Bash script- which code is better?

[blur xc]

If I have a script that might create a temporary text file, and when it's done displays the contents of the text file, if it exists. Which code would be better?


if [ -f file.txt ]; then
cat file.txt
fior

cat file.txt 2> /dev/nullI think the if statement makes prettier code, but passing the error to /dev/null if file.txt doesn't exist works too.

Thoughts?

Thanks,
BM

[howlingmadhowie]

i'd go for the first example because it's clearer what it means. i doubt it would be much slower than the second code either.

[kaibob]

I would use the following:


[ -f file ] && cat file

If the above can't be used, I would go with the if command.

[d3v1150m471c]

if [ -f file ]; then
cat file
else
echo "file does not exist"
fi

or


[ -f file] && cat file || echo "file does not exist"

[blur xc]

Well, I thought I'd do a benchmark the the results are more than surprising-

I wrote this script:

#!/bin/bash

for (( i = 1; i < 500; i++ )); do
# cat missingfiles.txt 2> /dev/null
if [ -f missingfile.txt ]; then
cat missingfile.txt
fi
done

and by commenting out / uncommenting the different lines of code I was able to see which is better and the 2> /dev/null is about 6000% slower to execute. In reality, running the command just once would yield an imperceptible difference, but it is good to see anyway-

the /dev/null bit:

time ./test1.sh

real 0m3.203s
user 0m0.956s
sys 0m2.128s

and the if statement:

time ./test1.sh

real 0m0.050s
user 0m0.012s
sys 0m0.012s

Thanks-
BM

[howlingmadhowie]

Well, I thought I'd do a benchmark the the results are more than surprising-

I wrote this script:

#!/bin/bash

for (( i = 1; i < 500; i++ )); do
# cat missingfiles.txt 2> /dev/null
if [ -f missingfile.txt ]; then
cat missingfile.txt
fi
done

and by commenting out / uncommenting the different lines of code I was able to see which is better and the 2> /dev/null is about 6000% slower to execute. In reality, running the command just once would yield an imperceptible difference, but it is good to see anyway-

the /dev/null bit:

time ./test1.sh

real 0m3.203s
user 0m0.956s
sys 0m2.128s

and the if statement:

time ./test1.sh

real 0m0.050s
user 0m0.012s
sys 0m0.012s

Thanks-
BM

gosh, that is actually rather interesting. i wouldn't have expected such a huge difference.

thinking about it, this could be because cat starts a new process with all the overhead this entails, while [ test ] is a shell built-in.

[falconindy]

If you're specifying Bash in the shebang, you really should be using Bash syntax, which is even faster since [[ is a keyword. Had to up the test size by a factor of 20 to get a meaningful difference in my results:


# Using untested cat
real 0m5.272s
user 0m1.801s
sys 0m3.034s

# Using [ -f missingfiles.txt ]
real 0m0.130s
user 0m0.121s
sys 0m0.008s

# Using [[ -f missingfiles.txt ]]
real 0m0.090s
user 0m0.081s
sys 0m0.008s

Also interesting is changing the for loop declaration to Bash syntax:

for i in {1..10000} do;
if [[ -f missingfiles.txt ]]; then
cat missingfiles.txt
fi
done

real 0m0.052s
user 0m0.045s
sys 0m0.007s

[cubeist]

If you're specifying Bash in the shebang, you really should be using Bash syntax, which is even faster since [[ is a keyword. Had to up the test size by a factor of 20 to get a meaningful difference in my results:


# Using untested cat
real 0m5.272s
user 0m1.801s
sys 0m3.034s

# Using [ -f missingfiles.txt ]
real 0m0.130s
user 0m0.121s
sys 0m0.008s

# Using [[ -f missingfiles.txt ]]
real 0m0.090s
user 0m0.081s
sys 0m0.008s

Also interesting is changing the for loop declaration to Bash syntax:

for i in {1..10000} do;
if [[ -f missingfiles.txt ]]; then
cat missingfiles.txt
fi
done

real 0m0.052s
user 0m0.045s
sys 0m0.007s

This is really useful! Can you tell us where you found this syntax reference? The last Bash book I bought (The bash cookbook by O'reilly) published last year doesn't use this syntax anywhere. Neither do any of my other books... is there a down side to using [[ over [ ?

[howlingmadhowie]

this page looks interesting for bash testing:
http://www.ibm.com/developerworks/library/l-bash-test.html

the basic differences seem to be:

with [ and ] you have to escape lots of stuff (for example brackets, < and > and *) so you end up writing stuff like:


[ian@pinguino ~]$ [ ! \( "a" = "$HOME" -o 3 -lt 4 \) ]; echo $?
1

which checks if a==$HOME or 3 is less than 4

with [[ and ]] you don't have to escape stuff so instead you can write:


[[ ( -d "$HOME" ) && ( -w "$HOME" ) ]]

which is a lot easier on the eyes.

(( and )) are used for arithmetic testing on integer values.

[falconindy]

This is really useful! Can you tell us where you found this syntax reference? The last Bash book I bought (The bash cookbook by O'reilly) published last year doesn't use this syntax anywhere. Neither do any of my other books... is there a down side to using [[ over [ ?
Not sure where I picked up the {x..y} syntax initially, but it's used all over. The BashFAQ (http://mywiki.wooledge.org/BashFAQ) is a phenomenal resource that no scripter should go without, lest their Bash-cred be questioned.

The only reason not to use double square brackets is portability, since it's only valid in Bash. They're faster and far more flexible as Howie has shown above. You're even allowed to use Bash regex inside [[ ]] comparisons.

[blur xc]

If you're specifying Bash in the shebang, you really should be using Bash syntax, which is even faster since [[ is a keyword. Had to up the test size by a factor of 20 to get a meaningful difference in my results:


# Using untested cat
real 0m5.272s
user 0m1.801s
sys 0m3.034s

# Using [ -f missingfiles.txt ]
real 0m0.130s
user 0m0.121s
sys 0m0.008s

# Using [[ -f missingfiles.txt ]]
real 0m0.090s
user 0m0.081s
sys 0m0.008sAlso interesting is changing the for loop declaration to Bash syntax:

for i in {1..10000} do;
if [[ -f missingfiles.txt ]]; then
cat missingfiles.txt
fi
done
real 0m0.052s
user 0m0.045s
sys 0m0.007s

Cool- thanks for that. I would have used bash syntax for the for loop, but just didn't know how. I'm still new at this, and I'm reading Wiley's Linux Command Line and Bash Scripting Bible May 2008 (long title) and I first ran across the C syntax for-loops and just went with it.

I first read about the {x..y} syntax in the Ubuntu Pocket Guide by Keir Thomas, of all places for stuff like "mkdir photos{1..10}" or "touch file{1..10}.txt".

BM