EPUB Thumbnailing support (#561)

* EPUB Thumbnailing support

---------

Signed-off-by: ed <s@ocv.me>
Co-authored-by: ed <s@ocv.me>

copyparty/__main__.py

@@ -1443,13 +1443,13 @@ def add_thumbnail(ap):
     # https://pillow.readthedocs.io/en/stable/handbook/image-file-formats.html
     # https://github.com/libvips/libvips
     # ffmpeg -hide_banner -demuxers | awk '/^ D  /{print$2}' | while IFS= read -r x; do ffmpeg -hide_banner -h demuxer=$x; done | grep -E '^Demuxer |extensions:'
-    ap2.add_argument("--th-r-pil", metavar="T,T", type=u, default="avif,avifs,blp,bmp,cbz,dcx,dds,dib,emf,eps,fits,flc,fli,fpx,gif,heic,heics,heif,heifs,icns,ico,im,j2p,j2k,jp2,jpeg,jpg,jpx,pbm,pcx,pgm,png,pnm,ppm,psd,qoi,sgi,spi,tga,tif,tiff,webp,wmf,xbm,xpm", help="image formats to decode using pillow")
+    ap2.add_argument("--th-r-pil", metavar="T,T", type=u, default="avif,avifs,blp,bmp,cbz,dcx,dds,dib,emf,eps,epub,fits,flc,fli,fpx,gif,heic,heics,heif,heifs,icns,ico,im,j2p,j2k,jp2,jpeg,jpg,jpx,pbm,pcx,pgm,png,pnm,ppm,psd,qoi,sgi,spi,tga,tif,tiff,webp,wmf,xbm,xpm", help="image formats to decode using pillow")
     ap2.add_argument("--th-r-vips", metavar="T,T", type=u, default="avif,exr,fit,fits,fts,gif,hdr,heic,jp2,jpeg,jpg,jpx,jxl,nii,pfm,pgm,png,ppm,svg,tif,tiff,webp", help="image formats to decode using pyvips")
-    ap2.add_argument("--th-r-ffi", metavar="T,T", type=u, default="apng,avif,avifs,bmp,cbz,dds,dib,fit,fits,fts,gif,hdr,heic,heics,heif,heifs,icns,ico,jp2,jpeg,jpg,jpx,jxl,pbm,pcx,pfm,pgm,png,pnm,ppm,psd,qoi,sgi,tga,tif,tiff,webp,xbm,xpm", help="image formats to decode using ffmpeg")
+    ap2.add_argument("--th-r-ffi", metavar="T,T", type=u, default="apng,avif,avifs,bmp,cbz,dds,dib,epub,fit,fits,fts,gif,hdr,heic,heics,heif,heifs,icns,ico,jp2,jpeg,jpg,jpx,jxl,pbm,pcx,pfm,pgm,png,pnm,ppm,psd,qoi,sgi,tga,tif,tiff,webp,xbm,xpm", help="image formats to decode using ffmpeg")
     ap2.add_argument("--th-r-ffv", metavar="T,T", type=u, default="3gp,asf,av1,avc,avi,flv,h264,h265,hevc,m4v,mjpeg,mjpg,mkv,mov,mp4,mpeg,mpeg2,mpegts,mpg,mpg2,mts,nut,ogm,ogv,rm,ts,vob,webm,wmv", help="video formats to decode using ffmpeg")
     ap2.add_argument("--th-r-ffa", metavar="T,T", type=u, default="aac,ac3,aif,aiff,alac,alaw,amr,apac,ape,au,bonk,dfpwm,dts,flac,gsm,ilbc,it,itgz,itxz,itz,m4a,mdgz,mdxz,mdz,mo3,mod,mp2,mp3,mpc,mptm,mt2,mulaw,oga,ogg,okt,opus,ra,s3m,s3gz,s3xz,s3z,tak,tta,ulaw,wav,wma,wv,xm,xmgz,xmxz,xmz,xpk", help="audio formats to decode using ffmpeg")
     ap2.add_argument("--th-spec-cnv", metavar="T", type=u, default="it,itgz,itxz,itz,mdgz,mdxz,mdz,mo3,mod,s3m,s3gz,s3xz,s3z,xm,xmgz,xmxz,xmz,xpk", help="audio formats which provoke https://trac.ffmpeg.org/ticket/10797 (huge ram usage for s3xmodit spectrograms)")
-    ap2.add_argument("--au-unpk", metavar="E=F.C", type=u, default="mdz=mod.zip, mdgz=mod.gz, mdxz=mod.xz, s3z=s3m.zip, s3gz=s3m.gz, s3xz=s3m.xz, xmz=xm.zip, xmgz=xm.gz, xmxz=xm.xz, itz=it.zip, itgz=it.gz, itxz=it.xz, cbz=jpg.cbz", help="audio/image formats to decompress before passing to ffmpeg")
+    ap2.add_argument("--au-unpk", metavar="E=F.C", type=u, default="mdz=mod.zip, mdgz=mod.gz, mdxz=mod.xz, s3z=s3m.zip, s3gz=s3m.gz, s3xz=s3m.xz, xmz=xm.zip, xmgz=xm.gz, xmxz=xm.xz, itz=it.zip, itgz=it.gz, itxz=it.xz, cbz=jpg.cbz, epub=jpg.epub", help="audio/image formats to decompress before passing to ffmpeg")
 
