizontalPV and noVerticalPV:
if rotatedPage:
order = [1, 2]
else:
if options.righttoleft:
order = [2, 1]
else:
order = [1, 2]
boxes = ["PV-L", "PV-R"]
else:
order = []
boxes = []
for i in range(0, len(boxes)):
f.writelines(["<div id=\"" + boxes[i] + "\">\n",
"<a style=\"display:inline-block;width:100%;height:100%;\" class=\"app-amzn-magnify\" "
"data-app-amzn-magnify='{\"targetId\":\"" + boxes[i] +
"-P\", \"ordinal\":" + str(order[i]) + "}'></a>\n",
"</div>\n"])
f.write("</div>\n")
for box in boxes:
f.writelines(["<div class=\"PV-P\" id=\"" + box + "-P\" style=\"" + additionalStyle + "\">\n",
"<img style=\"" + boxStyles[box] + "\" src=\"", "../" * backref, "Images/", postfix,
imgfile, "\" width=\"" + str(size[0]) + "\" height=\"" + str(size[1]) + "\"/>\n",
"</div>\n"])
f.writelines(["</body>\n",
"</html>\n"])
f.close()
return path, imgfile
def buildNCX(dstdir, title, chapters, chapterNames):
ncxfile = os.path.join(dstdir, 'OEBPS', 'toc.ncx')
f = open(ncxfile, "w", encoding='UTF-8')
f.writelines(["<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n",
"<ncx version=\"2005-1\" xml:lang=\"en-US\" xmlns=\"http://www.daisy.org/z3986/2005/ncx/\">\n",
"<head>\n",
"<meta name=\"dtb:uid\" content=\"urn:uuid:", options.uuid, "\"/>\n",
"<meta name=\"dtb:depth\" content=\"1\"/>\n",
"<meta name=\"dtb:totalPageCount\" content=\"0\"/>\n",
"<meta name=\"dtb:maxPageNumber\" content=\"0\"/>\n",
"<meta name=\"generated\" content=\"true\"/>\n",
"</head>\n",
"<docTitle><text>", escape(title), "</text></docTitle>\n",
"<navMap>\n"])
for chapter in chapters:
folder = chapter[0].replace(os.path.join(dstdir, 'OEBPS'), '').lstrip('/').lstrip('\\\\')
filename = getImageFileName(os.path.join(folder, chapter[1]))
navID = folder.replace('/', '_').replace('\\', '_')
if options.chapters:
title = chapterNames[chapter[1]]
navID = filename[0].replace('/', '_').replace('\\', '_')
elif os.path.basename(folder) != "Text":
title = chapterNames[os.path.basename(folder)]
f.write("<navPoint id=\"" + navID + "\"><navLabel><text>" +
escape(title) + "</text></navLabel><content src=\"" + filename[0].replace("\\", "/") +
".xhtml\"/></navPoint>\n")
f.write("</navMap>\n</ncx>")
f.close()
def buildNAV(dstdir, title, chapters, chapterNames):
navfile = os.path.join(dstdir, 'OEBPS', 'nav.xhtml')
f = open(navfile, "w", encoding='UTF-8')
f.writelines(["<?xml version=\"1.0\" encoding=\"utf-8\"?>\n",
"<!DOCTYPE html>\n",
"<html xmlns=\"http://www.w3.org/1999/xhtml\" xmlns:epub=\"http://www.idpf.org/2007/ops\">\n",
"<head>\n",
"<title>" + escape(title) + "</title>\n",
"<meta charset=\"utf-8\"/>\n",
"</head>\n",
"<body>\n",
"<nav xmlns:epub=\"http://www.idpf.org/2007/ops\" epub:type=\"toc\" id=\"toc\">\n",
"<ol>\n"])
for chapter in chapters:
folder = chapter[0].replace(os.path.join(dstdir, 'OEBPS'), '').lstrip('/').lstrip('\\\\')
filename = getImageFileName(os.path.join(folder, chapter[1]))
if options.chapters:
title = chapterNames[chapter[1]]
elif os.path.basename(folder) != "Text":
title = chapterNames[os.path.basename(folder)]
f.write("<li><a href=\"" + filename[0].replace("\\", "/") + ".xhtml\">" + escape(title) + "</a></li>\n")
f.writelines(["</ol>\n",
"</nav>\n",
"<nav epub:type=\"page-list\">\n",
"<ol>\n"])
for chapter in chapters:
folder = chapter[0].replace(os.path.join(dstdir, 'OEBPS'), '').lstrip('/').lstrip('\\\\')
filename = getImageFileName(os.path.join(folder, chapter[1]))
if options.chapters:
title = chapterNames[chapter[1]]
elif os.path.basename(folder) != "Text":
title = chapterNames[os.path.basename(folder)]
f.write("<li><a href=\"" + filename[0].replace("\\", "/") + ".xhtml\">" + escape(title) + "</a></li>\n")
f.write("</ol>\n</nav>\n</body>\n</html>")
f.close()
def buildOPF(dstdir, title, filelist, cover=None):
