sed match and ignore new lines

ted_chou12 · 02-24-2012, 06:53 PM

Hi, I am trying to make this match sed to ignore new lines:

Code:

torrentlink=($(sed -rn '/target=/s/.*href="([^"]+)".*>.*<\/a.*/\1/p' "$html"))

It would perfectly match contents like this:

Code:

<a style="color: white;" href="http://www.p2pnow.net" target="_blank">p2pnow.net</a>

p2pnow.net
but in html code that has new lines in it:

Code:

<a href="/topics/view/168577_Dymy_10_Fairy_Tail_23_BIG5_1024X576_RMVB.html"  target="_blank" >
				【Dy】【10】【<span class="keyword">Fairy</span> <span class="keyword">Tail</span>_】【】【BIG5】【1024X576】【<span class="keyword">RMVB</span>】</a>

This would not match anything. But I wish it to match:

【Dy】【10】【Fairy Tail_】【】【BIG5】【1024X576】【RMVB】
Thanks,
Ted

firstfire · 02-24-2012, 08:36 PM

Hi.

Sed is not good at multiline matches... You may try to remove all newlines from input or try the following

Code:

sed -rn '/target=/{:a; /<\/a>/!{N;ba}; s/.*href="([^"]+)".*>.*<\/a.*/\1/p}'

Bolded commands should append lines to pattern space until '</a>' is found.

Hope that helps.

ted_chou12 · 02-24-2012, 11:35 PM

Hi, thanks for your help. The command seems to return the href="value" rather than the content between <a>value</a>. The later one is the one i am searching for. I have tried to use tr to remove new lines before passing throught the command, but it seems to slow everything down.
Thanks,
Ted

firstfire · 02-25-2012, 12:22 AM

Hi.

Code:

$ cat infile.txt 
<a href="/topics/view/168577_Dymy_10_Fairy_Tail_23_BIG5_1024X576_RMVB.html"  target="_blank" >
【Dy】【10】【<span class="keyword">Fairy</span> <span class="keyword">Tail</span>_】【】【BIG5】【1024X576】【<span class="keyword">RMVB</span>】</a>
$ sed -rn '/target=/{:a; /<\/a>/!{N;ba}; s/\n//g; s/.*href="[^"]+"[^>]*>(.*)<\/a.*/\1/p}' infile.txt 
【Dy】【10】【<span class="keyword">Fairy</span> <span class="keyword">Tail</span>_】【】【BIG5】【1024X576】【<span class="keyword">RMVB</span>】

Modifications are in bold font. I also added a code to remove newlines.

ntubski · 02-25-2012, 08:37 AM

I would suggest not using regex to parse html. Here is a possible solution with xmlstarlet:

Code:

xml fo --html "$html" | xml sel -t -m '//a/node()' -c . 
# -T or --text will drop the <span>s
xml fo --html "$html" | xml sel -T -t -m '//a/node()' -c .

David the H. · 02-26-2012, 11:33 AM

As ntubski's link makes clear, the problem with html and xml is that their open, nested structures are inherently difficult to process with regular expressions and tools that are designed to work line-by-line such as awk and sed. For complex jobs you should really be using something with a dedicated parser.

That said, I believe it's still ok to use sed/awk/whatever for simple parsing on known, cleanly-structured html. But to do so you should start by using a tool like htmltidy to clean up and regularize the formatting, so that it can be more easily handled by these tools. Start by getting everything on one line before you try any extracting.