simple command to strip header and footer from a file

Question

I want a simple little command to strip an XML-Header and Footer from a file:

<?xml version="1.0" encoding="UTF-8"?>
<conxml>
<MsgPain001>
    <HashValue>A9C72997C702A2F841B0EEEC3BD274DE1CB7BEA4B813E030D068CB853BCFECA6</HashValue>
    <HashAlgorithm>SHA256</HashAlgorithm>
    <Document>
                ...
    </Document>
    <Document>
                ...
    </Document>
</MsgPain001>
</conxml>

...

Should become just

<Document>
         ...
    </Document>
    <Document>
          ...
    </Document>

(note the indenting, the indent of the first document-tag should be stripped of.

This sounds like a (greedy) regex

<Document>.*</Document>

But I dont get it due to the linefeeds.

Could someone provide a simple sed script or similar to get it?

I need it in a pipe to compute a hash over the contained documents.

TIA!

Answer 1

Using sed:

 sed -n '/<Document>/,/<\/Document>/ p' yourfile.xml

Explanation:

• -n makes sed silent, meaning it does not output the whole file contents,
• /pattern/ searches for lines including specified pattern,
• a,b (the comma) tells sed to perform an action on the lines from a to b (where a and b get defined by matching the above patterns),
• p stands for print and is the action performed on the lines that matched the above.

---

Edit: If you'd like to additionally strip the whitespace before <Document>, it can be done this way:

 sed -ne '/ <Document>/s/^ *//' -e '/<Document>/,/<\/Document>/ p' yourfile.xml

Answer 2

To prevent text from being stripped between </Document> and the next <Document> you may have to use a series of sed commands (cf. Gilles' comment above).

Essentially sed reads the entire file into the hold buffer (so that the file contents can be treated as a single line) and marks the first and last Document tags for further processing.

# version 1
# marker: HERE
cat file.xml | 
sed -n '1h;1!H;${;g;s/\(<Document>.*<\/Document>\)/HERE\1HERE/g;p;}' | 
sed -n -e '/HERE<Document>/,/<\/Document>HERE/ p' | 
sed -e 's/^ *HERE\(<Document>\)/\1/' -e 's/\(<\/Document>\)HERE *$/\1/'

# version 2    (using the Bash shell)
# marker: $'\001'
cat file.xml | 
sed -n $'1h;1!H;${;g;s/\\(<Document>.*<\\/Document>\\)/\001\\1\001/g;p;}' | 
sed -n -e $'/\001<Document>/,/<\\/Document>\001/ p' | 
sed -e $'s/^ *\001//' -e $'s/\001 *$//' | 
cat -vet

... but I guess all this could be done more elegantly (& reliably) using xmlstarlet!