Question on Bash parameter extension

DaneelGiskard · 01-10-2008, 10:17 AM

Hi all,

I'm trying to use the bash "parameter extension" mechanism to extract a substring which matches a certain pattern from a string.

So far I'm succesfully using the following in a bash script:

Code:

test='this is my example video with a resolution of 400x500 pixel.'
echo ${test//???x???/}

This will remove the '400x500' substring from the test string and output the rest.

What I want to achieve is exactly the opposite: I want to remove everything _except_ the '400x500' substring, thus using regular expressions to extract a specific substring from a string in a bash script.

Is such a "negation" possible in the above script fragment? Is there a better way to extract a substring which matches a specific regex pattern from a string in bash?

Many thanks,
Michael

Hko · 01-10-2008, 11:35 AM

Regular expressions are not supported that way in bash. It is probably possible to do it in pure bash, but it will become complex AFAICS.

Using sed it is quite easy, but not "pure bash" (if that is what you asking):

Code:

echo $test | sed -n 's/.* \([0-9]*x[0-9]*\) .*/\1/p'

Hko · 01-10-2008, 11:48 AM

Anyway, doing it "pure bash" is possible. But a bit more work.

Code:

test='this is my example video with a resolution of 400x500 pixel.'
head=${test%%x[0-9]*}
tail=${test##*[0-9]x}
X_res=${head##* }
Y_res=${tail%% *}
res_string=${X_res}x${Y_res}
echo $res_string

jozyba · 01-10-2008, 12:09 PM

Firstly, DaneelGiskard, your script doesn't work. It doesn't only strip out the '400x500' substring, it strips out any 'x' and the 3 characters either side of it. It should be something like:

Code:

echo ${test//[0-9][0-9][0-9]x[0-9][0-9][0-9]/}

I agree with Hko, string manipulation easily becomes over-complicated in bash, it's better to use tools that are designed for the job. Here's the simplest version I could come up with using pure bash:

Code:

x=${test//${test%%[0-9][0-9][0-9]x[0-9][0-9][0-9]*}/}; echo ${x:0:7}

Hko · 01-10-2008, 01:37 PM

Quote:

Originally Posted by jozyba

Here's the simplest version I could come up with using pure bash:

Code:

x=${test//${test%%[0-9][0-9][0-9]x[0-9][0-9][0-9]*}/}; echo ${x:0:7}

This does not work correctly if the resolution has different length, e.g. 1024x768 or 50x40.

Neither does the original code from DaneelGiskard by the way. It will not remove, say, "1024x768" entirely.

jozyba · 01-10-2008, 02:06 PM

Yes, you're right. This should work...

Code:

x=${test//${test%%[0-9]*x[0-9]*}/}; echo ${x// */}

DaneelGiskard · 01-14-2008, 06:15 AM

Many thanks, the sed version works beautifully for me.

All the best,
Michael