opffile = os.path.join(dstdir, 'OEBPS', 'content.opf')
deviceres = options.profileData[1]
if options.righttoleft:
writingmode = "horizontal-rl"
else:
writingmode = "horizontal-lr"
f = open(opffile, "w", encoding='UTF-8')
f.writelines(["<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n",
"<package version=\"3.0\" unique-identifier=\"BookID\" ",
"xmlns=\"http://www.idpf.org/2007/opf\">\n",
"<metadata xmlns:opf=\"http://www.idpf.org/2007/opf\" ",
"xmlns:dc=\"http://purl.org/dc/elements/1.1/\">\n",
"<dc:title>", title, "</dc:title>\n",
"<dc:language>en-US</dc:language>\n",
"<dc:identifier id=\"BookID\">urn:uuid:", options.uuid, "</dc:identifier>\n",
"<dc:contributor id=\"contributor\">KindleComicConverter-" + __version__ + "</dc:contributor>\n"])
if len(options.summary) > 0:
f.writelines(["<dc:description>", options.summary, "</dc:description>\n"])
for author in options.authors:
f.writelines(["<dc:creator>", author, "</dc:creator>\n"])
f.writelines(["<meta property=\"dcterms:modified\">" + strftime("%Y-%m-%dT%H:%M:%SZ", gmtime()) + "</meta>\n",
"<meta name=\"cover\" content=\"cover\"/>\n",
"<meta property=\"rendition:orientation\">portrait</meta>\n",
"<meta property=\"rendition:spread\">portrait</meta>\n",
"<meta property=\"rendition:layout\">pre-paginated</meta>\n"])
if options.iskindle and options.profile != 'Custom':
f.writelines(["<meta name=\"original-resolution\" content=\"",
str(deviceres[0]) + "x" + str(deviceres[1]) + "\"/>\n",
"<meta name=\"book-type\" content=\"comic\"/>\n",
"<meta name=\"RegionMagnification\" content=\"true\"/>\n",
"<meta name=\"primary-writing-mode\" content=\"" + writingmode + "\"/>\n",
"<meta name=\"zero-gutter\" content=\"true\"/>\n",
"<meta name=\"zero-margin\" content=\"true\"/>\n",
"<meta name=\"ke-border-color\" content=\"#ffffff\"/>\n",
"<meta name=\"ke-border-width\" content=\"0\"/>\n"])
f.writelines(["</metadata>\n<manifest>\n<item id=\"ncx\" href=\"toc.ncx\" ",
"media-type=\"application/x-dtbncx+xml\"/>\n",
"<item id=\"nav\" href=\"nav.xhtml\" ",
"properties=\"nav\" media-type=\"application/xhtml+xml\"/>\n"])
if cover is not None:
filename = getImageFileName(cover.replace(os.path.join(dstdir, 'OEBPS'), '').lstrip('/').lstrip('\\\\'))
if '.png' == filename[1]:
mt = 'image/png'
else:
mt = 'image/jpeg'
f.write("<item id=\"cover\" href=\"Images/cover" + filename[1] + "\" media-type=\"" + mt +
"\" properties=\"cover-image\"/>\n")
reflist = []
for path in filelist:
folder = path[0].replace(os.path.join(dstdir, 'OEBPS'), '').lstrip('/').lstrip('\\\\').replace("\\", "/")
filename = getImageFileName(path[1])
uniqueid = os.path.join(folder, filename[0]).replace('/', '_').replace('\\', '_')
reflist.append(uniqueid)
f.write("<item id=\"page_" + str(uniqueid) + "\" href=\"" +
folder.replace('Images', 'Text') + "/" + filename[0] +
".xhtml\" media-type=\"application/xhtml+xml\"/>\n")
if '.png' == filename[1]:
mt = 'image/png'
else:
mt = 'image/jpeg'
f.write("<item id=\"img_" + str(uniqueid) + "\" href=\"" + folder + "/" + path[1] + "\" media-type=\"" +
mt + "\"/>\n")
f.write("<item id=\"css\" href=\"Text/style.css\" media-type=\"text/css\"/>\n")
if options.righttoleft:
f.write("</manifest>\n<spine page-progression-direction=\"rtl\" toc=\"ncx\">\n")
else:
f.write("</manifest>\n<spine page-progression-direction=\"ltr\" toc=\"ncx\">\n")
for entry in reflist:
f.write("<itemref idref=\"page_" + entry + "\"/>\n")
f.write("</spine>\n</package>\n")
f.close()
os.mkdir(os.path.join(dstdir, 'META-INF'))
f = open(os.path.join(dstdir, 'META-INF', 'container.xml'), 'w', encoding='UTF-8')
f.writelines(["<?xml version=\"1.0\"?>\n",
"<container version=\"1.0\" xmlns=\"urn:oasis:names:tc:opendocument:xmlns:container\">\n",
"<rootfiles>\n",
