robust total size of a list of files
 

Hi,

There are different methods but what is the most robust method to make the total size of the files and/or directories from a list of files (`ls -1 `)?



thanks

du also shows total count of file sizes


check "-c" parameters in du man pages.

You really don't want to use ls for this. What's the exact
requirement of this exercise?

For a list of files, I'd use
but for complete directory trees, I'd use
If you want the output in millions of bytes, with one decimal digit, use printf("%.1f MB\n", s/1000000.0)

Remember that du lists disk usage, not file sizes.

@Nominal

du lists disk usage, but by whom?? ofcourse a file.. everything is file in linux. than what is the difference between size listed by du for a file and file size showing using stat command ?

Exactly. Observe:
This produces a sparse file. You can read it normally, but the initial zeros you see are not stored on disk, and do not consume disk space.

So, even though ls -l example-file outputs
and find example-file -printf '%s\n' and stat -c '%s' example-file both output
the actual disk usage is much less than the file size. My ext4 partition I used for this example utilizes 4k blocks, so du -hs example-file outputsDo you now see the difference between file sizes and disk usage?

Directories also consume disk space. The amount depends on the number of entries, the length of the file and directory names in it, and the number of POSIX and extended attributes used. If the du command you run includes the directories, the result will reflect the actual disk space used, and will be significantly different to the total size of the relevant files. You cannot even say which one (disk usage or total file size) is greater, unless you check!

Someone should also point out that this line will never work as expected seeing the while loop is in a sub-shell and so any changes to the variables inside the loop
will not be reported outside the loop.

That is correct. However, look at the -b and --apparent-size switches for du which will show the correct result.

What about?

cat * | wc -c

well typically to compare directories I do

but I am
a) lazy
b) a guru
c) lateral thinking

not sure which

Only if the parameter list does not refer to any directories. When usingthe result includes the size of the directories, not just the files. It is therefore not the total size of the files, as specified in the title of this thread; it yields the total size of the specified files and directories.

Isn't a directory also a file (everything's a file in unix)? By the way, the original question was (emphasis added):

if you don't want directories, you can use find . -type f or whatever.

There are enough differences that make the distinction meaningful.

Directory size is almost always an integral number of disk allocation blocks, making it useless for detecting e.g. additions or removals. (Note: Symlink size is almost always the length of the symlink target path.)

Directories cannot be hardlinked, only files can. (You can, however, bind-mount directories and files over other directories and files, so only the "top one" is accessible.)

Execute right to a directory allows you to stat() the directory and any file you know the name of in that directory. (Without read rights, you cannot scan the directory, so I guess read and write rights are analogous between files and directories.)

For mount points, /path/to/directory and /path/to/directory/. have different device and inode numbers.

thank you so so so much !!!!!

I like the --apparent-size way and too the find frmo unspawn