DEPRECATED

This project is deprecated.

XML Magritte Generator

With this library you can take a "nice" XML and generate Pharo class hierarchy with Magritte Descriptions for each variable. I use Magritte XML Bindings, which gives the hierarchy free XML import/export.

The generator will try to guess basic types (Boolean, Integer, Float, String), and the relationship between the parent and child elements (attribute, collection (oneToMany), reference (oneToOne)).

For example if you feed it "nice" XMLs like this one

<machine id="12">
	<version value="1.2.7" />
	<anyOnline>false</anyOnline>
	<nodes>
		<node id="node-1">
			<isOnline>false</isOnline>
			<address>
				<host>127.0.0.1</host>
				<port>80</port>
			</address>
		</node>
	</nodes>
</machine>

it will exhaust class hierarchy and populate it with the content

Unsupported

Since it is mostly magic, it doesn't always work, for example in the scenarios below. Note that it may fail in a variety of unforeseen circumstances, so feel free to open an issue.

Namespaces / namespaced names:

I didn't look much into how namespaces are actually propagated through the document; not just prefixing, but also NS inheriting. With consistent document the names can be easily mapped.

<xmi:element />
<element xml:attr="value" />

Elements with non-uniform repeating children:

Because I use the container element as variable name to hold child references.

This is ok:

<root>
  <container>
    <child />
    <child />
  </container>
</root>

This is not ok:

<root>
  <container>
    <child />
    <child />
    <otherThing />
  </container>
</root>

Cross-tree references

This is much easier for XMI, but for generic XML there's a lot of ambiguity. Maybe a type override specifying that source and target are #REF ids to /state-machien/states/state.

<state-machine>
	<states>
		<state id="10" />
		<state id="15" />
	</states>
	<transitions>
		<transition source="10" target="15" />
	</transitions>
<state-machine>

Usage

Use the generator along those lines:

dom is an instance of XMLDocument

Generator

gen := XOGStructureGenerator new.
"optional attributes, they have their default values"
gen packageName: 'MachineGenerated'.
gen classPrefix: 'XXX'. "to avoid name conflicts"
gen rootClassName: 'MXObject'. "a class from which all the classes will inherit"

"run the generator; this will NOT create the code yet"
gen processDocument: dom.

"retrieve CBChangesSet¹, so additional modifications can be performed"
changes := gen changes.

(CBChangesBrowser changes: changes refactoringChanges) open.

¹ Documentation on CBChangesSet is available in Changes Builder project

Connecting with XML nodes

NOTE that all the generated classes must contain this method:

descriptionContainer
	<magritteContainer>
	^ super descriptionContainer
		xmlElementName: self class xmlElementName;
		yourself

you can either:

• use 'MXObject' which implements it
• add the method to your own root class
• use TMXDescription trait
• do whatever you want

Reading / Writing

XML → Object

obj := XXXClassForRootElement new magritteDescription fromXmlNode: dom root.

Object → XML

obj magritteDescription toXmlDocument: obj.

Inference

The inferencers guess the element/attribute value type (String, Boolean, ...) as well as their type that relates them to their attributes and children (list, complex, inlined, ...).

types := XOGTypeClassification new classificationFor: dom.
valueTypes := XOGValueTypeInference new inferDocument: dom.
classification := XOGElementClassification new classifyTypes: types andValues: valueTypes.

Installation

Metacello new
    baseline: #XMLMAGenerator;
    repository: 'github://peteruhnak/xml-magritte-generator/repository';
    load

Dependencies:

BaselineOfXMLMAGenerator project latestVersion projects asArray collect: #name. "#('XMLParser' 'XPath' 'Magritte' 'MagritteXMLBindings' 'ChangesBuilder')"