Execute a command with a Bash variable in it and store the result

Question

I don't know how to execute a command with a variable and get the result of this. I have many .h and .c files and I need convert each from ISO-8859 to UTF-8.

So I make:

ls | grep "\.[ch]" | xargs myscript.sh

And in my script, the filename is in variable $1. Now I need to perform

iconv -f ISO-8859 -t UTF-8 $1

and store result of this, because iconv prints to stdout.

result=`iconv -f ISO-8859 -t UTF-8 $1`

echo $result

This seems to be not working, because it gives me some mismatch instead of converted $1.

"Some mismatch"? Please quote the exact error message you're getting. — hmakholm left over Monica, Nov 29 '11 at 22:40
It is not error. It is srabled content of more files then only $1, i dont understand that. — designerrr, Nov 29 '11 at 22:44
do any of the files emitted from you `ls | grep ..` have spaces in the filename? That will blow things up. Try changing to"$1" in your `result=...`, but you may need to come up with a way to handle spaces in file names. Please edit your question with approriate example values for $1 AND the exact text of scramble content. Good luck. — shellter, Nov 29 '11 at 23:15
If your files are small enough to capture, then you need to use `echo "$result"` with the double quotes to preserve the internal spacing. Without the double quotes, you get a single line of output with each sequence of one or more white spaces reduced to a single blank. — Jonathan Leffler, Nov 30 '11 at 20:44
The canonical is *[How do I set a variable to the output of a command in Bash?](https://stackoverflow.com/questions/4651437)*. Despite the unspecific title, it covers the case of ***(variable) input to the external command*** (in a (Bash) variable. — Peter Mortensen, Nov 10 '20 at 00:57

Michael · Accepted Answer · 2011-11-29T22:50:06.163

8

If you need to do some kind of transformation on the data first, you can "capture" output with the following syntax:

result="$(iconv -f ISO-8859 -t UTF-8 $1)"

There is a gotcha here as well: if you are going to be storing large amounts of data with potential whitespace or other meddlesome characters in it, be sure to always quote the variable ("$result" instead of $result) to ensure it gets treated as a single string.

edited Nov 29 '11 at 22:50

answered Nov 29 '11 at 22:44

Michael

880
2
6
16

This `$cmd $1 > $1` will not just overwrite the file, it will _clobber_ it **first**. This is a good way to truncate all your source files. – sehe Nov 29 '11 at 22:47
Fixed again. Man, my shell is rusty. – Michael Nov 29 '11 at 22:52
Ok, -1 removed :) **P.S.:** these are presumably C source files, so yes, onless it was the submission for an obfuscated code contest, there is _definitely_ going to be _some_ whitespace in there :) – sehe Nov 29 '11 at 22:53
Thank you man, it give me some mismatch because i dont quote $result when i echoes it. – designerrr Nov 29 '11 at 23:00

sehe · Answer 2 · 2011-11-29T22:48:27.290

3

I'd do as such:

while read filename; 
do
    mv "$filename" "$filename.bck" && \
        iconv -f ISO-8859 -t UTF-8 "$filename.bck" > "$filename"
done < find -iname '*.[hc]'

This creates backups on the fly and also handles files with whitespace (not newline characters).

edited Nov 29 '11 at 22:48

answered Nov 29 '11 at 22:43

sehe

374,641
47
450
633

Double quotes handle newlines in strings (file names). – Jonathan Leffler Nov 30 '11 at 17:20
1

@JonathanLeffler: I think sehe refers to `while read` when talking about newlines. IIRC, in bash you could do `while read -d $'\x00'`. – ninjalj Nov 30 '11 at 19:22
Fair enough - yes; `while read filename` won't handle newlines. – Jonathan Leffler Nov 30 '11 at 20:46

score 3 · Answer 3 · edited Nov 09 '20 at 23:50

Here is a solution that even handles newlines:

find -name '*.[ch]' \
    -exec mv '{}' '{}.backup' \; \
    -exec iconv -f ISO-8859 -t UTF-8 '{}.backup' -o '{}' \;

Generally, never parse filenames if you are going to use the results. The only sane ways I know of are

Use shell globs, e.g. for file in ./*.[ch] ; do echo "$file" ; done. Only works for one directory.
Use find in combination with -exec
Use 'find' in combination with -print0 (which prints the filenames as \0-separated strings) and use the output to build commandlines with xargs -0 and probably a helper script. This is quite cumbersome, though.

Also, make sure that relative filenames you use are prefixed with ./. Calling mv -from -to isn't safe, but mv ./-from ./-to is, and does what you want. E.g. when globbing, go with ./*.c rather than *.c.

Execute a command with a Bash variable in it and store the result

3 Answers3