Requirements

  • Java 8 or later is required

Overview

A number of API enhancements have been made to take advantage of evolution in the Java language, for example try-with-resource and foreach constructs. The use of these make code simpler and more reliable.

Deprecated classes and methods have been removed, and the module structure and third-party jar use has been improved.

Java WebStart has been deprecated as of Java 9. As such, we no longer utilize WebStart.

Quick Navigation

netCDF-Java API Changes (5.0.x)

Point release notes:

Unsigned Types

  • DataType now has unsigned types: UBYTE, USHORT, UINT, ULONG
  • Array, ArrayScalar, ArrayByte, ArrayInt, ArrayShort, ArrayLong factory and constructor methods now require isUnsigned parameter.
  • Array.factory(class, shape) → Array.factory(DataType, shape) or Array.factory(dataType, class, shape)
  • Array.get1DJavaArray(Class) → Array.get1DJavaArray(DataType) or Array.get1DJavaArray(Class, isUnsigned)
  • Remove Array.setUnsigned(), Variable.setUnsigned()
  • Variable.getUnsigned() → Variable.getDataType().getUnsigned()
  • new Attribute(String name, List values) → new Attribute(String name, List values, boolean isUnsigned)
  • StructureDataScalar.addMember(String name, String desc, String units, DataType dtype, boolean isUnsigned, Number val) → StructureDataScalar.addMember(String name, String desc, String units, DataType dtype, Number val)

Variable Length (vlen) Dimensions and Variables

  • The CDM data model is clarified to allow vlen dimensions only in the outermost (fastest changing) Dimension.
  • Reading a Variable with only a single vlen Dimension will result in a regular Array.
  • Reading a Variable with a nested vlen Dimension will result in an ArrayObject containing regular Arrays of independent lengths.
  • In both cases the returned array’s DataType will be the primitive type.
  • Previously the exact Array class and DataType returned from a read on a vlen was not well-defined.
  • Use Array.isVlen() to discover if an Array represents vlen data.
  • ArrayObject.factory(Class classType, Index index) is now ArrayObject.factory(DataType dtype, Class classType, boolean isVlen, Index index)
  • Use Array.makeVlenArray(int[] shape, Array[] data) to construct vlen data.

AutoCloseable

AutoCloseable was introduced in Java 7, along with the try-with-resources language feature. Use of this feature makes code more readable and more reliable in ensuring that resources (like file handles) are released when done. We strongly recommend that you modify your code to take advantage of it wherever possible. For example:

 try (NetcdfFile ncfile = NetcdfFile.open(location)) {
    ...
 } catch (IOException ioe) {
   // handle ioe here, or propagate by not using catch clause
 }
  • The following now implement AutoCloseable, so can be the target of try-with-resource:
  • NetcdfFile, NetcdfFileWriter
  • HTTPMethod, HTTPSession
  • FeatureDataset, CoverageDataset, CoverageDatasetCollection
  • CFPointWriter, Grib2NetcdfFile
  • ArrayStructure (deprecate finish())
  • PointFeatureCollectionIterator (finish() method now deprecated)
  • StructureDataIterator, PointFeatureIterator, PointFeatureCollectionIterator, NestedPointFeatureCollectionIterator
  • thredds.client.catalog.tools.DataFactory.Result

Iterable

Iterable was introduced in Java 7, along with the foreach language feature, and makes code more readable with less boilerplate. For example:

for (StructureData sdata : myArrayStructure) {
    // ...
}
  • The following now implement Iterable<>, and so can be the target of foreach:
  • Range implements Iterable<Integer> (replace first(), last(), stride() methods)
  • ArrayStructure implements Iterable<StructureData> (replace getStructureDataIterator() method)
  • PointFeatureIterator extends Iterator<PointFeature> (finish() method deprecated)
  • In order for PointFeatureIterator to implement Iterator<PointFeature>, the hasNext() and next() methods cannot throw IOException. The interface is changed to remove throws IOException, which will now be wrapped in RuntimeException.
  • PointFeatureCollection implements Iterable<PointFeature> (replace hasNext(), next(), finish(), resetIteration() methods)
  • StationTimeSeriesFeatureCollection implements Iterable<StationTimeSeriesFeature> (replace hasNext(), next(), finish(), resetIteration() methods)
  • ProfileFeatureCollection implements Iterable<ProfileFeature> (replace hasNext(), next(), finish(), resetIteration() methods)
  • TrajectoryFeatureCollection implements Iterable<TrajectoryFeature> (replace hasNext(), next(), finish(), resetIteration() methods)
  • StationProfileFeature implements Iterable<ProfileFeature> (replace hasNext(), next(), finish(), resetIteration() methods)
  • StationProfileFeatureCollection implements Iterable<StationProfileFeature> (replace hasNext(), next(), finish(), resetIteration() methods)
  • SectionFeature implements Iterable<ProfileFeature> (replace hasNext(), next(), finish(), resetIteration() methods)
  • SectionFeatureCollection implements Iterable<SectionFeature> (replace hasNext(), next(), finish(), resetIteration() methods)

