Skip to content

robinfriedli/JXP

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

95 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

JXP

Java XML Persistence API

Maven

    <dependency>
      <groupId>com.github.robinfriedli</groupId>
      <artifactId>JXP</artifactId>
      <version>2.0.5</version>
      <type>pom</type>
    </dependency>

    <repository>
        <name>jitpack.io</name>
        <url>https://jitpack.io</url>
    </repository>

Gradle

    dependencies {
        implementation "com.github.robinfriedli:JXP:2.0.5"
    }

    repositories {
        maven { url "https://jitpack.io" }
    }

Maven (1.2.1 or earlier)

    <repository>
      <id>JXP</id>
      <url>https://raw.github.com/robinfriedli/JXP/repository/</url>
    </repository>

    <dependency>
      <groupId>net.robinfriedli.JXP</groupId>
      <artifactId>JXP</artifactId>
      <version>1.2.1</version>
    </dependency>

Setup

Creating the JxpBackend object:

    JxpBackend jxp = new JxpBuilder()
        .addListeners(new MyEventListener(), new CountryListener())
        // map your XmlElement implementation to an XML tag name to have it instantiated automatically
        .mapClass("city", City.class)
        .mapClass("country", Country.class)
        .build();

Initializing a Context:

    // build a Context based on an existing XML file
    Context context = jxp.getContext("./resources/countries.xml");

    // build a Context based on a dom Document instance, this Context can later be saved to a file using Context#persist
    Document document = DocumentBuilderFactory.newInstance().newDocumentBuilder().newDocument();
    Element rootElem = document.createElement("tests");
    rootElem.setAttribute("xmlns", "testSpace");
    document.appendChild(rootElem);
    Context context = jxp.getContext(document);

AbstractXmlElement and BaseXmlElement

AbstractXmlElement is the class to extend for any type you want to persist as an XML element. Its default implementation, BaseXmlElement, is instantiated for each XML element in your file when creating a new CachedContext (as opposed to LazyContext (CachedContext is default), more about that later) via the StaticXmlElementFactory#instantiateAllElements method if no class is mapped to the corresponding XML tag name in the JxpBackend. It is recommended however that you create your own implementation for AbstractXmlElement which allows you to define a unique id for your element (BaseXmlElement has null as id), which is useful for loading XmlElements more easily and generally helps with managing and querying XML elements. Additionally it obviously allows you to add your methods. To make JXP instantiate your own class simply map your class with the matching tag name and add it to the JxpBackend using JxpBuilder#mapClass, you can also add additional classes later using StaticXmlElementFactory#mapClass.

When extending AbstractXmlElement you need to pass the tag name of your XML element plus optionally a Map with all attributes (with the attribute name as key and value as value), a List of sub elements and / or to the constructor of the super class. Your XmlElement should have two constructors; one to create a new XmlElement with the provided values and one to instantiate the XmlElement when loading it from the file, to which you pass the corresponding org.w3c.dom.Element

Example:

    package net.robinfriedli.jxp.api;

    import java.util.HashMap;
    import java.util.Map;

    import javax.annotation.Nullable;

    import net.robinfriedli.jxp.api.AbstractXmlElement;
    import net.robinfriedli.jxp.api.XmlElement;
    import net.robinfriedli.jxp.collections.NodeList;
    import net.robinfriedli.jxp.persist.Context;

    public class City extends AbstractXmlElement {

        public City(String name, int population) {
            super("city", buildAttributes(name, population));
        }

        // constructor invoked by StaticXmlElementFactory when instantiating a persistent XmlElement
        public City(Element element, NodeList subElements, Context context) {
            super(element, subElements, context);
        }

        @Nullable
        @Override
        public String getId() {
            return getAttribute("name").getValue();
        }

        private static Map<String, ?> buildAttributes(String name, int population) {
            Map<String, Object> attributes = new HashMap<>();
            attributes.put("name", name);
            attributes.put("population", population);
            return attributes;
        }
    }
    package net.robinfriedli.jxp.api;

    import java.util.HashMap;
    import java.util.List;
    import java.util.Map;

    import javax.annotation.Nullable;

    import com.google.common.collect.Lists;
    import net.robinfriedli.jxp.api.AbstractXmlElement;
    import net.robinfriedli.jxp.api.XmlElement;
    import net.robinfriedli.jxp.collections.NodeList;
    import net.robinfriedli.jxp.persist.Context;

    public class Country extends AbstractXmlElement {

        public Country(String name, String englishName, boolean sovereign, List<City> cities) {
            super("country", buildAttributes(name, englishName, sovereign), cities);
        }

        // constructor invoked by StaticXmlElementFactory when instantiating a persistent XmlElement
        public Country(Element element, NodeList subElements, Context context) {
            super(element, subElements, context);
        }

