[SOLVED] Bash, find : How to avoid [...] pattern matching in file names expanded from "$var"?

Telengard · 04-21-2011, 04:17 PM

Quote:

Originally Posted by grail

Code:

while read -r FILE
do
    find dir2 -type f -name "$FILE"
done< <(find dir1 -type f -printf "%f\n" | sed 's/[^[:alnum:]]/\\&/g')

It seems to work and *bonus* even works if files under dir1 are buried in a completely different hierarchy than files under dir2. It looks a little over-engineered though and I'm not sure all that complexity is really doing anything useful.

You're escaping all non-alphanumeric characters. I cannot think of a scenario where this would not work, but it goes further than necessary for my scenario.

Is it really necessary to use process substitution for gathering the file names from dir1? If so, then why?

I tried piping the dir1 file names into the loop and it seemed to work perfectly. When might piping not work?

Code:

foo$ find dir1 -type f -printf "%f\n" | sed 's/[^[:alnum:]]/\\&/g' |
> while read -r FILE
> do
> find dir2 -type f -name "$FILE"
> done
dir2/subdir2/file2
dir2/subdir2/file[a1]
dir2/subdir2/deepdir/messed[[up]name
dir2/subdir1/file1

Thanks very much for this example, although I still prefer

Quote:

Originally Posted by http://www.linuxquestions.org/questions/programming-9/bash-find-how-to-avoid-%5B-%5D-pattern-matching-in-file-names-expanded-from-%24var-875990/#post4331213

konsolebox's solution

because IMHO it is more graceful and easier to type.

grail · 04-21-2011, 11:53 PM

The main reason I prefer substitution over piping is that should you also require data from within the loop, ie. maybe you add a counter, it
will not be accessible from outside the loop due to the pipe creating a subshell, but in the simplest of cases it works just fine

Not sure I would agree with easier to type ... but I do like keeping it mostly all bash

Telengard · 04-22-2011, 04:15 PM

Quote:

Originally Posted by grail

should you also require data from within the loop, ie. maybe you add a counter, it will not be accessible from outside the loop due to the pipe creating a subshell,

Couldn't that be solved by marking the counter variable for export before creating the subshell? I guess maybe not because variable normally export parent to child (downstream) and not child to parent (upstream).

Code:

tmp$ export cmts=0
tmp$ awk '/^#/' ~/.bashrc |
> while read
> do
>   cmts=$((cmts + 1))
>   echo -n "$cmts "
> done ; echo -e "\n$cmts"
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
0
tmp$

Quote:

but I do like keeping it mostly all bash

Bash is a general purpose interpreter, so it probably is slower at specialized tasks than a dedicated program would be. For example, Bash can do what tr can do, but I'd be willing to bet that tr does it faster. At some point though, the expense of invoking additional processes within a loop will likely outweigh the benefit of using a specialized program.

grail · 04-23-2011, 01:27 AM

Quote:

For example, Bash can do what tr can do, but I'd be willing to bet that tr does it faster.

Not to be too nit picky, but this is probably not a good example as you need to first call bash builtin echo and bash now supports parameter substitution.

Code:

x="this is a test"

echo $( echo "$x" | tr ' ' '_' )

echo "${x// /_}"

But I would agree when looking at something like sed or awk in comparison to bash for getting data from a file efficiently.

Telengard · 04-23-2011, 02:36 AM

Quote:

Originally Posted by grail

Not to be too nit picky, but this is probably not a good example as you need to first call bash builtin echo and bash now supports parameter substitution.

Code:

x="this is a test"

echo $( echo "$x" | tr ' ' '_' )

echo "${x// /_}"

I don't want to go too far afield here, but I can't understand why you're making it so complex.

No need for parameter substitution. No need for echo.

Code:

foo$ x='this is a test'
foo$ tr ' ' '_' <<< "$x"
this_is_a_test
foo$

Beyond that, I'm pretty sure tr can be invoked without involving Bash whatsoever. There are other shells, and other command interpreters. I don't think program1 even needs Bash (or any other shell) to invoke program2 (although I may be wrong about that.)

Lately I've been thinking of invoking programs within AWK. I manipulate the output of programs with other programs. If I'm using AWK anyway, do I really need so many other programs?

I'm sure an AWK solution to the problem in this thread is possible.