ucar.nc2.util.DiskCache2

  • All instances of DiskCache2 now have one cleanup thread
  • The DiskCache2.exit() method is now static and need only be called once when the application is exiting.
  • DiskCache2.setLogger() is removed.
  • DiskCache2.cleanCache(File dir, StringBuffer sbuff, boolean isRoot) is now DiskCache2.cleanCache(File dir, Formatter sbuff, boolean isRoot)
  • deprecated methods are removed: setCachePathPolicy(int cachePathPolicy, String cachePathPolicyParam), setPolicy(int cachePathPolicy)
  • logging of routine cache cleanup is now at DEBUG level

ucar.ma2.Range

  • Range.copy(String name) replaced by Range.setName(String name)
  • Range.getIterator() deprecated, use Range.iterator()

Currently a Range is specified by start:end:stride In the future, may be extended with subclasses RangeScatter and RangeComposite You should use the iterator now to ensure correct functionality. To iterate over the values of the Range:

for (int i=range.first(); i<=range.last(); i+= range.stride()) {    // REPLACE THIS
 // ...
}

for (int i : range) {  // USE THIS
 // ...
}

ucar.nc2.util.cache

  • FileCache and FileFactory take a DatasetUrl instead of a String location

ucar.nc2.dataset

In order to disambiguate remote protocols, all using http:, the utility method DatasetUrl.findDatasetUrl(location) is used to determine the protocol and capture the result in a DatasetUrl object. Because this can be expensive, the DatasetUrl should be calculated once and kept for the duration of the dataset access. When the protocol is already known, the DatasetUrl(ServiceType protocol, String location) constructor may be used. The API is changed to allow/require the application to compute these DatasetUrls.

  • NetcdfDataset.acquireDataset() takes a DatasetUrl instead of a String location.

  • the general method of NetcdfDataset.openDataset() takes a DatasetUrl instead of a String location. Variants use a String location, and call DatasetUrl.findDatasetUrl(location).

  • CoordinateAxis2D.getMidpoints() was deprecated and now removed, use getCoordValuesArray()

ucar.nc2.ft.PointFeature

  • Added method getTimeUnit(). An implementation exists in ucar.nc2.ft.point.PointFeatureImpl, so if your PointFeature extends it, you shouldn’t need to do any work.
  • Removed method getObservationTimeAsDate(). Instead, use getObservationTimeAsCalendarDate().toDate().
  • Removed method getNominalTimeAsDate(). Instead, use getNominalTimeAsCalendarDate().toDate().
  • Removed method getData(). Instead, use getDataAll().

ucar.ma2.MAMath

  • Added method equals(Array, Array). It is intended for use in Object.equals() implementations. This means, among other things, that corresponding floating-point elements must be exactly equal, not merely within some epsilon of each other.
  • Added method hashCode(Array array). It is intended for use in Object.hashCode() implementations and is compatible with equals(Array, Array).
  • Renamed isEqual(Array, Array) to nearlyEquals(Array, Array). This was done to avoid (some) confusion with the new equals(Array, Array), and to highlight that this method performs approximate comparison of floating-point numbers, instead of the exact comparison done by equals(Array, Array).

