sed inline not working to modify XML-style input

Running OP's current pipeline of sed scripts causes the contents of the input file (tracklist.txt) to be printed to stdout and then the pipeline hangs (ie, no other output, no return to a commnand prompt). I'm guessing this is what OP is referring to when stating it does not work ... ??

Primary issue: the input file (tracklist.txt) needs to be provided as an argument to the 1st sed script and not as an argument to the last sed script.

Recommentation:

# instead of this: sed 's/Documents//g' | ... | sed '/<file>/s|<file>/|<file>|' tracklist.txt ^^^^^^^^^^^^^ # do this: sed 's/Documents//g' tracklist.txt | ... | sed '/<file>/s|<file>/|<file>|' ^^^^^^^^^^^^^ 

Running this updated version of OP's sed pipeline generates:

<file>text1</file> <file>text2</file> <file>text3</file> 

While there are better tools for parsing HTML/XML, if OP must use sed then there are several ways to generate the same results but in a more efficient manner.

One idea requiring a single sed script:

sed -E 's|(<file>).*/([^/]+</file>)|\1\2|' tracklist.txt 

Where:

• -E - enables support for extended regexes
• (<file>) - (1st capture group) matches the string <file>
• ([^/]+</file>) - (2nd capture group) matches all characters that are not a / followed by the string </file>
• .*/ - everything between the two capture groups ending in a /
• \1\2 - replacement string consists of the two capture groups appended together
• NOTE: this works for the specific input provided by OP; it may need tweaking if the format of the input varies from what's shown in OP's sample input

For OP's sample input this generates:

<file>text1</file> <file>text2</file> <file>text3</file> 

Given your input XML file

Added a r root node:

<r> <file>Documents/time/text1</file> <file>Commun/text2</file> <file>Current/text3</file> </r> 

The code:

xidel --xquery ' <r>{ for $x in //file return <file>{tokenize($x, "/")[last()]}</file> }</r> ' --output-format=xml --output-node-indent file.xml 

Yields:

<?xml version="1.0" encoding="UTF-8"?> <r> <file>text1</file> <file>text2</file> <file>text3</file> </r> 

Explanations:

Here, instead of using the wrong tool: sed, I use XPath and XQuery (the former is a subset of the latter) proper XML parser.

xidel is the Swiss army knife for manipulating HTML/XML.

Usage:

xidel ... file.xml > new_file.xml 

If you want to edit on the fly:

xidel ... file.xml | sponge file.xml 

sponge from GNU more-utils.

Using Raku (formerly known as Perl_6)

~$ raku -MXML -e 'my $xml = open-xml( $*ARGFILES.Str ); for $xml.elements( :RECURSE(0), :TAG{"file"} ) -> $E { my $old = $E.contents[0]; my $new = XML::Text.new( text => $old.text.match(/ <?after "/"> <-[/]>+ $/) ); $E.replace( $old, $new ); }; .say for $xml;' file.xml 

OR:

% raku -MXML -e 'my $xml = open-xml( $*ARGFILES.Str ); for $xml.elements( :RECURSE(0), :TAG{"file"} ) -> $E { my $old = $E.contents[0]; my $new = XML::Text.new( text => $old.text.path.basename ); $E.replace( $old, $new ); }; .say for $xml;' file.xml 

Raku is a programming language in the Perl-family that features high-level Grammars for parsing text. Along with Raku/Rakudo itself, community members support modules in the Raku/Rakudo ecosystem. One of those modules is the (Raku-native) XML module.

Similar to the OP's other question, in Raku with the XML-module you can (for example) limit replacements to 1). the top-level and 2). within only the <file> TAG. This is done by setting the code to iterate through elements with the limitations :RECURSE(0), :TAG{"file"}. FYI, you can iterate through all TAGs at all depths if so desired: simply set :RECURSE(Inf) and remove the :TAG named-argument, which sets the :TAG restriction to False.

The first answer above identifies suitable TAGs/levels for replacement. Thus identified, each element's internal (i.e. non-TAG) contents[0] are assigned to the variable $old, which is actually an XML::Text object. The $old object is .text extracted into a string, and the desired match is found. A new (XML::Text.new) object is created ($new) with the now-corrected text => 'value' key/value pair. From here the XML-module's replace routine completes the job: replace( $old, $new ).

The second answer above is a clever twist on the first. Because the OP wants to edit path names, routines associated with Raku's IO::Path object class can be used. Raku's .IO routine understands the text as a valid path name, and Raku's .basename routine returns the final filename. This approach has the potential to increase code portability, because Raku has mechanisms for using the correct (/ or \ ) path-separator on different platforms.

Sample Input (thanks to @GillesQuénot!):

<r> <file>Documents/time/text1</file> <file>Commun/text2</file> <file>Current/text3</file> </r> 

Sample Output:

<?xml version="1.0"?><r> <file>text1</file> <file>text2</file> <file>text3</file> </r> 

https://github.com/raku-community-modules/XML
https://docs.raku.org/type/IO/Path
https://rakudo.org/
https://raku.org