use of org.apache.sis.image.DataType in project sis by apache.
the class ImageRenderer method setData.
/**
* Sets the data as vectors. The number of vectors must be equal to the {@linkplain #getNumBands() expected number of bands}.
* All vectors must be backed by arrays (indirectly, through {@linkplain Vector#buffer() buffers} backed by arrays) and have
* the same {@linkplain Vector#size() size}.
* This method wraps the underlying arrays of a primitive type into a Java2D buffer; data are not copied.
*
* <p><b>Implementation note:</b> the NIO buffers are set by a call to {@link #setData(DataType, Buffer...)},
* which can be overridden by subclasses if desired.</p>
*
* @param data the vectors wrapping arrays of primitive type.
* @throws NullArgumentException if {@code data} is null or one of {@code data} element is null.
* @throws MismatchedCoverageRangeException if the number of specified vectors is not equal to the number of bands.
* @throws UnsupportedOperationException if a vector is not backed by an accessible array or is read-only.
* @throws RasterFormatException if vectors do not have the same size.
* @throws ArithmeticException if a buffer position overflows the 32 bits integer capacity.
*/
public void setData(final Vector... data) {
ArgumentChecks.ensureNonNull("data", data);
ensureExpectedBandCount(data.length, true);
final Buffer[] buffers = new Buffer[data.length];
DataType dataType = null;
for (int i = 0; i < data.length; i++) {
final Vector v = data[i];
ArgumentChecks.ensureNonNullElement("data", i, v);
final DataType t = DataType.forPrimitiveType(v.getElementType(), v.isUnsigned());
if (dataType == null) {
dataType = t;
} else if (dataType != t) {
throw new RasterFormatException(Resources.format(Resources.Keys.MismatchedDataType));
}
buffers[i] = v.buffer().orElseThrow(UnsupportedOperationException::new);
}
setData(dataType, buffers);
}
use of org.apache.sis.image.DataType in project sis by apache.
the class ConvertedGridCoverage method getBandType.
/**
* Returns the data type for range of values of given sample dimensions.
* This data type applies to each band, not to a packed sample model
* (e.g. we assume no packing of 4 byte values in a single 32-bits integer).
*
* @param targets the sample dimensions for which to get the data type.
* @param converted whether the image will hold converted or packed values.
* @param source if the type can not be determined, coverage from which to inherit the type as a fallback.
* @return the data type (never null).
*
* @see GridCoverage#getBandType()
*/
static DataType getBandType(final List<SampleDimension> targets, final boolean converted, final GridCoverage source) {
NumberRange<?> union = null;
boolean allowsNaN = false;
for (final SampleDimension dimension : targets) {
final Optional<NumberRange<?>> c = dimension.getSampleRange();
if (c.isPresent()) {
final NumberRange<?> range = c.get();
if (union == null) {
union = range;
} else {
/*
* We do not want unit conversions for this union, because the union is used
* only for determining a data type having the capacity to store the values.
* The physical meaning of those values is not relevant here.
*/
if (union instanceof MeasurementRange<?>) {
union = new NumberRange<>(union);
}
union = union.unionAny(range);
}
}
if (!allowsNaN)
allowsNaN = dimension.allowsNaN();
}
if (union == null) {
return source.getBandType();
}
DataType type = DataType.forRange(union, !converted);
if (allowsNaN) {
type = type.toFloat();
}
return type;
}
use of org.apache.sis.image.DataType in project sis by apache.
the class DataSubset method readSlice.
/**
* Reads a two-dimensional slice of the data cube from the given input channel. This method is usually
* invoked for reading the tile in full, in which case the {@code lower} argument is (0,0) and the
* {@code upper} argument is the tile size. But those arguments may identify a smaller region if the
* {@link DataSubset} contains only one (potentially large) tile.
*
* <p>The length of {@code lower}, {@code upper} and {@code subsampling} arrays shall be 2.</p>
*
* <h4>Default implementation</h4>
* The default implementation in this base class assumes uncompressed data without band subset.
* Subsampling on the <var>X</var> axis is not supported if the image has interleaved pixels.
* Packed pixels (é.g. bilevel images with 8 pixels per byte) are not supported.
