Use the S4 approach for reading and writing XML about eml HOT 6 CLOSED

This issue to be addressed once we close #34 (e.g. can generate the class definitions).

from eml.

Among other things, I think that the S4 approach sidesteps our difficulties with the API construction.

For instance, currently we define eml_dataset struggles to list its arguments as the arguments to it's child nodes and then uses some ugly do.call and if statements to check if a child node is included

eml_dataset <- function(..., .title = "Unnamed", .creator=list(), .contact=list(), .methods=list(), .coverage=list(), .rights = eml$get("default_rights")){
  dataset <- newXMLNode("dataset")
  title <- newXMLNode("title", .title)
  rights <- newXMLNode("intellectualRights", .rights)
  addChildren(dataset, title)

  creator <- do.call(eml_person, c(person_type="creator", .contact))
  addChildren(dataset, creator)
  contact <- do.call(eml_person, c(person_type="contact", .creator)) 
  addChildren(dataset, contact)

  if(!is.empty(.coverage)){
   coverage <- do.call(eml_coverage, .coverage)
   addChildren(dataset, coverage)
  }

  ## Methods node
  methods_node <- newXMLNode("methods", parent=dataset)
  if(!is.empty(.methods)){
   methods_node <- 
     do.call(eml_methods, 
             c(node = methods_node,
               .methods))
  }
....

All of this is sidestepped when this creature is instead an S4 object with children defined in slots:

setClass('eml_dataset', 
         representation(title = "XMLInternalElementNode",
                        creator = "XMLInternalElementNode",
                        rights = "XMLInternalElementNode",
                        methods = "XMLInternalElementNode",
                        datatable = "XMLInternalElementNode"))

Which is easy to extend. (Of course we don't have to write it manually, and the abstract node type can be replaced with detailed node definitions as they are written / once they are generated). @duncantl What then would the arguments to the eml_write or eml_dataset functions look like (to maintain a clean API)? Can we just map/match arguments to S4 slots?

from eml.

The XMLSchema package generates the constructor functions for the classes corresponding to eml_dataset, eml_write, etc. Yes, they provide a map from arguments to slots. They also get the optional slots and default values correct when they are specified in the schema.

from eml.

Why would the slots of eml_dataset be XMLInternalElementNode? The goal is to map the XML from an eml document to regular R objects and similarly, to map the R objects to an XML document that can be exported.

from eml.

Right, that was just me being lazy in the example. I think I comment above
that they would actually be the node classes (eg from XMLschema)

Carl Boettiger
http://carlboettiger.info

sent from mobile device; my apologies for any terseness or typos
On Jul 25, 2013 5:10 PM, "Duncan Temple Lang" [email protected]
wrote:

Why would the slots of eml_dataset be XMLInternalElementNode? The goal is
to map the XML from an eml document to regular R objects and similarly, to
map the R objects to an XML document that can be exported.

—
Reply to this email directly or view it on GitHubhttps://github.com/ropensci/reml/issues/38#issuecomment-21593718
.

from eml.

Ah, so we will get the constructor fns automatically too; cool. Will have
to see what they look like...

Carl Boettiger
http://carlboettiger.info

sent from mobile device; my apologies for any terseness or typos
On Jul 25, 2013 4:49 PM, "Duncan Temple Lang" [email protected]
wrote:

The XMLSchema package generates the constructor functions for the classes
corresponding to eml_dataset, eml_write, etc. Yes, they provide a map from
arguments to slots. They also get the optional slots and default values
correct when they are specified in the schema.

—
Reply to this email directly or view it on GitHubhttps://github.com/ropensci/reml/issues/38#issuecomment-21592928
.

from eml.