 
 def add_transcoding(ap):

copyparty/dxml.py

@@ -65,6 +65,9 @@ DXMLParser = _DXMLParser
 
 
 def parse_xml(txt: str) -> ET.Element:
+    """
+    Parse XML into an xml.etree.ElementTree.Element while defusing some unsafe parts.
+    """
     parser = DXMLParser()
     parser.feed(txt)
     return parser.close()  # type: ignore

copyparty/mtag.py

@@ -29,7 +29,7 @@ from .util import (
 )
 
 if True:  # pylint: disable=using-constant-test
-    from typing import Any, Optional, Union
+    from typing import IO, Any, Optional, Union
 
     from .util import NamedLogger, RootLogger
 
@@ -176,6 +176,9 @@ def au_unpk(
                 raise Exception("no images inside cbz")
             fi = zf.open(using)
 
+        elif pk == "epub":
+            fi = get_cover_from_epub(log, abspath)
+
         else:
             raise Exception("unknown compression %s" % (pk,))
 
@@ -365,6 +368,74 @@ def parse_ffprobe(txt: str) -> tuple[dict[str, tuple[int, Any]], dict[str, list[
     return zd, md
 
 
+def get_cover_from_epub(log: "NamedLogger", abspath: str) -> IO[bytes] | None:
+    import zipfile
+
+    from .dxml import parse_xml
+
+    try:
+        from urlparse import urljoin  # Python2
+    except ImportError:
+        from urllib.parse import urljoin  # Python3
+
+    with zipfile.ZipFile(abspath, "r") as z:
+        # First open the container file to find the package document (.opf file)
+        try:
+            container_root = parse_xml(z.read("META-INF/container.xml").decode())
+        except KeyError:
+            log("epub: no container file found in %s" % (abspath,))
+            return None
+
+        # https://www.w3.org/TR/epub-33/#sec-container.xml-rootfile-elem
+        container_ns = {"": "urn:oasis:names:tc:opendocument:xmlns:container"}
+        # One file could contain multiple package documents, default to the first one
+        rootfile_path = container_root.find("./rootfiles/rootfile", container_ns).get(
+            "full-path"
+        )
+
+        # Then open the first package document to find the path of the cover image
+        try:
+            package_root = parse_xml(z.read(rootfile_path).decode())
+        except KeyError:
+            log("epub: no package document found in %s" % (abspath,))
+            return None
+
+        # https://www.w3.org/TR/epub-33/#sec-package-doc
+        package_ns = {"": "http://www.idpf.org/2007/opf"}
+        # https://www.w3.org/TR/epub-33/#sec-cover-image
+        coverimage_path_node = package_root.find(
+            "./manifest/item[@properties='cover-image']", package_ns
+        )
+        if coverimage_path_node is not None:
+            coverimage_path = coverimage_path_node.get("href")
+        else:
+            # This might be an EPUB2 file, try the legacy way of specifying covers
+            coverimage_path = _get_cover_from_epub2(log, package_root, package_ns)
+
+        # This url is either absolute (in the .epub) or relative to the package document
+        adjusted_cover_path = urljoin(rootfile_path, coverimage_path)
+
+        return z.open(adjusted_cover_path)
+
+
+def _get_cover_from_epub2(log: "NamedLogger", package_root, package_ns) -> str | None:
+    # <meta name="cover" content="id-to-cover-image"> in <metadata>, then
+    # <item> in <manifest>
+    cover_id = package_root.find("./metadata/meta[@name='cover']", package_ns).get(
+        "content"
+    )
+
+    if not cover_id:
+        return None
+
+    for node in package_root.iterfind("./manifest/item", package_ns):
+        if node.get("id") == cover_id:
+            cover_path = node.get("href")
+            return cover_path
+
+    return None
+
+
 class MTag(object):
     def __init__(self, log_func: "RootLogger", args: argparse.Namespace) -> None:
         self.log_func = log_func