Implementing FeatureSource

Now with the setup out of the way we can get to work.

CSVDataStore

The first step is to create a basic DataStore that only supports feature extraction. We will read data from a CSV file into the GeoTools feature model.

../../_images/CSVDataStore.png

CSVDataStore

To implement a DataStore we will subclass ContentDataStore. This is a helpful base class for making new kinds of content available to GeoTools. The GeoTools library works with an interaction model very similar to a database - with transactions and locks. ContentDataStore is going to handle all of this for us - as long as we can teach it how to access our content.

ContentDataStore requires us to implement the following two methods:

  • createTypeNames()
  • createFeatureSource(ContentEntry entry)

The class ContentEntry is a bit of a scratch pad used to keep track of things for each type.

Our initial implementation will result in a read-only datastore for accessing CSV content:

  1. Set up a org.geotools.tutorial.csv package in src/main/java.

  2. To begin create the file CSVDataStore extending ContentDataStore

    package org.geotools.tutorial.csv;
    
    import com.csvreader.CsvReader;
    import java.io.File;
    import java.io.FileReader;
    import java.io.IOException;
    import java.io.Reader;
    import java.util.Collections;
    import java.util.List;
    import org.geotools.data.Query;
    import org.geotools.data.store.ContentDataStore;
    import org.geotools.data.store.ContentEntry;
    import org.geotools.data.store.ContentFeatureSource;
    import org.geotools.feature.NameImpl;
    import org.opengis.feature.type.Name;
    
    /**
     * DataStore for Comma Seperated Value (CSV) files.
     *
     * @author Jody Garnett (Boundless)
     */
    public class CSVDataStore extends ContentDataStore {
    
  3. We are going to be working with a single CSV file

        File file;
    
        public CSVDataStore(File file) {
            this.file = file;
        }
    
  4. Add the reader

        /**
         * Allow read access to file; for our package visible "friends". Please close the reader when
         * done.
         *
         * @return CsvReader for file
         */
        CsvReader read() throws IOException {
            Reader reader = new FileReader(file);
            CsvReader csvReader = new CsvReader(reader);
            return csvReader;
        }
    
  5. Listing TypeNames

    A DataStore may provide access to several different data products. The method createTypeNames provides a list of the information being published.

    After all that lead-in you will be disappointed to note that our list will be a single value - the name of the CSV file.

        protected List<Name> createTypeNames() throws IOException {
            String name = file.getName();
            name = name.substring(0, name.lastIndexOf('.'));
    
            Name typeName = new NameImpl(name);
            return Collections.singletonList(typeName);
        }
    
  6. Next we have the createFeatureSource method.

    This is used to create a FeatureSource which is used by client code to access content. There is no cache for FeatureSource instances as they are managed directly by client code. Don’t worry that is not as terrible as it sounds, we do track all the information and resources made available to a FeatureSource in a ContentEntry data structure. You can think of the FeatureSource instances sent out into the wild as light weight wrappers around ContentEntry.

    It is worth talking a little bit about ContentEntry which is passed into this method as a parameter. ContentEntry is used as a scratchpad holding all recorded information about the content we are working with.

    ../../_images/ContentEntry.png

    ContentEntry

    ContentEntry also contains a back pointer to the ContentDataStore in case your implementation of FeatureSource needs to phone home.

  7. Implement createFeatureSource. Technically the ContentEntry is provided as “parameter object” holding the type name requested by the user, and any other context known to the DataStore.

        @Override
        protected ContentFeatureSource createFeatureSource(ContentEntry entry) throws IOException {
            return new CSVFeatureSource(entry, Query.ALL);
        }
    

CSVFeatureSource

Next we can create the CSVFeatureSource mentioned above. This class is responsible for providing access to the contents of our CSVDataStore.

Note

The distinction between DataStore and FeatureSource can be difficult to demonstrate as our example consists of a single file. If it helps DataStore is an object representing the file, service or database. FeatureSource meanwhile represents the contents, data product, or table being published.