"<rootfile full-path=\"OEBPS/content.opf\" media-type=\"application/oebps-package+xml\"/>\n",
"</rootfiles>\n",
"</container>"])
f.close()
def buildEPUB(path, chapterNames, tomeNumber):
filelist = []
chapterlist = []
cover = None
os.mkdir(os.path.join(path, 'OEBPS', 'Text'))
f = open(os.path.join(path, 'OEBPS', 'Text', 'style.css'), 'w', encoding='UTF-8')
f.writelines(["@page {\n",
"margin: 0;\n",
"}\n",
"body {\n",
"display: block;\n",
"margin: 0;\n",
"padding: 0;\n",
"}\n",
"#PV {\n",
"position: absolute;\n",
"width: 100%;\n",
"height: 100%;\n",
"top: 0;\n",
"left: 0;\n",
"}\n",
"#PV-T {\n",
"top: 0;\n",
"width: 100%;\n",
"height: 50%;\n",
"}\n",
"#PV-B {\n",
"bottom: 0;\n",
"width: 100%;\n",
"height: 50%;\n",
"}\n",
"#PV-L {\n",
"left: 0;\n",
"width: 49.5%;\n",
"height: 100%;\n",
"float: left;\n",
"}\n",
"#PV-R {\n",
"right: 0;\n",
"width: 49.5%;\n",
"height: 100%;\n",
"float: right;\n",
"}\n",
"#PV-TL {\n",
"top: 0;\n",
"left: 0;\n",
"width: 49.5%;\n",
"height: 50%;\n",
"float: left;\n",
"}\n",
"#PV-TR {\n",
"top: 0;\n",
"right: 0;\n",
"width: 49.5%;\n",
"height: 50%;\n",
"float: right;\n",
"}\n",
"#PV-BL {\n",
"bottom: 0;\n",
"left: 0;\n",
"width: 49.5%;\n",
"height: 50%;\n",
"float: left;\n",
"}\n",
"#PV-BR {\n",
"bottom: 0;\n",
"right: 0;\n",
"width: 49.5%;\n",
"height: 50%;\n",
"float: right;\n",
"}\n",
".PV-P {\n",
"width: 100%;\n",
"height: 100%;\n",
"top: 0;\n",
"position: absolute;\n",
"display: none;\n",
"}\n"])
f.close()
for dirpath, dirnames, filenames in os.walk(os.path.join(path, 'OEBPS', 'Images')):
chapter = False
dirnames, filenames = walkSort(dirnames, filenames)
for afile in filenames:
filelist.append(buildHTML(dirpath, afile, os.path.join(dirpath, afile)))
if not chapter:
chapterlist.append((dirpath.replace('Images', 'Text'), filelist[-1][1]))
chapter = True
if cover is None:
cover = os.path.join(os.path.join(path, 'OEBPS', 'Images'),
'cover' + getImageFileName(filelist[-1][1])[1])
options.covers.append((image.Cover(os.path.join(filelist[-1][0], filelist[-1][1]), cover, options,
tomeNumber), options.uuid))
# Overwrite chapternames if tree is flat and ComicInfo.xml has bookmarks
if not chapterNames and options.chapters:
chapterlist = []
globaldiff = 0
for aChapter in options.chapters:
pageid = aChapter[0]
for x in range(0, pageid + globaldiff + 1):
if '-kcc-b' in filelist[x][1]:
pageid += 1
if '-kcc-c' in filelist[pageid][1]:
pageid -= 1
filename = filelist[pageid][1]
chapterlist.append((filelist[pageid][0].replace('Images', 'Text'), filename))
chapterNames[filename] = aChapter[1]
globaldiff = pageid - (aChapter[0] + globaldiff)
buildNCX(path, options.title, chapterlist, chapterNames)
buildNAV(path, options.title, chapterlist, chapterNames)
buildOPF(path, options.title, filelist, cover)
def imgDirectoryProcessing(path):
global workerPool, workerOutput
workerPool = Pool()
workerOutput = []
options.imgMetadata = {}
options.imgOld = []
work = []
pagenumber = 0
for dirpath, _, filenames in os.walk(path):
for afile in filenames:
pagenumber += 1
work.append([afile, dirpath, options])
if GUI:
GUI.progressBarTick.emit(str(pagenumber))
if len(work) > 0:
for i in work:
workerPool.apply_async(func=imgFileProcessing, args=(i, ), callback=imgFileProcessingTick)
workerPool.close()
workerPool.join()
if GUI and not GUI.conversionAlive:
rmtree(os.path.join(path, '..', '..'), True)
raise UserWarning("Conversion interrupted.")
if len(workerOutput) > 0:
rmtree(os.path.join(path, '..', '..'), True)
raise RuntimeError("One of workers crashed. Cause: " + workerOutput[0][0], workerOutput[0][1])
for file in options.imgOld:
if os.path.isfile(file):
os.remove(file)
else:
rmtree(os.path.join(path, '..', '..'), True)
raise UserWarning("Source directory is empty.")
