Explore moving to fast_xml completely

[oubiwann]

With ticket #15 we've needed to start using fast_xml. This ticket considers what it would take to move lxml entirely to fast_xml.

[oubiwann]

Test XML file:

<?xml version="1.0"?>
<customers>
   <customer id="55000">
      <name>Charter Group</name>
      <address>
         <street>100 Main</street>
         <city>Framingham</city>
         <state>MA</state>
         <zip>01701</zip>
      </address>
      <address>
         <street>720 Prospect</street>
         <city>Framingham</city>
         <state>MA</state>
         <zip>01701</zip>
      </address>
      <address>
         <street>120 Ridge</street>
         <state>MA</state>
         <zip>01760</zip>
      </address>
   </customer>
</customers>

As parsed by erlsom:simple_form:

#(ok
  #("customers"
    ()
    (#("customer"
       (#("id" "55000"))
       (#("name" () ("Charter Group"))
        #("address"
          ()
          (#("street" () ("100 Main"))
           #("city" () ("Framingham"))
           #("state" () ("MA"))
           #("zip" () ("01701"))))
        #("address"
          ()
          (#("street" () ("720 Prospect"))
           #("city" () ("Framingham"))
           #("state" () ("MA"))
           #("zip" () ("01701"))))
        #("address"
          ()
          (#("street" () ("120 Ridge"))
           #("state" () ("MA"))
           #("zip" () ("01760"))))))))
  "\n")

As parsed by `fast_xml:

#(xmlel
  #"customers"
  ()
  (#(xmlcdata #"\n   ")
   #(xmlel
     #"customer"
     (#(#"id" #"55000"))
     (#(xmlcdata #"\n      ")
      #(xmlel #"name" () (#(xmlcdata #"Charter Group")))
      #(xmlcdata #"\n      ")
      #(xmlel
        #"address"
        ()
        (#(xmlcdata #"\n         ")
         #(xmlel #"street" () (#(xmlcdata #"100 Main")))
         #(xmlcdata #"\n         ")
         #(xmlel #"city" () (#(xmlcdata #"Framingham")))
         #(xmlcdata #"\n         ")
         #(xmlel #"state" () (#(xmlcdata #"MA")))
         #(xmlcdata #"\n         ")
         #(xmlel #"zip" () (#(xmlcdata #"01701")))
         #(xmlcdata #"\n      ")))
      #(xmlcdata #"\n      ")
      #(xmlel
        #"address"
        ()
        (#(xmlcdata #"\n         ")
         #(xmlel #"street" () (#(xmlcdata #"720 Prospect")))
         #(xmlcdata #"\n         ")
         #(xmlel #"city" () (#(xmlcdata #"Framingham")))
         #(xmlcdata #"\n         ")
         #(xmlel #"state" () (#(xmlcdata #"MA")))
         #(xmlcdata #"\n         ")
         #(xmlel #"zip" () (#(xmlcdata #"01701")))
         #(xmlcdata #"\n      ")))
      #(xmlcdata #"\n      ")
      #(xmlel
        #"address"
        ()
        (#(xmlcdata #"\n         ")
         #(xmlel #"street" () (#(xmlcdata #"120 Ridge")))
         #(xmlcdata #"\n         ")
         #(xmlel #"state" () (#(xmlcdata #"MA")))
         #(xmlcdata #"\n         ")
         #(xmlel #"zip" () (#(xmlcdata #"01760")))
         #(xmlcdata #"\n      ")))
      #(xmlcdata #"\n   ")))
   #(xmlcdata #"\n")))

[oubiwann]

That's a little cumbersome for fast_xml ... but with some hacking and cleanup, I can get results like this:

#(customers
  #M()
  (#(customer
     #M(#"id" #"55000")
     (#(name #M() (#"Charter Group"))
      #(address
        #M()
        (#(street #M() (#"100 Main"))
         #(city #M() (#"Framingham"))
         #(state #M() ())
         #(zip #M() (#"01701"))))
      #(address
        #M()
        (#(street #M() (#"720 Prospect"))
         #(city #M() (#"Framingham"))
         #(state #M() ())
         #(zip #M() (#"01701"))))
      #(address
        #M()
        (#(street #M() (#"120 Ridge"))
         #(state #M() ())
         #(zip #M() (#"01760"))))))))

Note: this command was left in draft mode for a day or more ... since this hack, support has landed for this in lxml.

[oubiwann]

I think it's fair to say that once streaming support is added, we're on our way.

Probably 0.5.0 can be the release where erlsom is dropped and fast_xml is used exclusively.

[oubiwann]

Closing this as "successfully explored". Barring any insurmountably difficulties with streaming support, we'll be moving forward on a switch to fast_xml.