        @Nullable
        @Override
        public String getId() {
            return getAttribute("englishName").getValue();
        }

        public static Map<String, ?> buildAttributes(String name, String englishName, boolean sovereign) {
            Map<String, Object> attributes = new HashMap<>();
            attributes.put("name", name);
            attributes.put("englishName", englishName);
            attributes.put("sovereign", sovereign);
            return attributes;
        }
    }

Context, JxpBackend and BindableContext

The Context class represents the context of an XML document and / or File and is used to make any changes to it. It stores all XmlElement instances (in case of a CachedContext) of that document / file and applies changes to them by starting a Transaction, adding changes to it and then commit it using the invoke() method. This is required for any action that creates, deletes or changes an XmlElement. BindableContext is a Context that can be bound to any object and can be retrieved from the JxpBackend using that object. The JxpBackend holds all Context instances created by it, all BindableContexts and all EventListeners.

Using the invoke method:

parameter description
first parameter (optional): commit (boolean), set true if all changes made during this task should be committed to the XML file. If false the transaction will be added to the Context's uncommitted transactions. Default: true
second parameter (optional): instantApply (boolean). Defines whether all changes should be applied to the XmlElement instance upon adding it to the transaction immediately. That means all changes are available within the invoked task and not only after the transaction. E.g. if false changing an attribute using elem.setAttribute("test", "value") and then calling elem.getAttribute("test").getValue() will still return the old value if still inside the transaction
third parameter: the actual task to run, a Callable or Runnable depending on whether your task should return something or not

There are two other variants of the invoke method: invokeWithoutListeners() and futureInvoke(). futureInvoke, if used within a Transaction, adds a QueuedTask to the Transaction which will be executed after the Transaction has finished. This is necessary if the current Transaction is in a state where it is not recording changes anymore, e.g. in listeners. (For instant-apply transactions this is only required for the transactionCommitted event, otherwise no invoke method call is required at all). Alternatively, if there is no current Transaction, the method can still be used to invoke tasks asynchronously, in which case the implement has to execute the task at some point. futureInvoke has two additional boolean parameters after commit and instantApply: cancelOnFailure and triggerListeners. The first one is relevant when queueing the task to the transaction and defines whether the task gets cancelled if the transaction fails (rolls back). triggerListeners defines whether the task will be executed within a normal invoke or invokeWithoutListeners if false. The default option for both is true.

The quickest way to add a new XmlElement to the file:

    context.invoke(() -> new Country("Italia", "Italy", true, Lists.newArrayList(rome, florence, venice)).persist(context));

Example:

    Context context = contextManager.getContext();
    context.invoke(() -> {
        Country ch = new Country("Schweiz", "Switzerland", true, Lists.newArrayList(this));
        ch.persist(context);

        Country unitedKingdom = context.getElement("United Kingdom", Country.class);
        unitedKingdom.setAttribute("sovereign", true);

        Country france = context.getElement("France", Country.class);
        france.delete();
    });

Async example:

    QueuedTask<Void> future = context.futureInvoke(() -> {
        City london = new City("London", 8900000);
        context.requireElement("England").addSubElement(london);
        Thread.sleep(100);
        return null;
    });
    new Thread(future).start();
    long count1 = Query.evaluate(attribute("name").is("London")).execute(context.getElementsRecursive()).count();
    System.out.println("Count " + count1 + ", expected: 0");
    future.get();
    long count2 = Query.evaluate(attribute("name").is("London")).execute(context.getElementsRecursive()).count();
    System.out.println("Count " + count2 + ", expected: 1");

The parameters are used to build the mode the executed task will be wrapped in. For more advanced options you can create the mode yourself or even implement your own ModeWrapper to create additional modes the task may be invoked with.

Example:

Invoker.Mode mode = Invoker.Mode
    .create()
    .with(new ListenersMutedMode(context.getBackend()))
    .with(AbstractTransactionalMode.Builder.create().setInstantApply(false).writeToFile(false).build(context));
context.invoke(mode, () -> {
    City zurich = context.requireElement("Zurich", City.class);
    zurich.setAttribute("population", 410000);
});

LazyContext, sequential Transactions and XPath Queries

When working with huge XML files with millions of elements it might not be feasible to instantiate and store all XmlElements for the lifetime of the Context. LazyContext only stores XlmElement currently in state CONCEPTION or DELETION during the current Transaction until it's flushed. XmlElement instances are only instantiated upon calling Context#getElements and this each time it gets called, so you might need to manage storing them yourself when needed.