def imgFileProcessingTick(output):
if isinstance(output, tuple):
workerOutput.append(output)
workerPool.terminate()
else:
for page in output:
if page is not None:
options.imgMetadata[page[0]] = page[1]
options.imgOld.append(page[2])
if GUI:
GUI.progressBarTick.emit('tick')
if not GUI.conversionAlive:
workerPool.terminate()
def imgFileProcessing(work):
try:
afile = work[0]
dirpath = work[1]
opt = work[2]
output = []
workImg = image.ComicPageParser((dirpath, afile), opt)
for i in workImg.payload:
img = image.ComicPage(opt, *i)
if opt.cropping == 2 and not opt.webtoon:
img.cropPageNumber(opt.croppingp)
if opt.cropping > 0 and not opt.webtoon:
img.cropMargin(opt.croppingp)
img.autocontrastImage()
img.resizeImage()
if opt.forcepng and not opt.forcecolor:
img.quantizeImage()
output.append(img.saveToDir())
return output
except Exception:
return str(sys.exc_info()[1]), sanitizeTrace(sys.exc_info()[2])
def getWorkFolder(afile):
if os.path.isdir(afile):
workdir = mkdtemp('', 'KCC-')
try:
os.rmdir(workdir)
fullPath = os.path.join(workdir, 'OEBPS', 'Images')
copytree(afile, fullPath)
sanitizePermissions(fullPath)
return workdir
except:
rmtree(workdir, True)
raise UserWarning("Failed to prepare a workspace.")
elif os.path.isfile(afile) and afile.lower().endswith('.pdf'):
pdf = pdfjpgextract.PdfJpgExtract(afile)
path, njpg = pdf.extract()
if njpg == 0:
rmtree(path, True)
raise UserWarning("Failed to extract images from PDF file.")
elif os.path.isfile(afile):
workdir = mkdtemp('', 'KCC-')
cbx = cbxarchive.CBxArchive(afile)
if cbx.isCbxFile():
try:
path = cbx.extract(workdir)
except:
rmtree(workdir, True)
raise UserWarning("Failed to extract archive.")
else:
rmtree(workdir, True)
raise UserWarning("Failed to detect archive format.")
else:
raise UserWarning("Failed to open source file/directory.")
sanitizePermissions(path)
newpath = mkdtemp('', 'KCC-')
copytree(path, os.path.join(newpath, 'OEBPS', 'Images'))
rmtree(path, True)
return newpath
def getOutputFilename(srcpath, wantedname, ext, tomeNumber):
if srcpath[-1] == os.path.sep:
srcpath = srcpath[:-1]
if 'Ko' in options.profile and options.format == 'EPUB':
ext = '.kepub.epub'
if wantedname is not None:
if wantedname.endswith(ext):
filename = os.path.abspath(wantedname)
elif os.path.isdir(srcpath):
filename = os.path.join(os.path.abspath(options.output), os.path.basename(srcpath) + ext)
else:
filename = os.path.join(os.path.abspath(options.output),
os.path.basename(os.path.splitext(srcpath)[0]) + ext)
elif os.path.isdir(srcpath):
filename = srcpath + tomeNumber + ext
else:
if 'Ko' in options.profile and options.format == 'EPUB':
path = srcpath.split(os.path.sep)
path[-1] = ''.join(e for e in path[-1].split('.')[0] if e.isalnum()) + tomeNumber + ext
if not path[-1].split('.')[0]:
path[-1] = 'KCCPlaceholder' + tomeNumber + ext
filename = os.path.sep.join(path)
else:
filename = os.path.splitext(srcpath)[0] + tomeNumber + ext
if os.path.isfile(filename):
counter = 0
basename = os.path.splitext(filename)[0]
while os.path.isfile(basename + '_kcc' + str(counter) + ext):
counter += 1
filename = basename + '_kcc' + str(counter) + ext
return filename
def getComicInfo(path, originalPath):
xmlPath = os.path.join(path, 'ComicInfo.xml')
options.authors = ['KCC']
options.remoteCovers = {}
options.chapters = []
options.summary = ''
titleSuffix = ''
if options.title == 'defaulttitle':
defaultTitle = True
if os.path.isdir(originalPath):
options.title = os.path.basename(originalPath)
else:
options.title = os.path.splitext(os.path.basename(originalPath))[0]
else:
defaultTitle = False
if os.path.exists(xmlPath):
try:
xml = metadata.MetadataParser(xmlPath)
except Exception:
os.remove(xmlPath)
return
options.authors = []
if defaultTitle:
if xml.data['Series']:
options.title = escape(xml.data['Series'])
if xml.data['Volume']:
titleSuffix += ' V' + xml.data['Volume'].zfill(2)
if xml.data['Number']:
titleSuffix += ' #' + xml.data['Number'].zfill(3)
options.title += titleSuffix
for field in ['Writers', 'Pencillers', 'Inkers', 'Colorists']:
for person in xml.data[field]:
options.authors.append(escape(person))
if len(options.authors) > 0:
options.authors = list(set(options.authors))
options.authors.sort()
else:
options.authors = ['KCC']
if xml.data['MUid']:
options.remoteCovers = getCoversFromMCB(xml.data['MUid'])
if xml.data['Bookmarks']:
options.chapters = xml.data['Bookmarks']
if xml.data['Summary']:
options.summary = escape(xml.data['Summary'])
os.remove(xmlPath)
def getCoversFromMCB(mangaID):
covers = {}
try:
jsonRaw = urlopen(Request('http://mcd.iosphe.re/api/v1/series/' + mangaID + '/',
headers={'User-Agent': 'KindleComicConverter/' + __version__}))
jsonData = loads(jsonRaw.read().decode('utf-8'))
