UNIX for Dummies Questions & Answers

I am very new to Unix, and am managing a digital asset management system. The GUI lets me do certain things, but I am finding that the real power is in the Solaris back-end. I have been asked to report how many and how much space the xml files in our repository are taking up. Is there a way to do this with a Unix command?

Thanks in advance for your help.

PF

That depends.. if they're in one folder, try this:

Code:

cd /some/dir
l *.xml | awk '{total += $5} END {print total}'

If they're in multiple files, try:

Code:

cd /some/dir
l `find . -name "*.xml" -print` | awk '{total += $5} END {print total}'

If there are a lot of files (more than your shell will allow on a command line), this will work. Actually, it works even if there aren't more files than are allowed, but that's a different matter ...

Code:

find . -name '*.xml' | xargs ls -l |
awk 'NF > 5 {t += $5 ; c++ }
     END { print c, "files take up", t, "bytes" }'

Some systems (like Irix, but there's probably a version that's installable for any system, like *BSD) come with a 'stat' program that will print just the size of a file with something like stat -qs file, in which case the awk bit above would sum on $1 instead of $5.

But your boss's question might mean, "what percentage of the disk is taken up by all these XML files?" This is a bit trickier, since ASCII files don't, in general, fill out the block they're in, and it depends on how many bytes belong to a given inode. F'rinstance, if the particular file system allocates in 2Kbyte segments, this is more useful:

Code:

find . -name '*.xml' |
    xargs du -k |
    awk '$1 % 2 { $1++ }
                { t += $1 ; c++ }
         END { print c, "files take up", t, "Kbytes"}'

du -k reports the file size in Kbytes, rouding up for partial Kbytes. The awk bit checks for partial 2Kbyte allocations and rounds them up because of the prior knowledge about the system. If your system allocates in 8Kbyte segments, just change $1 % 2 { $1++ } to m = $1 % 8 { $1 += 8 - m }. If you want bytes, multiply by 1024; if you want Kbytes, just print t as shown; and if you want Mbytes, divide by 1024 ...

Yes, you can assign to a variable in the pattern part of awk. And yes, in sh, ksh and zsh (and bash, I presume), you can break lines after the pipe symbol and it will do the right thing without an eol backslash, but don't try it in csh.