Bash – How to store the human-friendly size of a file in a variable

Yes, you can use find to look for non-executable files of the right size and then use file to check for ASCII. Something like:

find . -type f -size 1033c ! -executable -exec file {} + | grep ASCII

The question, however, isn't as simple as it sounds. 'Human readable' is a horribly vague term. Presumably, you mean text. OK, but what kind of text? Latin character ASCII only? Full Unicode? For example, consider these three files:

$ cat file1
abcde
$ cat file2
αβγδε
$ cat file3
abcde
αβγδε
$ cat file4
#!/bin/sh
echo foo

These are all text and human readable. Now, let's see what file makes of them:

$ file *
file1: ASCII text
file2: UTF-8 Unicode text
file3: UTF-8 Unicode text
file4: POSIX shell script, ASCII text executable

So, the find command above will only find file1 (for the sake of this example, let's imagine those files had 1033 characters). You could expand the find to look for the string text:

find . -type f -size 1033c ! -executable -exec file {} + | grep -w text

With the -w, grep will only print lines where text is found as a stand-alone word. That should be pretty close to what you want, but I can't guarantee that there is no other file type whose description might also include the string text.

There's no built-in tar option, but you can filter its output. For example, using humanize:

#!/usr/bin/env python

import fileinput
import humanize

for line in fileinput.input():
    (perm, owner, size, date, time, filename) = tuple(line.split())
    print '{0} {1} {2:>9} {3} {4} {5}'.format(perm, owner, humanize.naturalsize(size, gnu=True), date, time, filename)

Save this as e.g. humantvf, then

tar tvf ... | ./humantvf