  1. Create the file CSVFeatureSource.

    package org.geotools.tutorial.csv;
    
    import com.csvreader.CsvReader;
    import java.io.IOException;
    import org.geotools.data.FeatureReader;
    import org.geotools.data.Query;
    import org.geotools.data.store.ContentEntry;
    import org.geotools.data.store.ContentFeatureSource;
    import org.geotools.feature.simple.SimpleFeatureTypeBuilder;
    import org.geotools.geometry.jts.ReferencedEnvelope;
    import org.geotools.referencing.crs.DefaultGeographicCRS;
    import org.locationtech.jts.geom.Point;
    import org.opengis.feature.simple.SimpleFeature;
    import org.opengis.feature.simple.SimpleFeatureType;
    import org.opengis.filter.Filter;
    
    /**
     * Read-only access to CSV File.
     *
     * @author Jody Garnett (Boundless)
     */
    public class CSVFeatureSource extends ContentFeatureSource {
    
        public CSVFeatureSource(ContentEntry entry, Query query) {
            super(entry, query);
        }
    
  2. To assist others we can type narrow our getDataStore() method to explicitly to return a CSVDataStore. In addition to being accurate, this prevents a lot of casts resulting in more readable code.

        /** Access parent CSVDataStore. */
        public CSVDataStore getDataStore() {
            return (CSVDataStore) super.getDataStore();
        }
    
  3. The method getReaderInternal( Query ) used to provide streaming access to out data - reading one feature at a time. The CSVFeatureReader returned is similar to an iterator, and is implemented in the next section.

        protected FeatureReader<SimpleFeatureType, SimpleFeature> getReaderInternal(Query query)
                throws IOException {
            return new CSVFeatureReader(getState(), query);
        }
    

    Note

    The DataStore interface provides a wide range of functionality for client code access feature content.

    Here at the implementation level we provide a single implementation of getReaderInternal. This method is used by the superclass ContentFeatureSource to access our content. All the additional functionality from filtering to transaction independence is implemented using a combination of wrappers and post-processing.

  4. ContentFeatureSource supports two common optimisations out of the box.

    You are required to implement the abstract method getCountInternal( Query ) using any tips or tricks available to return a count of available features. If there is no quick way to generate this information returning -1 indicates that they Query must be handled feature by feature.

    For CSV files we can check to see if the Query includes all features - in which case we can skip over the header and quickly count the number of lines in our file. This is much faster than reading and parsing each feature one at a time.

        protected int getCountInternal(Query query) throws IOException {
            if (query.getFilter() == Filter.INCLUDE) {
                CsvReader reader = getDataStore().read();
                try {
                    boolean connect = reader.readHeaders();
                    if (connect == false) {
                        throw new IOException("Unable to connect");
                    }
                    int count = 0;
                    while (reader.readRecord()) {
                        count += 1;
                    }
                    return count;
                } finally {
                    reader.close();
                }
            }
            return -1; // feature by feature scan required to count records
        }
    
  5. The second optimisation requires an implementation of getBoundsInternal(Query) making use of any spatial index, or header, record the data bounds. This value is used when rendering to determine the clipping area.

        /**
         * Implementation that generates the total bounds (many file formats record this information in
         * the header)
         */
        protected ReferencedEnvelope getBoundsInternal(Query query) throws IOException {
            return null; // feature by feature scan required to establish bounds
        }
    
  6. The next bit of work involves declaring what kind of information we have available.

    In database terms the schema for a table is defined by the columns and the order they are declared in.

    The FeatureType generated here is based on the CSV Header, along with a few educated guesses to recognise LAT and LON columns as comprising a single Location.

        protected SimpleFeatureType buildFeatureType() throws IOException {
    
            SimpleFeatureTypeBuilder builder = new SimpleFeatureTypeBuilder();
            builder.setName(entry.getName());
    
            // read headers
            CsvReader reader = getDataStore().read();
            try {
                boolean success = reader.readHeaders();
                if (success == false) {
                    throw new IOException("Header of CSV file not available");
                }
    
                // we are going to hard code a point location
                // columns like lat and lon will be gathered into a
                // Point called Location
                builder.setCRS(DefaultGeographicCRS.WGS84); // <- Coordinate reference system
                builder.add("Location", Point.class);
    
                for (String column : reader.getHeaders()) {
                    if ("lat".equalsIgnoreCase(column)) {
                        continue; // skip as it is part of Location
                    }
                    if ("lon".equalsIgnoreCase(column)) {
                        continue; // skip as it is part of Location
                    }
                    builder.add(column, String.class);
                }
    
                // build the type (it is immutable and cannot be modified)
                final SimpleFeatureType SCHEMA = builder.buildFeatureType();
                return SCHEMA;
            } finally {
                reader.close();
            }
        }
    

CSVFeatureReader

FeatureReader is similar to the Java Iterator construct, with the addition of FeatureType (and IOExceptions).

