Reading from file works, but not stdin?

[David]

Just trying to get my feet wet playing with epubs; trying to use XMLStarlet to update the metadata file inside an epub.  Somehow, when I extract the metadata file to disk, then use that, it works.  But if I extract to stdout and pipe to XMLStarlet, it doesn't.  Clueless here …

xml sel  -N opf="http://www.idpf.org/2007/opf"    \
         -N dc="http://purl.org/dc/elements/1.1/" \
         -t -c '/opf:package/opf:metadata' content.opf

works and returns

<metadata xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:opf="http://www.idpf.org/2007/opf" xmlns:calibre="http://calibre.kovidgoyal.net/2009/metadata" xmlns="http://www.idpf.org/2007/opf">
        <dc:title>March to the Sea</dc:title>
        <dc:creator opf:role="aut" opf:file-as="Weber, David">David Weber</dc:creator><dc:creator opf:role="aut">John Ringo</dc:creator>
        <dc:contributor opf:role="bkp" opf:file-as="calibre">calibre (0.5.4) [http://calibre.kovidgoyal.net]</dc:contributor>
        <dc:identifier opf:scheme="calibre" id="calibre_id">f7ad94e9-85d2-4528-a8e8-5087bc03f3ca</dc:identifier>
        <dc:date>2001-08-01T00:00:00</dc:date>
        <dc:language>en-us</dc:language>
        <dc:publisher>Baen Books</dc:publisher>
        <dc:identifier opf:scheme="ISBN">0-671-31826-8</dc:identifier>
        <meta name="calibre:series_index" content="1"/>
        <dc:subject>Science Fiction</dc:subject>
    </metadata>

but

unzip -c -q test.epub |                              \
   xml sel  -N opf="http://www.idpf.org/2007/opf"    \
            -N dc="http://purl.org/dc/elements/1.1/" \
            -t -c '/opf:package/opf:metadata'

does not.  Instead, I just get the usage text.

- It's not a problem with the unzip output; if I replace the unzip command with 'cat content.opf' I get the same results. 
- If I have only 1 namespace declaration, it does work.
- Changing the order of the namespace declarations doesn't affect the result.

[David]

Correction: the second listing should read

unzip -c -q test.epub content.opf |                  \
   xml sel  -N opf="http://www.idpf.org/2007/opf"    \
            -N dc="http://purl.org/dc/elements/1.1/" \
            -t -c '/opf:package/opf:metadata'

to extract only the metadata file from the epub.

[Noam Postavsky]

I can't reproduce your problem, can you post the ouput from -version and the exact command that fails. Does it work if you use "-" (indicating stdin) as the file name?

[David]

OK, I don't know where those semicolons crept in.  The version is 1.0.1, as packaged for openSUSE 11.3.

The following command works:

xml sel -N opf="http://www.idpf.org/2007/opf"      \
        -N dc="http://purl.org/dc/elements/1.1/"   \
        -t -c '/opf:package/opf:metadata/dc:title' metadata.opf

But

cat metadata.opf | xml sel -N opf="http://www.idpf.org/2007/opf"      \
        -N dc="http://purl.org/dc/elements/1.1/"   \
        -t -c '/opf:package/opf:metadata/dc:title'

does not.  If I only define a single namespace, it works.  But then I can't select individual elements within /package/metadata:

cat metadata.opf | xml sel -N opf="http://www.idpf.org/2007/opf"      \
        -t -c '/opf:package/opf:metadata'

However, it turns out the multiple namespaces work when I explicitly set stdin as the file using "-" as suggested, thanks!  My amateur skills showing …

[David]

Why does it keep putting those semicolons in there?

[Noam Postavsky]

1.0.1 is pretty old, you might be hitting bug #1722425: many -N options on command line bug.

It seems like the forum adds semicolons every time it finds a url inside quotes, you can block it by putting the url inside a url bbcode tag.