Making a large amount of changes or creating millions of new elements in one transaction might require using a sequential Transaction using the Context#invokeSequential method to make sure the Transaction does not eat up your entire ram. This type of Transaction will flush, i.e. commit and clear all changes, itself each time the amount changes reaches the provided threshold.

It is possible to only instantiate part of your document's elements if your ram does not allow storing all of them at any time using the Context#xPathQuery method, which will instantiate an XmlElement for each result it finds. The XQueryBuilder provides some convenient methods to generate an XPath query. In case of a CachedContext this method will not actually instantiate new XmlElements but match the results with the instances that are cached in the context but it's still faster to use the Queries described in the next section if the XmlElement instances are already stored somewhere (e.g. running a big XPath query over 6100000 elements takes up to 20 seconds while running a similar jxp query over a collection of XmlElements of the same size takes 12 milliseconds). Use the XConditions class to build nodes for the query. Note that JXP puts the tag name of the root element (so in this case "countries/") automatically, so when using XQueryBuilder#find you need to put the path starting from the first element after root.

XPath Query Examples:

XQuery query1 = XQueryBuilder.find("country").where(and(
        or(
                and(
                        attribute("sovereign").is(false),
                        attribute("englishName").startsWith("Eng"),
                        attribute("name").contains("ngl"),
                        attribute("name").endsWith("land")
                ),
                attribute("name").startsWith("United"),
                attribute("englishName").endsWith("land")
        ),
        not(
                and(
                        attribute("name").startsWith("United"),
                        attribute("name").endsWith("States")
                )
        )
));
String xPath1 = query1.getXPath(context);
XQuery query2 = XQueryBuilder.find("country/city").where(or(
        and(
                not(
                        and(
                                attribute("population").in(200000, 400000, 1000000, 8900000),
                                attribute("name").in("Birmingham", "New York")
                        )
                ),
                attribute("population").greaterThan(200000),
                attribute("population").lowerThan(1000000)
        ),
        and(
                or(
                        attribute("population").greaterThan(8000000),
                        attribute("population").lowerThan(1000000)
                ),
                attribute("name").in("London", "Manchester", "Birmingham"),
                not(
                        or(
                                attribute("name").in("New York"),
                                attribute("name").contains("ches"),
                                attribute("name").endsWith("ham")
                        )
                )
        )
));
String xPath2 = query2.getXPath(context);

Queries

The Conditions class offers many static methods to build predicates to find XmlElements with. Use Context#query to run a query over a context's elements (including all subelements recursively), which returns a ResultStream, or use the static method Query#evaluate to build a query and then run it over any Collection of XmlElements using the Query#execute method. Use ResultStream#count to get the amount of results a query returns or ResultStream#collect(Collector) to select what kind of Collection to collect the results with. You can also use Query#order (or later ResultStream#order) to order the results by an attribute or text content. Note that ResultStream operates on a Stream, meaning its methods (except #order and #getResultStream) terminate the stream and can only be invoked once. If you do need to invoke several of its methods for some reason you can collect it to a QueryResult<E, C> with ResultStream#getResult which offers similar methods but operates on the collected Collection of type C.

This type of Query works with cached XmlElement instances and is recommended with CachedContexts as it performs much better than XPath Queries. In case of LazyContext however the XmlElement instances need to be instantiated for each query (unless they are cached in a collection by the implementer and the Query is executed with this collection) while XPath queries run over the Context's document and only instantiate the XmlElements based on the Elements the query returns.

Examples:

    City london = context.query(
            and(
                    attribute("population").greaterThan(800000),
                    subElementOf(england),
                    instanceOf(City.class),
                    textContent().isEmpty()
            )
    , City.class).requireOnlyResult();
    Set<City> largeCities = Query.evaluate(attribute("population").greaterThan(8000000))
        .execute(context.getElementsRecursive(), City.class)
        .collect(Collectors.toSet());
    context.invoke(() -> {
        if (Query.evaluate(attribute("name").is("France")).execute(context.getElements()).count() == 0) {
            City paris = new City("Paris", 2000000);
            Country france = new Country("France", "France", true, Lists.newArrayList(paris));
            france.persist(context);
        }
        if (Query.evaluate(attribute("englishName").startsWith("Swe")).execute(context.getElements()).count() == 0) {
            City stockholm = new City("Stockholm", 950000);
            Country sweden = new Country("Sverige", "Sweden", true, Lists.newArrayList(stockholm));
            sweden.persist(context);
        }
    });
    XmlElement london = Query.evaluate(attribute("population").greaterEquals(400000))
            .order(Order.attribute("name"))
            .execute(context.getElementsRecursive())
            .requireFirstResult();

    XmlElement zurich = Query.evaluate(attribute("population").greaterEquals(400000))
            .order(Order.attribute("name", Order.Direction.DESCENDING))
            .execute(context.getElementsRecursive())
            .requireFirstResult();

