Linux / Cygwin：将模式替换为其他模式匹配的结果（sed / find？）

Question

I have a large website of documents that look like this:
<title>DOCTITLE</title>
<h1>Some Title</h1>

I'm trying to use Cygwin to replace DOCTITLE with Some Title in every file.

To be more specific, I need to extract whatever text is between <h1> tags from each file and replace the literal string "DOCTITLE" with the extracted text.

Here's one thought that doesn't work but illustrates the spirit of what I'm after:

find . -name "*html"  
       -exec sed -i 
                's/DOCTITLE/'$(grep "h1" | sed 's/<h1>\(.*\)<\/h1>/\1/')'/'
'{}' /;

Unsurprisingly, this fails because grep has no input and it would destroy the <h1> .

Any ideas?

Thanks for your time and expertise!

Answer 1

This might work for you (GNU sed):

find . -name "*html" -exec sed -i '$!N;s/DOCTITLE\([^\n]*\n<h1>\([^<]*\)<\/h1>\)/\2\1/;P;D' {}\;

This will need extensive testing first!

Answer 2

Your approach—using $( … ) —won't work, since sed 's -exec argument doesn't handle that syntax. What we can do instead, however, is call bash to do that bit of work for us:

find . -name '*.html' -exec /bin/bash -c 'sed "s/DOCTITLE/$(sed -n '\''\,<h1>.*</h1>,{s,<h1>\(.*\)</h1>,\1,p;q}'\'' '\''{}'\'')/" "{}"' \;

The outer sed does exactly what your sed command does. The inner $( … ) part is expanded by bash to produce only the text between the first <h1> s (it'd be much simpler if it didn't need to only get that first match).

Specifically, that inner sed prints nothing by default (the -n ), then for lines that match the regex <h1>.*</h1> , it runs s,<h1>\\(.*\\)</h1>,\\1,p;q , ie strip the HTML tags, print the result, then quit; that q ensures we only print out the first match.

Note I've avoided needing to use grep by using sed -n ; you could alteratively do the same thing with the below command, with the -m option to grep limiting the command to the first match.

find . -name '*.html' -exec /bin/bash -c 'sed "s/DOCTITLE/$(grep -m1 '\''<h1>.*</h1>'\'' '\''{}'\'' | sed '\''s,<h1>\(.*\)</h1>,\1,'\'')/" "{}"' \;

In both cases, there's some mildly horrific quoting going on: the '\\'' sequences are to insert a single quote into a single-quoted string. We need to quote the sed statements to ensure any spaces in the titles don't cause problems, and we need to quote the file names to be able to handle spaces in file names.