../../_images/CSVFeatureReader.png

CSVFeatureReader and Support Classes

The class ContentState is available to store any state required. Out of the box ContentState provides a cache of FeatureType, count and bounds. You are encouraged to create your own subclass of ContentState to track additional state - examples include security credentials or a database connection.

Note

Subclassing ContentState is a key improvement made for ContentDataStore. In our earlier base class we noticed many developers creating HashMaps to cache individual results in an effort to improve performance. Inevitability there would be difficulty keeping these caches in sync. Breaking out an object to handle the state required for data access is vast improvement.

FeatureReader interface:

  • FeatureReader.getFeatureType()
  • FeatureReader.next()
  • FeatureReader.hasNext()
  • FeatureReader.close()

To implement our FeatureReader, we will need to do several things: open a File and read through it line by line, parsing Features as we go. Because this class actually does some work, we are going to include a few more comments in the code to keep our heads on straight.

  1. Create the class CSVFeatureReader as follows:

    /* GeoTools - The Open Source Java GIS Toolkit
     * http://geotools.org
     *
     * (C) 2010-2014, Open Source Geospatial Foundation (OSGeo)
     *
     * This file is hereby placed into the Public Domain. This means anyone is
     * free to do whatever they wish with this file. Use it well and enjoy!
     */
    package org.geotools.tutorial.csv;
    
    import com.csvreader.CsvReader;
    import java.io.IOException;
    import java.util.NoSuchElementException;
    import org.geotools.data.FeatureReader;
    import org.geotools.data.Query;
    import org.geotools.data.store.ContentState;
    import org.geotools.feature.simple.SimpleFeatureBuilder;
    import org.geotools.geometry.jts.JTSFactoryFinder;
    import org.locationtech.jts.geom.Coordinate;
    import org.locationtech.jts.geom.GeometryFactory;
    import org.opengis.feature.IllegalAttributeException;
    import org.opengis.feature.simple.SimpleFeature;
    import org.opengis.feature.simple.SimpleFeatureType;
    
    public class CSVFeatureReader implements FeatureReader<SimpleFeatureType, SimpleFeature> {
    
        /** State used when reading file */
        protected ContentState state;
    
        /**
         * Current row number - used in the generation of FeatureId. TODO: Subclass ContentState to
         * track row
         */
        private int row;
    
        protected CsvReader reader;
    
        /** Utility class used to build features */
        protected SimpleFeatureBuilder builder;
    
        /** Factory class for geometry creation */
        private GeometryFactory geometryFactory;
    
        public CSVFeatureReader(ContentState contentState, Query query) throws IOException {
            this.state = contentState;
            CSVDataStore csv = (CSVDataStore) contentState.getEntry().getDataStore();
            reader = csv.read(); // this may throw an IOException if it could not connect
            boolean header = reader.readHeaders();
            if (!header) {
                throw new IOException("Unable to read csv header");
            }
            builder = new SimpleFeatureBuilder(state.getFeatureType());
            geometryFactory = JTSFactoryFinder.getGeometryFactory(null);
            row = 0;
        }
        /** Access FeatureType (documenting available attributes) */
        public SimpleFeatureType getFeatureType() {
            return (SimpleFeatureType) state.getFeatureType();
        }
    }
    
  2. Implement the iterator next() and hasNext() methods using a field to hold the value to return next.

        /** The next feature */
        private SimpleFeature next;
    
        /**
         * Access the next feature (if available).
         *
         * @return SimpleFeature read from property file
         * @throws IOException If problem encountered reading file
         * @throws IllegalAttributeException for invalid data
         * @throws NoSuchElementException If hasNext() indicates no more features are available
         */
        public SimpleFeature next()
                throws IOException, IllegalArgumentException, NoSuchElementException {
            SimpleFeature feature;
            if (next != null) {
                feature = next;
                next = null;
            } else {
                feature = readFeature();
            }
            return feature;
        }
        /**
         * Check if additional content is available.
         *
         * @return <code>true</code> if additional content is available
         * @throws IOException
         */
        public boolean hasNext() throws IOException {
            if (next != null) {
                return true;
            } else {
                next = readFeature(); // read next feature so we can check
                return next != null;
            }
        }
    

    Note

    The next() and hasNext() methods are allowed to throw IOExceptions making these methods easy to implement. Most client code will use this implementation behind a FeatureIterator wrapper that converts any problems to a RuntimeException. A classic easy of implementation vs easy of use tradeoff.