Transaction

All changes made by a task are added to the Context's current transaction. At the end of the task all changes will be applied to the in memory XmlElements (if instantApply is false, else the changes are applied upon adding them to the transaction until the Transaction is not recording anymore, i.e. until Transaction#commit is called) and then, if commit is true, persisted to the XML document and written to the file. If commit is false the transaction will be added to the Context's uncommitted transactions. These can later be committed by calling Context#commitAll() or reverted by calling Context#revertAll(). If a commit fails all changes made by the invoked task will be reverted. If committing a change is known to fail (e.g. (before v0.7) because the affected XmlElement is duplicate) you can use Context#apply to create an apply-only Transaction to make sure the change never gets committed and commit the change manually using the XmlPersister. You can get the XmlPersister via context.getPersistenceManager().getXmlPersister()

JxpEventListener

The EventListener offers 3 methods that get fired when the corresponding Event gets applied plus 2 collecting events for when a Transaction gets applied or committed. If the transaction is an InstantApplyTx transactionApplied gets fired after the invoked task is done, however it can still apply changes made by listeners.

If you are using an InstantApplyTx you can simply make changes within the Listener without having to worry about using the invoke method, safe for the transactionCommitted event. Otherwise you need to use the Context#futureInvoke method to invoke your task in a new Transaction after the current one has finished since it is not recording and applying changes anymore.

 package net.robinfriedli.jxp.events;
 
 
 import java.util.EventListener;
 
 import net.robinfriedli.jxp.persist.SequentialTx;
 import net.robinfriedli.jxp.persist.Transaction;
 
 @SuppressWarnings("unused")
 public abstract class JxpEventListener implements EventListener {
 
     private final boolean mayInterrupt;
 
     protected JxpEventListener() {
         this(false);
     }
 
     protected JxpEventListener(boolean mayInterrupt) {
         this.mayInterrupt = mayInterrupt;
     }
 
     /**
      * Fired when an {@link ElementCreatedEvent} was applied
      *
      * @param event the ElementCreatedEvent
      */
     public void elementCreating(ElementCreatedEvent event) {
     }
 
     /**
      * Fired when an {@link ElementDeletingEvent} was applied
      *
      * @param event the ElementDeletingEvent
      */
     public void elementDeleting(ElementDeletingEvent event) {
     }
 
     /**
      * Fired when an {@link ElementChangingEvent} was applied
      *
      * @param event the ElementChangingEvent
      */
     public void elementChanging(ElementChangingEvent event) {
     }
 
     /**
      * Fired after a {@link Transaction} has been applied (to the in memory elements)
      *
      * @param transaction the transaction that has been applied
      */
     public void transactionApplied(Transaction transaction) {
     }
 
     /**
      * Fired after a {@link Transaction} has been committed (to the XML file). Note that this happens after the flush, so
      * you can't access the changes committed by this transaction anymore. If you need to access these changes see
      * {@link #onBeforeFlush(Transaction)}
      *
      * @param transaction the transaction that has been committed
      */
     public void transactionCommitted(Transaction transaction) {
     }
 
     /**
      * Fired before a {@link Transaction} is flushed. Usually happens during a commit or during a {@link SequentialTx}.
      *
      * @param transaction the transaction that has been flushed
      */
     public void onBeforeFlush(Transaction transaction) {
     }
 
     /**
      * the mayInterrupt parameter defines if an exception thrown by this listener should interrupt the transaction or
      * be ignored
      */
     public boolean mayInterrupt() {
         return mayInterrupt;
     }
 }

StringConverter

StringConverter is a static utility method to convert a String into any other class used by XmlAttribute#getValue and the ValueComparator for queries. You can extend to support more classes by using StringConverter#map

Thread safety

JXP offers thread safety by synchronising transactions and thus write access across all threads using a global instance of the net.robinfriedli.jxp.exec.MutexSync class. This class maps mutex objects to the canonical file path (for persistent contexts) or hashcode of the dom Document instance. Furthermore MutexSync automatically counts how many threads are using the mutex and drops unused mutexes to avoid leaking memory. Thread safety can only be established when using the propert synchronisation modes, all Context#invoke methods use synchronisation unless specified otherwise but when using a custom execution mode be sure to add the net.robinfriedli.jxp.exec.MutexSyncMode, this especially important to keep in mind when using Context#futureInvoke to run tasks in a separate thread as this method does not normally apply the MutexSyncMode because it is commonly used to queue tasks to run after the current transaction within the same thread thus assuming it is already running inside a synchronised parent task.

JXP often creates copies of internal collections before iterating for stability in a highly concurrent environment.