gt-corpus-xmldiff.sh prev-converted/ new-converted/

gt-corpus-xmldiff.sh

#!/bin/sh

set -u

c14n_sort ()
{
    # The order of <parallel_text> elements does not matter. Assuming
    # input has been through c14n, we can just sort those lines.
    awk '
/<version>XSLtemplate/{next}
/<parallel_text/ { p[$0]++; seen++; next }
seen { a[NR]=$0; next }
{b[NR]=$0}
END{
    for(i in b)print b[i]
    n=asorti(p, sorted)
    for (i=1; i<=n; i++) {
        print sorted[i]
    }
    for(i in a)print a[i]
}
' | awk '
/<language/ { p[$0]++; seen++; next }
seen { a[NR]=$0; next }
{b[NR]=$0}
END{
    for(i in b)print b[i]
    n=asorti(p, sorted)
    for (i=1; i<=n; i++) {
        print sorted[i]
    }
    for(i in a)print a[i]
}
'
}

c14n_plus () {
    xmllint --c14n - | c14n_sort
}

trap 'rm -f "$c14a" "$c14b"' EXIT
c14a=$(mktemp -t xmldiff.XXXXXXXXXX)
c14b=$(mktemp -t xmldiff.XXXXXXXXXX)

c14n_diff () {
    c14n_plus < "$1" > "$c14a"
    c14n_plus < "$2" > "$c14b"
    diff -u -b "$c14a" "$c14b"
    #xmldiff -c "$c14a" "$c14b"
}

dir1=$1
dir2=$2
find "$dir1" -type f -print0 |\
    while read -r -d '' f; do
        #if [[ ! $f = *housing* ]]; then continue; fi
        m=$dir2/${f##${dir1}}
        if [[ -f $m && -f $f ]]; then
            if ! d=$(c14n_diff "$f" "$m"); then
                echo "$m"
                echo "$d"
                echo
            fi
        fi
    done