New versions of create_root, new_child, etc with attributes and inner text

[315234]

This is a feature request to add methods for creating new nodes with an attribute dictionary, like this:

function create_root( xdoc::XMLDocument
                    , name::AbstractString
                    , attrs::Dict{ASCIIString,ASCIIString} )
    xroot = create_root(xdoc, name)
    set_attributes(xroot, attrs)
    xroot
end
function new_child( parent::XMLElement
                  , name::AbstractString
                  , attrs::Dict{ASCIIString,ASCIIString} )
    child = new_child(parent, name)
    set_attributes(child, attrs)
    child
end
function new_child( parent::XMLElement
                  , name::AbstractString
                  , attrs::Dict{ASCIIString,ASCIIString}
                  , text::ASCIIString )
    child = new_child2(parent, name, attrs)
    add_text(child, text)
    child
end

This would remove the need for additional calls to set_attributes or add_text and make the creation of new elements much more streamlined, for example:

xgeom = new_child(xgrid, "Geometry", Dict("Type"=>"VXVYVZ"))
xxdat = new_child(xgeom, "DataItem", Dict("Dimensions"=>"4 2", "Format"=>"HDF", "Precision"=>"8"), xspacedata)

[tkelman]

Could be additional methods using keyword arguments, I suppose. I'd look over a pull request.

[315234]

Would there be an issue with having them defined as methods with positional arguments, rather than keyword arguments, and letting dispatch choose the appropriate one?

I will try and submit a pull request if I can figure out how AbstractString and friends are actually supposed to work, and if I can get it working for both 0.3 and 0.4.

[tkelman]

The issue with positional arguments is whether the particular order is a clear API. Sometimes it's clearer to have named inputs. For example with attributes and text, what if you want to add text but not have any attributes? I also lean towards separating functionality out into smaller, more composable pieces, so I'm not sure what trying to put everything into the same list of arguments for a single function call accomplishes.

[nalimilan]

Wouldn't it make sense to pass all attributes as keyword arguments, like this?

function create_root(xdoc::XMLDocument,
                     name::AbstractString; # Notice the semicolon
                     attrs...)

@315234 See Compat.jl to get things working on both 0.3 and 0.4.

[tkelman]

That would only work for assigning to attributes whose names are valid Julia identifiers (so no keywords, no punctuation), but otherwise would be consistent with one of the signatures of set_attributes

[nalimilan]

Right. There could be an alternative method taking a Dict{AbstractString, AbstractString} for corner cases. (At least it could be added later if somebody requests it.)

[nalimilan]

Actually, doing f(symbol("/")=>1) works.