How to sort XML file using Ruby nokogiri - ruby-on-rails

Question

I want to sort this XML such that same type of demographics show first like all staty_type="REACH" appear on top, then all clicks and so on.
Here is an example object:
<?xml version="1.0"?>
<properties date="2020-06-23">
<property>
<order start="2020-06-23" end="2020-06-23">52658</order>
<demographics demographic="Age" stat_type="REACH">
<value category="18-24">36</value>
<value category="25-34">149</value>
</demographics>
<demographics demographic="Age" stat_type="CLICK">
<value category="18-24">6</value>
<value category="25-34">37</value>
</demographics>
<demographics demographic="Gender" stat_type="REACH">
<value category="female">402</value>
<value category="male">188</value>
</demographics>
<demographics demographic="Gender" stat_type="CLICK">
<value category="female">107</value>
<value category="male">44</value>
</demographics>
</property>
</properties>
I'm able to iterate XML. However, unable to perform sorting.
#doc = Nokogiri::XML(File.open("public/test.xml"))
builder = #doc.xpath("//property")
builder.search('./demographics').sort_by{|t| puts t['stat_type']}.each do |table|
puts table.to_s
end
I need the final XML in this form.
<?xml version="1.0"?>
<properties date="2020-06-23">
<property>
<order start="2020-06-23" end="2020-06-23">PBNI152658</order>
<demographics demographic="Age" stat_type="REACH">
<value category="18-24">36</value>
<value category="25-34">149</value>
</demographics>
<demographics demographic="Gender" stat_type="REACH">
<value category="female">402</value>
<value category="male">188</value>
</demographics>
<demographics demographic="Age" stat_type="CLICK">
<value category="18-24">6</value>
<value category="25-34">37</value>
</demographics>
<demographics demographic="Gender" stat_type="CLICK">
<value category="female">107</value>
<value category="male">44</value>
</demographics>
</property>
</properties>

Answer 1

When you do things like builder.search('./demographics') you just create a new nodeset with some nodes filtered from the initial XML document. Even if you sort this new nodeset you don't affect the initial document itself.
To sort the nodes of the initial document you have to rebuild the children of the node in question (<property> in your case). And here comes a tiny additional challenge - there are more nodes parsed by Nokogiri to take into account, not only the ones to sort:
pry(main)> #doc.at_xpath("//property").children.map(&:node_name)
=> ["text", "order", "text", "demographics", "text", "demographics", "text", "demographics", "text", "demographics", "text"]
So, what we have to do is to sort demographics nodes only and keep everything else untouched. One of the ways to do this is:
property_node = #doc.at_xpath("//property")
nodes_to_sort = property_node.children.dup
# My sorting logic is dumb here, apply your own as necessary
sorted_demographics = nodes_to_sort.select { |n| n.node_name == "demographics" }.sort_by { |n| n.attr("stat_type") }.reverse
# Create an empty nodeset. There should be a more idiomatic and readable way but this trick works too
new_nodeset = nodes_to_sort - nodes_to_sort
nodes_to_sort.each do |n|
case n.node_name
when "demographics"
new_nodeset << sorted_demographics.shift
else
new_nodeset << n
end
end
property_node.children = new_nodeset
And voila! - we are sorted now:
pry(main)> puts #doc
<?xml version="1.0"?>
<properties date="2020-06-23">
<property>
<order start="2020-06-23" end="2020-06-23">52658</order>
<demographics demographic="Gender" stat_type="REACH">
<value category="female">402</value>
<value category="male">188</value>
</demographics>
<demographics demographic="Age" stat_type="REACH">
<value category="18-24">36</value>
<value category="25-34">149</value>
</demographics>
<demographics demographic="Gender" stat_type="CLICK">
<value category="female">107</value>
<value category="male">44</value>
</demographics>
<demographics demographic="Age" stat_type="CLICK">
<value category="18-24">6</value>
<value category="25-34">37</value>
</demographics>
</property>
</properties>
NB. Take the solution above with a grain of salt - I don't know nokogiri's XML building capabilities well, so chances are there are some ways to achieve the same result with less code/in a more idiomatic way.