for volume in jsonData['Covers']['a']:
if volume['Side'] == 'front':
covers[int(volume['Volume'])] = volume['Raw']
except Exception:
return {}
return covers
def getDirectorySize(start_path='.'):
total_size = 0
for dirpath, _, filenames in os.walk(start_path):
for f in filenames:
fp = os.path.join(dirpath, f)
total_size += os.path.getsize(fp)
return total_size
def getTopMargin(deviceres, size):
y = int((deviceres[1] - size[1]) / 2) / deviceres[1] * 100
return str(round(y, 1))
def getPanelViewResolution(imageSize, deviceRes):
scale = float(deviceRes[0]) / float(imageSize[0])
return int(deviceRes[0]), int(scale * imageSize[1])
def getPanelViewSize(deviceres, size):
x = int(deviceres[0] / 2 - size[0] / 2) / deviceres[0] * 100
y = int(deviceres[1] / 2 - size[1] / 2) / deviceres[1] * 100
return str(int(x)), str(int(y))
def sanitizeTree(filetree):
chapterNames = {}
for root, dirs, files in os.walk(filetree, False):
for name in files:
splitname = os.path.splitext(name)
slugified = slugify(splitname[0])
while os.path.exists(os.path.join(root, slugified + splitname[1])) and splitname[0].upper()\
!= slugified.upper():
slugified += "A"
newKey = os.path.join(root, slugified + splitname[1])
key = os.path.join(root, name)
if key != newKey:
os.replace(key, newKey)
for name in dirs:
tmpName = name
slugified = slugify(name)
while os.path.exists(os.path.join(root, slugified)) and name.upper() != slugified.upper():
slugified += "A"
chapterNames[slugified] = tmpName
newKey = os.path.join(root, slugified)
key = os.path.join(root, name)
if key != newKey:
os.replace(key, newKey)
return chapterNames
def sanitizeTreeKobo(filetree):
pageNumber = 0
for root, dirs, files in os.walk(filetree):
dirs, files = walkSort(dirs, files)
for name in files:
splitname = os.path.splitext(name)
slugified = str(pageNumber).zfill(5)
pageNumber += 1
while os.path.exists(os.path.join(root, slugified + splitname[1])) and splitname[0].upper()\
!= slugified.upper():
slugified += "A"
newKey = os.path.join(root, slugified + splitname[1])
key = os.path.join(root, name)
if key != newKey:
os.replace(key, newKey)
def sanitizePermissions(filetree):
for root, dirs, files in os.walk(filetree, False):
for name in files:
os.chmod(os.path.join(root, name), S_IWRITE | S_IREAD)
for name in dirs:
os.chmod(os.path.join(root, name), S_IWRITE | S_IREAD | S_IEXEC)
def splitDirectory(path):
level = -1
for root, _, files in os.walk(os.path.join(path, 'OEBPS', 'Images')):
for f in files:
if f.endswith('.jpg') or f.endswith('.jpeg') or f.endswith('.png') or f.endswith('.gif'):
newLevel = os.path.join(root, f).replace(os.path.join(path, 'OEBPS', 'Images'), '').count(os.sep)
if level != -1 and level != newLevel:
level = 0
break
else:
level = newLevel
if level > 0:
splitter = splitProcess(os.path.join(path, 'OEBPS', 'Images'), level)
path = [path]
for tome in splitter:
path.append(tome)
return path
else:
raise UserWarning('Unsupported directory structure.')
def splitProcess(path, mode):
output = []
currentSize = 0
currentTarget = path
if options.webtoon:
targetSize = 104857600
else:
targetSize = 419430400
if options.batchsplit == 2 and mode == 2:
mode = 3
if mode < 3:
for root, dirs, files in walkLevel(path, 0):
for name in files if mode == 1 else dirs:
if mode == 1:
size = os.path.getsize(os.path.join(root, name))
else:
size = getDirectorySize(os.path.join(root, name))
if currentSize + size > targetSize:
currentTarget, pathRoot = createNewTome()
output.append(pathRoot)
currentSize = size
else:
currentSize += size
if path != currentTarget:
move(os.path.join(root, name), os.path.join(currentTarget, name))
else:
firstTome = True
for root, dirs, _ in walkLevel(path, 0):
for name in dirs:
if not firstTome:
currentTarget, pathRoot = createNewTome()
output.append(pathRoot)
move(os.path.join(root, name), os.path.join(currentTarget, name))
else:
firstTome = False
return output
def detectCorruption(tmpPath, orgPath):
imageNumber = 0
imageSmaller = 0
for root, _, files in os.walk(tmpPath, False):
for name in files:
if getImageFileName(name) is not None:
path = os.path.join(root, name)
pathOrg = orgPath + path.split('OEBPS' + os.path.sep + 'Images')[1]
if os.path.getsize(path) == 0:
rmtree(os.path.join(tmpPath, '..', '..'), True)
raise RuntimeError('Image file %s is corrupted.' % pathOrg)
try:
img = Image.open(path)
img.verify()
img = Image.open(path)
img.load()
imageNumber += 1
if options.profileData[1][0] > img.size[0] and options.profileData[1][1] > img.size[1]:
imageSmaller += 1
except Exception as err:
rmtree(os.path.join(tmpPath, '..', '..'), True)
if 'decoder' in str(err) and 'not available' in str(err):
raise RuntimeError('Pillow was compiled without JPG and/or PNG decoder.')