  3. Using the CSVReader library to parse the content saves a lot of work - and lets us focus on building features. The utility class FeatureBuilder gathers up state, employing a FeatureFactory on your behalf to construct each feature.

        /** Read a line of content from CSVReader and parse into values */
        SimpleFeature readFeature() throws IOException {
            if (reader == null) {
                throw new IOException("FeatureReader is closed; no additional features can be read");
            }
            boolean read = reader.readRecord(); // read the "next" record
            if (read == false) {
                close(); // automatic close to be nice
                return null; // no additional features are available
            }
            Coordinate coordinate = new Coordinate();
            for (String column : reader.getHeaders()) {
                String value = reader.get(column);
                if ("lat".equalsIgnoreCase(column)) {
                    coordinate.y = Double.valueOf(value.trim());
                } else if ("lon".equalsIgnoreCase(column)) {
                    coordinate.x = Double.valueOf(value.trim());
                } else {
                    builder.set(column, value);
                }
            }
            builder.set("Location", geometryFactory.createPoint(coordinate));
    
            return this.buildFeature();
        }
    
        /** Build feature using the current row number to generate FeatureId */
        protected SimpleFeature buildFeature() {
            row += 1;
            return builder.buildFeature(state.getEntry().getTypeName() + "." + row);
        }
    

    Note

    A key API contact is the construction of a unique FeatureID for each feature in the system. Our convention has been to prefix the typename ahead of any native identifier (in this case row number). Each FeatureID being unique is a consequence of following the OGC Feature Model used for Web Feature Server. These identifiers created here are employed in the generation of XML documents and need to follow the restrictions on XML identifiers.

  4. Finally we can close() the CSVFeatureReader when no longer used. Returning any system resources (in this case an open file handle).

        /**
         * Close the FeatureReader when not in use.
         *
         * @throws IOException
         */
        public void close() throws IOException {
            if (reader != null) {
                reader.close();
                reader = null;
            }
            builder = null;
            geometryFactory = null;
            next = null;
        }
    

    Note

    The FeatureState is not closed or disposed - as several threads may be making concurrent use of the CSVDataStore.

CSVDataStoreFactory

Now that we have implemented accessing and reading content what could possibly be left?

This is GeoTools so we need to wire in our new creation to the Factory Service Provider (SPI) plug-in system so that application can smoothly integrate our new creation.

To make your DataStore truly independent and pluggable, you must create a class implementing the DataStoreFactorySPI interface.

This allows the Service Provider Interface mechanism to dynamically plug in your new DataStore with no implementation knowledge. Code that uses the DataStoreFinder can just add the new DataStore to the classpath and it will work!

The DataStoreFactorySpi provides information on the Parameters required for construction. DataStoreFactoryFinder provides the ability to create DataStores representing existing information and the ability to create new physical storage.

  1. Implementing DataStoreFactorySPI:

    • The “no argument” constructor is required as it will be used by the Factory Service Provider (SPI) plug-in system.
    • getImplementationHints() is used to report on any “Hints” used for configuration by our factory. As an example our Factory could allow people to specify a specific FeatureFactory to use when creating a feature for each line.

    Create CSVDataStoreFactory as follows:

    /* GeoTools - The Open Source Java GIS Toolkit
     * http://geotools.org
     *
     * (C) 2010-2014, Open Source Geospatial Foundation (OSGeo)
     *
     * This file is hereby placed into the Public Domain. This means anyone is
     * free to do whatever they wish with this file. Use it well and enjoy!
     */
    // header start
    package org.geotools.tutorial.csv;
    
    import java.awt.RenderingHints.Key;
    import java.io.File;
    import java.io.IOException;
    import java.io.Serializable;
    import java.util.Collections;
    import java.util.Map;
    import org.geotools.data.DataStore;
    import org.geotools.data.DataStoreFactorySpi;
    import org.geotools.util.KVP;
    
    /**
     * Provide access to CSV Files.
     *
     * @source $URL$
     */
    public class CSVDataStoreFactory implements DataStoreFactorySpi {
        /**
         * Public "no argument" constructor called by Factory Service Provider (SPI) entry listed in
         * META-INF/services/org.geotools.data.DataStoreFactorySPI
         */
        public CSVDataStoreFactory() {}
    
