#!/bin/bash

# Generates a basic XML sitemap by collecting hyperlinks from the specified URL.

file='./sitemap.xml'
url='https://cerium.cc/'
includelastmod='false'



root=$(echo "$url" | sed 's/[a-z]*:\?\/\/\([0-9A-Za-z\.-]\+\)\/\?.*/\1/')
tempfile1='./sitemap-url-list.temp'
tempfile2='./sitemap-xml.temp'

cat > $file << EOF
<?xml version="1.0" encoding="UTF-8"?>
<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9">
EOF

curl -s $url > $tempfile1
# Isolates <a> tags
sed -i -En 's/(<\/a>)[^<>]*(<a )|(<a )|(<\/a>)/\1\4\n\2\3/gp' $tempfile1
sed -i -En 's/(<a.*)/\1/gp' $tempfile1
# Isolates value of href attribute
sed -i 's/">.*//' $tempfile1
sed -i 's/<a href="//' $tempfile1
# Removes download links
sed -i '/ download>/d' $tempfile1

# Removes links starting with a protocol
sed -i '/^#\|^[a-z]*:\/\//d' $tempfile1
# Removes links pointing to other domains, appends https protocol to root domain
sed -i "s/\/\/$root/https:\/\/$root/" $tempfile1
sed -i '/^\/\//d' $tempfile1
# Prepends root domain URL to relative path URL's
sed -i "s/^\//https:\/\/$root\//" $tempfile1

cat $tempfile1 > $tempfile2

# Adds XML formatting
sed -i 's/^/\t<url>\n\t\t<loc>/' $tempfile2
sed -i 's/\([^>]\)$/\1<\/loc>\n\t<\/url>/' $tempfile2

# Adds <lastmod> tag with ISO8601 format
if [ $includelastmod == 'true' ]; then
    pages=$(wc -l $tempfile1 | sed 's/\([0-9]\+\).*/\1/')
    for (( i = 1 ; i < $pages + 1 ; i++ )); do
        page=$(sed "${i}q;d" $tempfile1);
        lastmod=$(curl -sI $page | grep 'last-modified' | sed 's/last-modified: \(.*\)/\1/');
        isolastmod=$(date -d date -ud "$lastmod" +"%FT%T%:z");
        regexurl=$(echo $page | sed 's/\//\\\//g');
        sed -i "s/\($regexurl<\/loc>\)/\1\n\t\t<lastmod>$isolastmod<\/lastmod>/" $tempfile2;
    done
fi

cat $tempfile2 >> $file
rm $tempfile1 $tempfile2

echo '</urlset>' >> $file