else:
raise RuntimeError('Image file %s is corrupted.' % pathOrg)
else:
os.remove(os.path.join(root, name))
if imageSmaller > imageNumber * 0.25 and not options.upscale and not options.stretch:
print("WARNING: More than 25% of images are smaller than target device resolution. "
"Consider enabling stretching or upscaling to improve readability.")
if GUI:
GUI.addMessage.emit('More than 25% of images are smaller than target device resolution.', 'warning', False)
GUI.addMessage.emit('Consider enabling stretching or upscaling to improve readability.', 'warning', False)
GUI.addMessage.emit('', '', False)
def createNewTome():
tomePathRoot = mkdtemp('', 'KCC-')
tomePath = os.path.join(tomePathRoot, 'OEBPS', 'Images')
os.makedirs(tomePath)
return tomePath, tomePathRoot
def slugify(value):
value = slugifyExt(value)
value = sub(r'0*([0-9]{4,})', r'\1', sub(r'([0-9]+)', r'0000\1', value, count=2))
return value
def makeZIP(zipFilename, baseDir, isEPUB=False):
zipFilename = os.path.abspath(zipFilename) + '.zip'
zipOutput = ZipFile(zipFilename, 'w', ZIP_DEFLATED)
if isEPUB:
zipOutput.writestr('mimetype', 'application/epub+zip', ZIP_STORED)
for dirpath, _, filenames in os.walk(baseDir):
for name in filenames:
path = os.path.normpath(os.path.join(dirpath, name))
aPath = os.path.normpath(os.path.join(dirpath.replace(baseDir, ''), name))
if os.path.isfile(path):
zipOutput.write(path, aPath)
zipOutput.close()
return zipFilename
def makeParser():
psr = OptionParser(usage="Usage: kcc-c2e [options] comic_file|comic_folder", add_help_option=False)
mainOptions = OptionGroup(psr, "MAIN")
processingOptions = OptionGroup(psr, "PROCESSING")
outputOptions = OptionGroup(psr, "OUTPUT SETTINGS")
customProfileOptions = OptionGroup(psr, "CUSTOM PROFILE")
otherOptions = OptionGroup(psr, "OTHER")
mainOptions.add_option("-p", "--profile", action="store", dest="profile", default="KV",
help="Device profile (Available options: K1, K2, K3, K45, KDX, KPW, KV, KoMT, KoG, KoGHD,"
" KoA, KoAHD, KoAH2O, KoAO) [Default=KV]")
mainOptions.add_option("-m", "--manga-style", action="store_true", dest="righttoleft", default=False,
help="Manga style (right-to-left reading and splitting)")
mainOptions.add_option("-2", "--two-panel", action="store_true", dest="autoscale", default=False,
help="Display two not four panels in Panel View mode")
mainOptions.add_option("-w", "--webtoon", action="store_true", dest="webtoon", default=False,
help="Webtoon processing mode"),
outputOptions.add_option("-o", "--output", action="store", dest="output", default=None,
help="Output generated file to specified directory or file")
outputOptions.add_option("-t", "--title", action="store", dest="title", default="defaulttitle",
help="Comic title [Default=filename or directory name]")
outputOptions.add_option("-f", "--format", action="store", dest="format", default="Auto",
help="Output format (Available options: Auto, MOBI, EPUB, CBZ) [Default=Auto]")
outputOptions.add_option("-b", "--batchsplit", type="int", dest="batchsplit", default="0",
help="Split output into multiple files. 0: Don't split 1: Automatic mode "
"2: Consider every subdirectory as separate volume [Default=0]")
processingOptions.add_option("-u", "--upscale", action="store_true", dest="upscale", default=False,
help="Resize images smaller than device's resolution")
processingOptions.add_option("-s", "--stretch", action="store_true", dest="stretch", default=False,
help="Stretch images to device's resolution")
processingOptions.add_option("-r", "--splitter", type="int", dest="splitter", default="0",
help="Double page parsing mode. 0: Split 1: Rotate 2: Both [Default=0]")
processingOptions.add_option("-g", "--gamma", type="float", dest="gamma", default="0.0",
help="Apply gamma correction to linearize the image [Default=Auto]")
processingOptions.add_option("-c", "--cropping", type="int", dest="cropping", default="2",
help="Set cropping mode. 0: Disabled 1: Margins 2: Margins + page numbers [Default=2]")
processingOptions.add_option("--cp", "--croppingpower", type="float", dest="croppingp", default="1.0",
help="Set cropping power [Default=1.0]")
processingOptions.add_option("--blackborders", action="store_true", dest="black_borders", default=False,
help="Disable autodetection and force black borders")
processingOptions.add_option("--whiteborders", action="store_true", dest="white_borders", default=False,
help="Disable autodetection and force white borders")
processingOptions.add_option("--forcecolor", action="store_true", dest="forcecolor", default=False,
help="Don't convert images to grayscale")
processingOptions.add_option("--forcepng", action="store_true", dest="forcepng", default=False,
help="Create PNG files instead JPEG")
customProfileOptions.add_option("--customwidth", type="int", dest="customwidth", default=0,
help="Replace screen width provided by device profile")
customProfileOptions.add_option("--customheight", type="int", dest="customheight", default=0,
help="Replace screen height provided by device profile")
otherOptions.add_option("-h", "--help", action="help",
help="Show this help message and exit")
psr.add_option_group(mainOptions)
psr.add_option_group(outputOptions)
psr.add_option_group(processingOptions)
psr.add_option_group(customProfileOptions)
psr.add_option_group(otherOptions)
return psr
def checkOptions():
global options
options.panelview = True
options.iskindle = False
options.bordersColor = None
if options.format == 'Auto':
if options.profile in ['K1', 'K2', 'K3', 'K45', 'KPW', 'KV']:
options.format = 'MOBI'
elif options.profile in ['OTHER', 'KoMT', 'KoG', 'KoGHD', 'KoA', 'KoAHD', 'KoAH2O', 'KoAO']:
options.format = 'EPUB'
elif options.profile in ['KDX']:
options.format = 'CBZ'
if options.profile in ['K1', 'K2', 'K3', 'K45', 'KPW', 'KV']:
options.iskindle = True
if options.white_borders:
options.bordersColor = 'white'
if options.black_borders:
options.bordersColor = 'black'
# Splitting MOBI is not optional
if options.format == 'MOBI' and options.batchsplit != 2:
options.batchsplit = 1
# Older Kindle models don't support Panel View.
if options.profile == 'K1' or options.profile == 'K2' or options.profile == 'KDX':
options.panelview = False
# Webtoon mode mandatory options
if options.webtoon:
options.panelview = False
options.righttoleft = False
options.upscale = True
# Disable all Kindle features for other e-readers
if options.profile == 'OTHER':
options.panelview = False
if 'Ko' in options.profile:
options.panelview = False
# CBZ files on Kindle DX/DXG support higher resolution
if options.profile == 'KDX' and options.format == 'CBZ':
options.customheight = 1200
# Override profile data
if options.customwidth != 0 or options.customheight != 0:
X = image.ProfileData.Profiles[options.profile][1][0]
Y = image.ProfileData.Profiles[options.profile][1][1]
if options.customwidth != 0:
X = options.customwidth
if options.customheight != 0:
Y = options.customheight
newProfile = ("Custom", (int(X), int(Y)), image.ProfileData.Palette16,
image.ProfileData.Profiles[options.profile][3])
image.ProfileData.Profiles["Custom"] = newProfile
options.profile = "Custom"
options.profileData = image.ProfileData.Profiles[options.profile]
def checkTools(source):
source = source.upper()
if source.endswith('.CBR') or source.endswith('.RAR'):
rarExitCode = Popen('unrar', stdout=PIPE, stderr=STDOUT, stdin=PIPE, shell=True)
rarExitCode = rarExitCode.wait()
if rarExitCode != 0 and rarExitCode != 7:
print('ERROR: UnRAR is missing!')
exit(1)
elif source.endswith('.CB7') or source.endswith('.7Z'):
sevenzaExitCode = Popen('7za', stdout=PIPE, stderr=STDOUT, stdin=PIPE, shell=True)
sevenzaExitCode = sevenzaExitCode.wait()
if sevenzaExitCode != 0 and sevenzaExitCode != 7:
print('ERROR: 7za is missing!')
exit(1)
if options.format == 'MOBI':
kindleGenExitCode = Popen('kindlegen -locale en', stdout=PIPE, stderr=STDOUT, stdin=PIPE, shell=True)
if kindleGenExitCode.wait() != 0:
print('ERROR: KindleGen is missing!')
exit(1)
def checkPre(source):
# Make sure that all temporary files are gone
for root, dirs, _ in walkLevel(gettempdir(), 0):
for tempdir in dirs:
if tempdir.startswith('KCC-'):
rmtree(os.path.join(root, tempdir), True)
# Make sure that target directory is writable
if os.path.isdir(source):
src = os.path.abspath(os.path.join(source, '..'))
else:
src = os.path.dirname(source)
try:
with TemporaryFile(prefix='KCC-', dir=src):
pass
except:
raise UserWarning("Target directory is not writable.")
def makeBook(source, qtGUI=None):
global GUI
GUI = qtGUI
if GUI:
GUI.progressBarTick.emit('1')
else:
checkTools(source)
checkPre(source)
print("Preparing source images...")
path = getWorkFolder(source)
print("Checking images...")
getComicInfo(os.path.join(path, "OEBPS", "Images"), source)
detectCorruption(os.path.join(path, "OEBPS", "Images"), source)
if options.webtoon:
if image.ProfileData.Profiles[options.profile][1][1] > 1024:
y = 1024
else:
y = image.ProfileData.Profiles[options.profile][1][1]
comic2panel.main(['-y ' + str(y), '-i', '-m', path], qtGUI)
print("Processing images...")