        /** No implementation hints required at this time */
        public Map<Key, ?> getImplementationHints() {
            return Collections.emptyMap();
        }
    
  2. We have a couple of methods to describe the DataStore.

    This isAvailable method is interesting in that it can become a performance bottleneck if not implemented efficiently. DataStoreFactorySPI factories are all called when a user attempts to connect, only the factories marked as available are shortlisted for further interaction.

        public String getDisplayName() {
            return "CSV";
        }
    
        public String getDescription() {
            return "Comma delimited text file.";
        }
    
        /** Confirm DataStore availability, null if unknown */
        Boolean isAvailable = null;
    
        /**
         * Test to see if this DataStore is available, for example if it has all the appropriate
         * libraries to construct an instance.
         *
         * <p>This method is used for interactive applications, so as to not advertise support for
         * formats that will not function.
         *
         * @return <tt>true</tt> if and only if this factory is available to create DataStores.
         */
        public synchronized boolean isAvailable() {
            if (isAvailable == null) {
                try {
                    Class cvsReaderType = Class.forName("com.csvreader.CsvReader");
                    isAvailable = true;
                } catch (ClassNotFoundException e) {
                    isAvailable = false;
                }
            }
            return isAvailable;
        }
    
  3. The user is expected to supply a Map of connection parameters when creating a datastore.

    The allowable connection parameters are described using Param[]. Each Param describes a key used to store the value in the map, and the expected Java type for the value. Additional fields indicate if the value is required and if a default value is available.

    This array of parameters form an API contract used to drive the creation of user interfaces.

    The API contract is open ended (we cannot hope to guess all the options needed in the future). The helper class KVP provides an easy to use implementation of Map<String,Object>. The keys used here are formally defined as static constants - complete with javadoc describing their use. If several authors agree on a new hint it will be added to these static constants.

        /** Parameter description of information required to connect */
        public static final Param FILE_PARAM =
                new Param(
                        "file",
                        File.class,
                        "Comma seperated value file",
                        true,
                        null,
                        new KVP(Param.EXT, "csv"));
    
        public Param[] getParametersInfo() {
            return new Param[] {FILE_PARAM};
        }
    

    Note

    Does anything really use this?

    The gt-swing module is able to construct a user interface based on these Param descriptions. The uDig and GeoServer projects have similar auto-generated screens.

    ../../_images/shapeWizard1.png

    Shapefile User Parameters

    ../../_images/shapeWizard2.png

    Shapefile Advanced Parameters

  4. Next we have some code to check if a set of provided connection parameters can actually be used.

        /**
         * Works for csv file.
         *
         * @param params connection parameters
         * @return true for connection parameters indicating a csv file
         */
        public boolean canProcess(Map<String, Serializable> params) {
            try {
                File file = (File) FILE_PARAM.lookUp(params);
                if (file != null) {
                    return file.getPath().toLowerCase().endsWith(".csv");
                }
            } catch (IOException e) {
                // ignore as we are expected to return true or false
            }
            return false;
        }
    
  5. Armed with a map of connection parameters we can now create a Datastore for an existing csv file.

    Here is the code that finally calls our CSVDataStore constructor:

        public DataStore createDataStore(Map<String, Serializable> params) throws IOException {
            File file = (File) FILE_PARAM.lookUp(params);
            return new CSVDataStore(file);
        }
    
  6. How about creating a DataStore for a new csv file?

    Since initially our DataStore is read-only we will just throw an UnsupportedOperationException at this time.

        public DataStore createNewDataStore(Map<String, Serializable> params) throws IOException {
            throw new UnsupportedOperationException("CSV Datastore is read only");
        }
    
  1. The Factory Service Provider (SPI) system operates by looking at the META-INF/services folder and checking for implementations of DataStoreFactorySpi

    To “register” our CSVDataStoreFactory please create the following in src/main/resources/:

    • META-INF/services/org.geotools.data.DataStoreFactorySpi

    This file requires the filename of the factory that implements the DataStoreSpi interface.

    Fill in the following content for your org.geotools.data.DataStoreFactorySpi file:

    org.geotools.tutorial.csv.CSVDataStoreFactory
    

That is it, in the next section we will try out your new DataStore.