[SOLVED] using sed and grep with a text string

boumphreyfr · 04-28-2011, 02:41 PM

Q: Is there any way to use grep and sed with a string variable rather than with a file?

The problem: Im running through a LARGE (about 10,000 lines) xhtml file and need to replace every instance of lines beginning <p>~

The following code works but takes a long time mainlly because an in/out operation needs to be carried out on each line. If I could read from a string rather than a file it would take a much shorter time!

Code:

#!/bin/bash 
count=$((0))
>tpf
echo "waiting....."

while read line; do
	echo $line>tpl
	echo $line>tpl
	if grep "<p>~" tpl >/dev/null 2>&1; then
		count=$((count+1))
		sed s/"<p>"/""/ tpl >tpl2;cp tpl2 tpl
		echo "<p style='font-size:1.4em; text-align: center; margin-top: 2em;' id='ch$count'>"$line >> tpf
		#echo $count$line
	else
		echo $line >> tpf
	fi
		
done<testbook

rm tpl
echo "done!! $count lines were changed"

I'm new to bash, and am trying to convert php and c programs to bash.

Any suggestions would be appreciated

Thanks,
Frank

crts · 04-28-2011, 03:42 PM

Quote:

Originally Posted by boumphreyfr

Q: Is there any way to use grep and sed with a string variable rather than with a file?

The problem: Im running through a LARGE (about 10,000 lines) xhtml file and need to replace every instance of lines beginning <p>~

The following code works but takes a long time mainlly because an in/out operation needs to be carried out on each line. If I could read from a string rather than a file it would take a much shorter time!

Code:

#!/bin/bash 
count=$((0))
>tpf
echo "waiting....."

while read line; do
	echo $line>tpl
	echo $line>tpl
	if grep "<p>~" tpl >/dev/null 2>&1; then
		count=$((count+1))
		sed s/"<p>"/""/ tpl >tpl2;cp tpl2 tpl
		echo "<p style='font-size:1.4em; text-align: center; margin-top: 2em;' id='ch$count'>"$line >> tpf
		#echo $count$line
	else
		echo $line >> tpf
	fi
		
done<testbook

rm tpl
echo "done!! $count lines were changed"

I'm new to bash, and am trying to convert php and c programs to bash.

Any suggestions would be appreciated

Thanks,
Frank

Hi,

first of all, you are checking for "<p>~" but your sed only removes "<p>". Do you really want to keep the "~"? Anyway, try this while-loop

Code:

while read line; do
	if [[ $line =~ ^\<p\>~ ]]; then
		echo "${line/<p>/<p style='font-size:1.4em; text-align: center; margin-top: 2em;' id='ch$count'>}" >> tpf
		#echo $count$line
	else
		echo $line >> tpf
	fi
		
done<testbook

Or how about just:

Code:

sed "/<p>~/ s/<p>/<p style='font-size:1.4em; text-align: center; margin-top: 2em;' id='ch$count'> " testbook > tpf

It might help if you post a small sample file and what you expect as output. I am having difficulties to understand what exactly you are trying to do.

cmnorton · 04-28-2011, 03:57 PM

I believe you are asking how grep and sed can work on the variable's contents. Here is a simple example of using grep with a test string. If you back tick your grep expression and assign it to a variable, you can test its contents.

Code:

TEST_STRING="This is a test."

RC=`echo $TEST_STRING | grep This`

if [ ! -z $? ]; then
    echo "This is contained in the test string."
fi

MTK358 · 04-29-2011, 09:07 AM

Quote:

Originally Posted by boumphreyfr

use grep and sed with a string variable rather than with a file?

Code:

echo "$variable" | sed '...'

Or if you want to save the output to another variable:

Code:

output=$(echo "$variable" | sed '...')

boumphreyfr · 04-29-2011, 04:26 PM

Thank you for all your replies, they taught me more than any book could!
When I entered:

Code:

#!/bin/bash 
	myvariable="Black Hat"
		echo $myvariable
		echo "$myvariable" | sed 's/Black/White/'
#stout should be 'White hat'

#enter result to new var and test
	mynewvariable=$(echo "$myvariable" | sed 's/Black/Red/')
	echo $mynewvariable
#stout should be 'Red hat'
exit

I got exactly the result I wanted!