Coordinate Systems

  • ucar.nc2.dataset.CoordTransBuilderIF is split into ucar.nc2.dataset.builder.HorizTransformBuilderIF and ucar.nc2.dataset.builder.VertTransformBuilderIF
  • HorizTransformBuilderIF now uses AttributeContainer instead of NetcdfDataset, Variable
  • CoordinateTransform.makeCoordinateTransform(NetcdfDataset ds, Variable ctv) is now ProjectionCT makeCoordinateTransform(AttributeContainer ctv)
  • Previously, the optional false_easting and false_northing should match the units of the x and y projection coordinates in ucar.nc2.dataset.CoordinateSystem
  • List<Dimension> getDomain() → Collection<Dimension> getDomain()
  • boolean isSubset(List<Dimension> subset, List<Dimension> set) → isSubset(Collection<Dimension> subset, Collection<Dimension> set)

Feature Datasets

  • ucar.nc2.dt.TypedDatasetFactory has been removed. Use ucar.nc2.ft.FeatureDatasetFactoryManager
  • ucar.nc2.dt.grid is deprecated (but not removed) and is replaced by ucar.nc2.ft2.coverage
  • ucar.nc2.dt.point and ucar.nc2.dt.trajectory have been removed, replaced by ucar.nc2.ft.\*
  • In FeatureDataset, deprecated methods getDateRange(), getStartDate(), getStartDate() have been removed
  • In FeatureDataset, mutating method removed: calcBounds()

Point Feature Datasets (ucar.nc2.ft and ucar.nc2.ft.point)

  • FeatureCollection has been renamed to ucar.nc2.ft.DsgFeatureCollection for clarity.
  • SectionFeature and SectionFeatureCollection have been renamed to TrajectoryProfileFeature, TrajectoryProfileFeatureCollection for clarity.
  • FeatureType.SECTION renamed to FeatureType.TRAJECTORY_PROFILE for clarity.
  • NestedPointFeatureCollection has been removed, use PointFeatureCC and PointFeatureCCC instead when working with DsgFeatureCollection in a general way.
  • In all the Point Feature classes, DateUnit, Date, and DateRange have been replaced by CalendarDateUnit, CalendarDate, and CalendarDateRange:
  • In PointFeature and subclasses, deprecated methods getObservationTimeAsDate(), getNominalTimeAsDate() have been removed
  • In ProfileFeature, getTime() returns CalendarDate instead of Date
  • In PointFeature implementations and subclasses, all constructors use CalendarDateUnit instead of DateUnit, and all subset() and flatten() methods use CalendarDateRange, not DateRange
  • In CFPointWriter subclasses, all constructors use CalendarDateUnit instead of DateUnit
  • In PointFeature, deprecated method getData() is removed; usually replace it with getDataAll()
  • In PointFeatureCollection, mutating methods are removed: setCalendarDateRange(), setBoundingBox(), setSize(), calcBounds()
  • The time and altitude units for the collection can be found in the DsgFeatureCollection, and you can get the collection object from PointFeature.getFeatureCollection()
  • In PointFeatureIterator and subclasses, methods setCalculateBounds(), getDateRange(), getCalendarDateRange(), getBoundingBox(), getSize() have been removed. That information is obtained from the DsgFeatureCollection.
  • In PointFeatureIterator and subclasses, setBufferSize() bas been removed.
  • In PointFeatureCollection and subclasses, getPointFeatureIterator() no longer accepts a bufferSize argument.

Coverage Feature Datasets (ucar.nc2.ft2.coverage)

  • Completely new package ucar.nc2.ft2.coverage that replaces ucar.nc2.dt.grid The class FeatureDatasetCoverage replaces GridDataset.
  • Uses of classes in car.nc2.dt.grid are deprecated, though the code is still in the core jar file for now.
  • For new API see Coverage Datasets
  • FeatureType.COVERAGE is the general term for GRID, FMRC, SWATH, CURVILINEAR types. Previously, GRID was used as the general type, now it refers to a specific type of Coverage. Affects FeatureDatasetFactoryManager.open(FeatureType wantFeatureType, ...)

Shared Dimensions

  • Group.addDimension and Group.addDimensionIfNotExists methods now throw an IllegalArgumentException if the dimension isn’t shared.
  • NetcdfFileWriter.addDimension methods no longer have an isShared parameter. Such dimensions should always be shared and allowing them to be private is confusing and error-prone.

Catalog API

  • All uses of classes in thredds.catalog are deprecated. If you still need these, you must add legacy.jar to your path.
  • TDS and CDM now use thredds.server.catalog and thredds.client.catalog. The APIs are different, but with equivalent functionality to thredds.catalog.
  • thredds.client.DatasetNode now has getDatasetsLogical() and getDatasetsLocal() that does or does not dereference a CatalogRef, respectively. You can also use getDatasets() which includes a dereferenced catalog if it has already been read.