awk and newlines in the input data

Question

I want to find files in a directory and identify by their mimetype, not by the extension of the files.

I'am using this command to determine the mime type:

% find . -type f -print0 | xargs -0 -I{} file --mime-type {}
./foo
bar.png: image/png
./OWoHp.png: image/png
./J7ZwV.png: image/png
./foo.txt: inode/x-empty
./bar: inode/x-empty
./hUXnc.png: image/png

The first file has a newline in the filename:

% ls foo$'\n'bar.png
foo?bar.png

That's ok and the file should not be renamed.

With the next command I want to filter all files that are not images.

% find . -type f -print0 | xargs -0 -I{} file --mime-type {} | awk -F$"\0" -F": " '/image/ {print $1}'
bar.png
./OWoHp.png
./J7ZwV.png
./hUXnc.png

and identify their sizes:

% find . -type f -print0 | xargs -0 -I{} file --mime-type {} | awk -F$"\0" -F":" '/image/ {print $1}' | xargs -I{} identify -format "%[fx:w*h] %i\n" {}
identify: unable to open image `bar.png': No such file or directory @ error/blob.c/OpenBlob/2709.
identify: unable to open file `bar.png' @ error/png.c/ReadPNGImage/3922.
26696 ./OWoHp.png
47275 ./J7ZwV.png
37975 ./hUXnc.png

But that does not work because there is no file with the name bar.png. The correct name is

./foo
bar.png

with a newline in the name.

Answer 1

I think your best bet will be to use a shell loop instead of xargs: Then you can control how commands are sent the filename argument.

find . -type f -print0 | 
while IFS= read -rd "" filename; do
    type=$( file --brief "$filename" )
    if [[ $type == *image* ]]; then
        identify -format "%[fx:w*h] %i\n" "$filename"
    fi
done

Answer 2

You could use the -exec sh -c '...' construct with find:

find . -type f -exec sh -c 'file --brief --mime-type "$0" | \
grep -q ^image/ && identify -format "%[fx:w*h] %i\n" "$0"' {} \;

or with exiftool:

exiftool -q -if '$mimetype =~ /image/' -p '$megapixels $directory/$filename' -r .

Answer 3

As steeldriver pointed out, your problem is not awk, it's file. There is no NUL in the input you are giving to awk because file ate it. I would do this whole thing in the shell instead:

find . -type f -print0 | while IFS= read -r -d '' file; do 
    file --mime-type "$file" | grep -qP "\bimage/" && 
        printf '%s %s\0' $(identify -format '%[fx:w*h]' "$file") "$file";
done | sort -gz | tr '\0' '\n'
256 ./file 10
256 ./file 15
484 ./file 16
576 ./file 11
576 ./file 17
1024 ./file 12
1024 ./file 19
2304 ./file 13
5625 ./file 14
15190 ./file 2
15680 ./file 1
16384 ./file 9
65536 ./file 18
145200 ./file 0
183531 ./file 6
364807 ./file
3
364807 ./file 4
364807 ./file 5
388245 ./file 8
550560 ./file 7

I included sort since I assume you're trying to improve your answer here. The example above was run on file names with spaces and one (file\n3 with a newline). For some reason, identify won't print \0-terminated lines so I used printf instead.