writing bash script to check if file present on that partition

sumeet inani · 06-13-2013, 10:10 AM

Aim is to use basic commands found in all distros.
Currently I am using tiny core which uses busybox.
I am trying to see in root directory of every partition if they have a file named 'i-have-oxford-kundli'

Code:

mounted=$(df | grep -i 'sd'| awk '{ print $(NF) }') 
echo mountpoints=$mounted
for word in $mounted; do 
echo looking in=$word
found=$(ls -l $word| grep -i i-have-oxford-kundli)
if [ $? -eq 0 ]; then 
echo $word has the desired file
echo over
else
echo not found
fi
done

I notice that 'over' is echoed for every mountpoint .
I want to run a set of commands when if condition is true not otherwise.
How can I do that ?

suicidaleggroll · 06-13-2013, 10:14 AM

Instead of running an ls, piping it to grep, and then checking the exit status, why not just check for the existence of the file directly?

Code:

if [ -f $word/i-have-oxford-kundli ]

druuna · 06-13-2013, 10:42 AM

@sumeet inani: You might want to add the -P flag to the df command. This will make sure that all entries stay on the same line and aren't wrapped over 2 lines.

Your script basically does what it is supposed to do, which is the most important thing. With that in mind: It can be done more efficient. Have a look at this:

Code:

while read MOUNTED
do
  echo "Looking in: $MOUNTED"
  if [[ -a "$MOUNTED/i-have-oxford-kundli" ]]
  then
    echo "Found it."
  else
    echo "Not found."
  fi
done < <(df -P | awk '/sd/ { print $(NF) }')

suicidaleggroll already mentioned -f (is it a regular file), -a (file exists) can also be used.

Firerat · 06-13-2013, 05:47 PM

Code:

busybox awk '{if (!system("test -a "$2"/i-have-oxford-kundli")){print "i-have-oxford-kundli exists in "$2}}' /proc/mounts

But, it fails for /
This is because I have been very lazy and 'hardcoded' / as a prefix to the filename

building on it

Code:

for hasOK in $(busybox awk '{if (!system("test -a "$2"/i-have-oxford-kundli")){print $2}}' /proc/mounts);do
    HasOK="$HasOK $hasOK"
done

Will fail if you have white space in mountpoints

sumeet inani · 06-14-2013, 08:44 AM

to suicidaleggroll , fireat , druuna

I will try your alternatives . My question was that : the second statement following 'then' got executed even when condition evaluated to false . How can that happen ?

Madhu Desai · 06-14-2013, 09:06 AM

Quote:

Originally Posted by sumeet inani

My question was that : the second statement following 'then' got executed even when condition evaluated to false . How can that happen ?

PHP Code:



mounted=$(df | grep -i 'sd'| awk '{ print $(NF) }') 
echo mountpoints=$mounted

for word in $mounted ; do 
    echo looking in=$word
    found=$(ls -l $word| grep -i i-have-oxford-kundli)

    if [ $? -eq 0 ] ; then 
        echo $word has the desired file
        echo over
    else
        echo not found
    fi

done

Code:

$ ./kundli
mountpoints=/boot /dc /mnt/Win8 /mnt/Backup /media/Madhu_USB
looking in=/boot
not found
looking in=/dc
not found
looking in=/mnt/Win8
/mnt/Win8 has the desired file
over
looking in=/mnt/Backup
not found
looking in=/media/Madhu_USB
/media/Madhu_USB has the desired file
over

I see no wrong in the script. 'over' is only echoed on the partitions that has the 'i-have-oxford-kundli' file.

Edit: Just noticed that if you have LVM partitions, they dont show up by this script.

druuna · 06-14-2013, 09:16 AM

Quote:

Originally Posted by sumeet inani

My question was that : the second statement following 'then' got executed even when condition evaluated to false . How can that happen ?

You're code works when it comes to the echo over part. Don't know why it happens to you.

Quote:

Originally Posted by mddesai

Edit: Just noticed that if you have LVM partitions, they dont show up by this script.

That's why I suggested using the -P option (df -P instead of plain df). df uses a width of 80 by default which will wrap lines.

Madhu Desai · 06-14-2013, 09:25 AM

Quote:

Originally Posted by druuna

That's why I suggested using the -P option (df -P instead of plain df). df uses a width of 80 by default which will wrap lines.

-P option will put all lines in their seperate line, but since OP has filtered only 'sd' in grep, LVM partitions wont show up. instead OP can use

Code:

mounted=$(df -P | grep -i 'sd\|mapper' | awk '{ print $(NF) }')

druuna · 06-14-2013, 09:32 AM

Quote:

Originally Posted by mddesai

but since OP has filtered only 'sd' in grep, LVM partitions wont show up. instead OP can use

Code:

mounted=$(df -P | grep -i 'sd\|mapper' | awk '{ print $(NF) }')

Very valid point!

This would be more elegant and resource friendly:

Code:

df -P | awk '/(mapper|sd)/ { print $(NF)}'

Madhu Desai · 06-14-2013, 10:11 AM

Quote:

Originally Posted by druuna

Code:

df -P | awk '/(mapper|sd)/ { print $(NF)}'

enhancing your code a little, it can be further made to pattern-match only from first field, so that search wont pickup 'sd' or 'mapper' from any other field, just in case...

Code:

df -P | awk '$1~/(mapper|sd)/ { print $(NF)}'

...just a thought.

David the H. · 06-14-2013, 02:42 PM

grep is lighter than awk, so it might be better to do the line filtering with that, and do the field splitting with read instead.

Code:

while read -r _ _ _ _ _ MOUNTED ; do
    ...
done < <( df -P | grep -E '^(mapper|sd)'

In fact, since df doesn't generally produce that much input, I think it might actually be even faster to do all the filtering in the loop.

Code:

match='^(mapper|sd)'

while read -r fs _ _ _ _ MOUNTED ; do
    [[ $fs =~ $match ]] || continue 
    ...
done < <( df -P )

Edit: I just noticed this is being done in busybox. I imagine the advanced bash features I used above may not be available.