Bash : rewriting a file which is read in same line

Question

I'm working on a bash script for partially automating router connection in command line. This is part where I have a problem:

#!/bin/bash
STRING2="ESSID"
MYPWD=$(pwd)
sudo iwlist wlan1 scan | grep $STRING2 | grep -n $STRING2 > $STRING2.txt
sed -r -e 's/\s+//g' -e 's/:ESSID//' $MYPWD"/"$STRING2".txt

The last line opens a text file, then deletes white spaces and also :ESSID's from the text file which is $MYPWD/$STRING2.txt. What I want to do is also write the output of sed to that file by piping. A previous approach was piping grep : > $MYPWD"/"$STRING2".txt" at the end such as :

sed -r -e 's/\s+//g' -e 's/:ESSID//' $MYPWD"/"$STRING2".txt | grep : > $MYPWD"/"$STRING2".txt"

But it ended up with an empty $MYPWD/$STRING2.txt. I'm not sure why this is happening. Any ideas?

Edit

Without the last piped grep, text file contains some text like this after execution:

1:___________ESSID:"somessid"

2:___________ESSID:"somessid"

3:___________ESSID:"somessid"

4:___________ESSID:"somessid"

ps: <__> stands for white space

#!/bin/bash
#declaring a string variable
STRING="MERHABA DUNYA"
STRING2="ESSID"
COUNTER="1"
MYPWD=$(pwd)
echo -e "Later you can check the recent networks around you from : " $MYPWD"/"$STRING2".txt\n"
#or MYPWD=${pwd}
#-e flag enables \n escape
echo -e $STRING2".txt is being updated\n. . .\n"
sudo iwlist wlan1 scan | grep $STRING2 | grep -n $STRING2 > $STRING2.txt # | sed -i# #s/ //g $STRING2.txt > $STRING2.txt
sed -r -e 's/\s+//g' -e 's/:ESSID//' $MYPWD"/"$STRING2".txt"
NUMLINE=$(sudo cat $MYPWD/$STRING2.txt | wc -l)
echo -e "\n. . ."
echo -e "\nThere are "$NUMLINE " visible networks\n"

Answer 1

When you have a command such as this

sed -e ... "datafile" >"datafile"

You will end up with a zero length result. The reason is that the shell sets up stdin and stdout before it executes the command. So stdout is sent to the file datafile, creating or truncating it in the process, and only then is the sed -e ... "datafile" run.

A common solution is something like this

sed -e ... "datafile" >"datafile.tmp" && mv -f "datafile.tmp" "datafile"

Some commands (including versions of sed) have the edit in place option:

sed -i.bak -e ... "datafile"
sed --in-place=.bak -e ... "datafile"