How to split a single XML files into multiple XML files based on a given tag and renamed based that tag attribute?

Question

I have several big archives in XML that I need to split the main node into files, and use the node's title attribute as name, eg:

<book title="ABC123" year="2000">
  <description>Some sentences...</description>
  <img src="image/cover_ABC123" />
</book>

And export it as ABC123.xml

I found a script that partially solves my help request, but it follows a numbered sequence and export files as 01.xml, 02.xml etc.; I would need to adapt it to my case, but I can't figure how:

(source: https://stackoverflow.com/a/56889282/17486393)

#!/usr/bin/env bash
xmlfile=file.xml

n=$(xmlstarlet sel -t -v 'count(//ORDER)' file.xml)
for i in $(seq 1 $n); do
   xmlstarlet sel -t -m "//ORDER[${i}]" -c . $xmlfile > "File${i}.xml"
done

I tried to add this option: -e "title" to extract also the name:

#!/usr/bin/env bash
xmlfile=file.xml

n=$(xmlstarlet sel -t -v 'count(//book -e "title")' file.xml)
for i in $(seq 1 $n); do
   xmlstarlet sel -t -m "//ORDER book -e "title"[${i}]" -c . $xmlfile > "File${i}.xml"
done

But I get:

xsl:for-each : could not compile select expression '//product -e title [1]

I tried to use this one instead, but I didn't understand this as well:

(https://stackoverflow.com/a/36156617/17486393)

$ for ((i=1; i<=`xmlstarlet sel -t -v 'count(/root/row)'  1.xml`; i++)); do \
          echo '<?xml version="1.0" encoding="UTF-8"?><root>' > NAME.xml;
          NAME=$(xmlstarlet sel -t -m '/root/row[position()='$i']' -v './NAME' 1.xml); \
          xmlstarlet sel -t -m '/root/row[position()='$i']' -c . -n 1.xml >> $NAME.xml; \
          echo '</root>' >> NAME.xml
       done

And I changed into:

$ for ((i=1; i<=`xmlstarlet sel -t -v 'count(/root/book)'  books01.xml`; i++)); do \
          NAME=$(xmlstarlet sel -t -m '/root/book[position()='$i']' -v './NAME' books01.xml); \
          xmlstarlet sel -t -m '/root/book[position()='$i']' -c . -n books01.xml >> $NAME.xml;
       done

It doesn't produce any files...

If possible I'd like to use xmlstarlet.

Answer 1

In XSLT 2.0 or later:

<xsl:transform version="2.0"
     xmlns:xsl="http://www.w3.org/1999/XSL/Transform"/>
      
   <xsl:template match="/">
      <xsl:for-each select="//book">
        <xsl:result-document href="{@title}.xml">
          <xsl:copy-of select="."/>
        </xsl:result-document>
      </xsl:for-each>
    </xsl:template>

</xsl:transform>

To execute this with SaxonJ from the command line (on one line):

java -jar dir/SaxonHE12-3J/saxon-he-12.3.jar 
  net.sf.saxon.Transform -s:input.xml -xsl:stylesheet.xsl 
  -o:out/output.xml -t

The resulting output.xml file will be essentially empty; the multiple files produced by xsl:result-document will be in the same directory as output.xml. The -t option logs each output file as it is written.

If you prefer a GUI tool, many popular XML editors have Saxon (or another XSLT 2.0 processor) integrated.

Answer 2

If possible I'd like to use xmlstarlet

xmlstarlet supports the EXSLT exsl:document element which is used to create multiple result documents in an existing directory, for example:

# shellcheck  shell=sh
cat <<'HERE' | xmlstarlet transform \
  /dev/stdin -s outDir="${TMPDIR}" input.xml
<xsl:transform version="1.0"
  xmlns:xsl="http://www.w3.org/1999/XSL/Transform" 
  xmlns:exsl="http://exslt.org/common"
  extension-element-prefixes="exsl"
>
  <xsl:param name="outDir" select="'/tmp'"/>
  <xsl:template match="/">
    <xsl:for-each select="//book">
      <exsl:document 
        href="{concat($outDir,'/',translate(@title,'/','_'),'.xml')}" 
        method="xml" 
        omit-xml-declaration="yes"
        indent="no"
      >
        <xsl:copy-of select="."/>
      </exsl:document>
    </xsl:for-each>
  </xsl:template>
</xsl:transform>
HERE

where

• xmlstarlet transform is an XSLT processor
• the XSLT stylesheet is read from stdin
• the output directory can be passed on the command line, e.g. ${TMPDIR} or file://${TMPDIR}
• any / (slash) characters in the book title are replaced with _ (underscore)
• if book titles are not unique only the last will survive
• the XPath 1.0 functions concat and translate are documented here and here

---

Or, if you're confident that lines in the input XML file are relatively short and therefore handled by standard text tools (hint: getconf LINE_MAX), you could have xmlstarlet select add a delimiter line before each XML <book> section and use awk to split the output into separate files:

# shellcheck  shell=sh
delim=$(uuidgen)
xmlstarlet select -t \
  -m '//book' \
    -o "${delim}$(printf '\t')${TMPDIR}/" -v 'translate(@title,"/","_")' -o '.xml' -n \
    -c '.' -n \
input.xml |
awk -F'[\t\n]' -v sep="${delim}" '$1==sep{close(out);out=$2;next;}{print>out;}'