cvmfscatalogsize

#!/bin/bash
# Figure out the size of cvmfs catalogs in a repository based on the
# .cvmfsdirtab file and a 'find * ! -type d' in the top level directory.
# Rewritten by Dave Dykstra October 4, 2016
# BUG: does not support '!' non-match wildcards

usage()
{
    echo "cvmfscatalogsize cmvfsdirtabpath cvmfsfindpathfile" >&2
    echo "Either path may be '-' to read from stdin" >&2
    echo "Example: $ cd /cvmfs/<repo>" >&2
    echo "         $ find * ! -type d|cvmfscatalogsize .cvmfsdirtab -" >&2    
    exit 2
}

if [ $# != 2 ]; then
    usage
fi


# convert shell wildcards into regular expressions for awk's convenience
awk -v EXPRS="`sort $1|sed '/^!/d;s,^/,,;s,\.,\\.,g;s,\*,[^/]*,g;s,?,.,g'|tr '\n' ','`" '
BEGIN { 
    NUMAEXPRS=split(EXPRS,AEXPRS,",")
    NUMAEXPRS-- # ignores element after trailing newline
    TOT=0
}
{
    TOT++
    # count down to match the longest one first
    for (n=NUMAEXPRS; n > 0; n--) {
        EXPR=AEXPRS[n]
        MATCH=gensub("^(" EXPR ")/.*","\\1",1,$0)
	if (MATCH!=$0) {
	    if(COUNT[MATCH]=="")
		COUNT[MATCH]=1
	    else
		COUNT[MATCH]++
	    break
	}
    }
}
END {
    for (MATCH in COUNT) {
	print COUNT[MATCH],"\t/" MATCH
	TOT-=COUNT[MATCH]
    }
    print TOT,"\t/"
}
' $2 | sort -k 2