Add real last modified dates, priorities, and change frequencies to XML Sitemaps-generated sitemaps for Omeka.

fix_sitemap_for_omeka.rb

#--------------------------------------------------------------------------------------------------------------
# This is the script I use to add real dates, change frequencies, and priorities to Omeka items,     
# exhibits, and simple pages to sitemaps generated with the XML Sitemap tool.
# https://www.xml-sitemaps.com/standalone-google-sitemap-generator.html
#
# I use the following settins in order to keep the number of URLs down (only include simple pages, exhibits 
# and exhibit pages, exhibit items, and collection pages).                                                       
#                                                                                                    
# Exclude from sitemap extensions:                                                                   
# 	divx flv zip m4a m4v rar tar bz2 tgz exe gif tif jpg png class jar mpeg mpg mp3 wav mp4 avi wmv  
#	gz mov mid ra ram css pdf xml                                                                    
#                                                                                                    
# Exclude URLs:                                                                                      
# 	sort_field=                                                                                      
# 	page=                                                                                            
# 	search                                                                                           
# 	files/show                                                                                       
# 	items/browse?advanced%5B0%5D%5B                                                                  
# 	output                                                                                           
# 	?advanced%5                                                                                      
# 	?collection=                                                                                     
# 	items/show                                                                                       
#                                                                                                    
# Maximum depth level:                                                                               
# 	5                                                                                                
#                                                                                                    
# To get the correct dates, I create one CSV file with last modified dates from the items table      
# 	(id, modified) and one CSV file with last modified dates from the exhibits and                   
# 	simple_pages_pages tables (slug, modified)                                                       
#--------------------------------------------------------------------------------------------------------------

require 'csv'
require 'benchmark'
require 'rexml/document'
include REXML

#Create XML Document
xml_file = File.new("PATH TO YOUR SITEMAP")
xml_doc = Document.new(xml_file)

#Read CSV files, turn into arays
item_nums = CSV.read('PATH TO YOUR ITEM DATES').to_a
slugs = CSV.read('PATH TO YOUR SLUGS').to_a


#Set output file
output = "PATH TO YOUR DESIRED OUTPUT FILE"


#Gather URLs
nodes = XPath.match(xml_doc, "//url")

#Start count, start time, Benchmark for STDOUT progress
count = 0
report_time = 0

btime = Benchmark.realtime {

	#Go through all the nodes
	nodes.each do |node|
		node_time = Benchmark.realtime {
		loc = node.elements["loc"].text

		#Remove inaccurate GMT modifier from lasmod timestamp
		node.elements["lastmod"].text = node.elements["lastmod"].text.gsub("+00:00","")
		
		item_num = nil
		

		#Set conditions for different pages' priority and changefreq
		if loc.match('/item')
			#Get item number from URL, get lastmod date from array by that number
			item_num = loc.gsub(/(http:\/\/YOUR_URL\/exhibits\/show\/)(.*\/)(item\/)([0-9]{1,})/,"\\4")
			node.elements["lastmod"].text = item_nums.select { |a| a[0] == item_num }.flatten[1] 

			node.elements["priority"].text = "0.4"
			node.elements["changefreq"].text = "yearly"
		#Collections pages
		elsif loc.match('/collections/show')
			node.elements["priority"].text = "0.4"
			node.elements["changefreq"].text = "monthly"
		#Exhibit summary pages
		elsif loc =~ /http:\/\/YOUR_URL\/exhibits\/show\/[\w\-]*\z/
			node.elements["priority"].text = "0.8"
			node.elements["changefreq"].text = "yearly"
		else
			node.elements["priority"].text = "0.6"
			node.elements["changefreq"].text = "monthly"
		end

		#Add real lastmods for Exhibits and Simple Pages. This won't work so well if some of your slugs
		#are contained within other slugs. (e.g. "about" and "about-collection")--you'll have to 
		#add some conditionals.
		slugs.each do |slug|
			node.elements["lastmod"].text = slug[1] if loc.match("#{slug[0]}")
		end

		#Change URL for production
		node.elements["loc"].text = loc.gsub("DEV SUBDOMAIN","www")

	 	#Add correct GMT modifier. Sub your local modifier for -05:00	 	
        	node.elements["lastmod"].text = node.elements["lastmod"].text.concat("-05:00").gsub(" ","T") unless node.elements["lastmod"].text.nil?

        #End node Benchmark time
		}

		#Add node time to time passed, print progress if number of nodes processed is divisible by 100, increment counter.
		report_time += node_time
		puts "Proceesed #{count} records in #{report_time} seconds." if count.modulo(100).zero?
		count += 1
	end

	#Output changed XML document to a new file
	formatter = Formatters::Default.new
	File.open(output,"w") do |result| 
		formatter.write(xml_doc,result)
	end
#End Benchmark
}

#Say we're done.
puts "Done in #{btime} seconds"