if GUI:
GUI.progressBarTick.emit('Processing images')
imgDirectoryProcessing(os.path.join(path, "OEBPS", "Images"))
if GUI:
GUI.progressBarTick.emit('1')
chapterNames = sanitizeTree(os.path.join(path, 'OEBPS', 'Images'))
if 'Ko' in options.profile and options.format == 'CBZ':
sanitizeTreeKobo(os.path.join(path, 'OEBPS', 'Images'))
if options.batchsplit > 0:
tomes = splitDirectory(path)
else:
tomes = [path]
filepath = []
tomeNumber = 0
if GUI:
if options.format == 'CBZ':
GUI.progressBarTick.emit('Compressing CBZ files')
else:
GUI.progressBarTick.emit('Compressing EPUB files')
GUI.progressBarTick.emit(str(len(tomes) + 1))
GUI.progressBarTick.emit('tick')
options.baseTitle = options.title
options.covers = []
for tome in tomes:
options.uuid = str(uuid4())
if len(tomes) > 9:
tomeNumber += 1
options.title = options.baseTitle + ' [' + str(tomeNumber).zfill(2) + '/' + str(len(tomes)).zfill(2) + ']'
elif len(tomes) > 1:
tomeNumber += 1
options.title = options.baseTitle + ' [' + str(tomeNumber) + '/' + str(len(tomes)) + ']'
if options.format == 'CBZ':
print("Creating CBZ file...")
if len(tomes) > 1:
filepath.append(getOutputFilename(source, options.output, '.cbz', ' ' + str(tomeNumber)))
else:
filepath.append(getOutputFilename(source, options.output, '.cbz', ''))
makeZIP(tome + '_comic', os.path.join(tome, "OEBPS", "Images"))
else:
print("Creating EPUB file...")
buildEPUB(tome, chapterNames, tomeNumber)
if len(tomes) > 1:
filepath.append(getOutputFilename(source, options.output, '.epub', ' ' + str(tomeNumber)))
else:
filepath.append(getOutputFilename(source, options.output, '.epub', ''))
makeZIP(tome + '_comic', tome, True)
move(tome + '_comic.zip', filepath[-1])
rmtree(tome, True)
if GUI:
GUI.progressBarTick.emit('tick')
if not GUI and options.format == 'MOBI':
print("Creating MOBI files...")
work = []
for i in filepath:
work.append([i])
output = makeMOBI(work, GUI)
for errors in output:
if errors[0] != 0:
print('Error: KindleGen failed to create MOBI!')
print(errors)
return filepath
k = kindle.Kindle()
if k.path and k.coverSupport:
print("Kindle detected. Uploading covers...")
for i in filepath:
output = makeMOBIFix(i, options.covers[filepath.index(i)][1])
if not output[0]:
print('Error: Failed to tweak KindleGen output!')
return filepath
else:
os.remove(i.replace('.epub', '.mobi') + '_toclean')
if k.path and k.coverSupport:
options.covers[filepath.index(i)][0].saveToKindle(k, options.covers[filepath.index(i)][1])
return filepath
def makeMOBIFix(item, uuid):
os.remove(item)
mobiPath = item.replace('.epub', '.mobi')
move(mobiPath, mobiPath + '_toclean')
try:
dualmetafix.DualMobiMetaFix(mobiPath + '_toclean', mobiPath, bytes(uuid, 'UTF-8'))
return [True]
except Exception as err:
return [False, format(err)]
def makeMOBIWorkerTick(output):
makeMOBIWorkerOutput.append(output)
if output[0] != 0:
makeMOBIWorkerPool.terminate()
if GUI:
GUI.progressBarTick.emit('tick')
if not GUI.conversionAlive:
makeMOBIWorkerPool.terminate()
def makeMOBIWorker(item):
item = item[0]
kindlegenErrorCode = 0
kindlegenError = ''
try:
if os.path.getsize(item) < 629145600:
output = Popen('kindlegen -dont_append_source -locale en "' + item + '"',
stdout=PIPE, stderr=STDOUT, stdin=PIPE, shell=True)
for line in output.stdout:
line = line.decode('utf-8')
# ERROR: Generic error
if "Error(" in line:
kindlegenErrorCode = 1
kindlegenError = line
# ERROR: EPUB too big
if ":E23026:" in line:
kindlegenErrorCode = 23026
if kindlegenErrorCode > 0:
break
if ":I1036: Mobi file built successfully" in line:
output.terminate()
else:
# ERROR: EPUB too big
kindlegenErrorCode = 23026
return [kindlegenErrorCode, kindlegenError, item]
except Exception as err:
# ERROR: KCC unknown generic error
kindlegenErrorCode = 1
kindlegenError = format(err)
return [kindlegenErrorCode, kindlegenError, item]
def makeMOBI(work, qtGUI=None):
global GUI, makeMOBIWorkerPool, makeMOBIWorkerOutput
GUI = qtGUI
makeMOBIWorkerOutput = []
availableMemory = virtual_memory().total / 1000000000
if availableMemory <= 2:
threadNumber = 1
elif 2 < availableMemory <= 4:
threadNumber = 2
else:
threadNumber = 4
makeMOBIWorkerPool = Pool(threadNumber)
for i in work:
makeMOBIWorkerPool.apply_async(func=makeMOBIWorker, args=(i, ), callback=makeMOBIWorkerTick)
makeMOBIWorkerPool.close()
makeMOBIWorkerPool.join()
return makeMOBIWorkerOutput