* Those restrictions are verified by {@link DataCube#canReadDirect(TiledGridResource.Subset)}.
* Subclasses must override for handling decompression or for resolving above-cited limitations.
*
* @todo It is possible to relax a little bit some restrictions. If the tile width is a divisor
* of the sample size, we could round {@code lower[0]} and {@code upper[0]} to a multiple
* of {@code sampleSize}. We would need to adjust the coordinates of returned image accordingly.
* This adjustment need to be done by the caller.
*
* @param offsets position in the channel where tile data begins, one value per bank.
* @param byteCounts number of bytes for the compressed tile data, one value per bank.
* @param lower (<var>x</var>, <var>y</var>) coordinates of the first pixel to read relative to the tile.
* @param upper (<var>x</var>, <var>y</var>) coordinates after the last pixel to read relative to the tile.
* @param subsampling (<var>sx</var>, <var>sy</var>) subsampling factors.
* @param location pixel coordinates in the upper-left corner of the tile to return.
* @return a single tile decoded from the GeoTIFF file.
* @throws IOException if an I/O error occurred.
* @throws DataStoreException if a logical error occurred.
* @throws RuntimeException if the Java2D image can not be created for another reason
* (too many exception types to list them all).
*
* @see DataCube#canReadDirect(TiledGridResource.Subset)
*/
Raster readSlice(final long[] offsets, final long[] byteCounts, final long[] lower, final long[] upper, final int[] subsampling, final Point location) throws IOException, DataStoreException {
final DataType type = getDataType();
// Assumed same as `SampleModel.getSampleSize(…)` by pre-conditions.
final int sampleSize = type.size();
final long width = subtractExact(upper[X_DIMENSION], lower[X_DIMENSION]);
final long height = subtractExact(upper[Y_DIMENSION], lower[Y_DIMENSION]);
/*
* The number of bytes to read should not be greater than `byteCount`. It may be smaller however if only
* a subregion is read. Note that the `length` value may be different than `capacity` if the tile to read
* is smaller than the "standard" tile size of the image. It happens often when reading the last strip.
* This length is used only for verification purpose so it does not need to be exact.
*/
final long length = ceilDiv(width * height * sourcePixelStride * sampleSize, Byte.SIZE);
final long[] size = new long[] { multiplyFull(sourcePixelStride, getTileSize(X_DIMENSION)), getTileSize(Y_DIMENSION) };
/*
* If we use an interleaved sample model, each "element" from `HyperRectangleReader` perspective is actually a
* group of `sourcePixelStride` values. Note that in such case, we can not handle subsampling on the first axis.
* Such case should be handled by the `CompressedSubset` subclass instead, even if there is no compression.
*/
assert sourcePixelStride == 1 || subsampling[X_DIMENSION] == 1;
lower[X_DIMENSION] *= sourcePixelStride;
upper[X_DIMENSION] *= sourcePixelStride;
/*
* Read each plane ("banks" in Java2D terminology). Note that a single bank contains all bands
* in the interleaved sample model case. This block assumes that each bank element contains
* exactly one sample value (verified by assertion), as documented in the Javadoc of this method.
* If that assumption was not true, we would have to adjust `capacity`, `lower[0]` and `upper[0]`
* (we may do that as an optimization in a future version).
*/
final HyperRectangleReader hr = new HyperRectangleReader(ImageUtilities.toNumberEnum(type.toDataBufferType()), input());
final Region region = new Region(size, lower, upper, subsampling);
final Buffer[] banks = new Buffer[numBanks];
for (int b = 0; b < numBanks; b++) {
if (b < byteCounts.length && length > byteCounts[b]) {
throw new DataStoreContentException(source.reader.resources().getString(Resources.Keys.UnexpectedTileLength_2, length, byteCounts[b]));
}
hr.setOrigin(offsets[b]);
// See above comment.
assert model.getSampleSize(b) == sampleSize;
final Buffer bank = hr.readAsBuffer(region, getBankCapacity(1));
fillRemainingRows(bank);
banks[b] = bank;
}
final DataBuffer buffer = RasterFactory.wrap(type, banks);
return Raster.createWritableRaster(model, buffer, location);
}
use of org.apache.sis.image.DataType in project sis by apache.
the class CompressedSubset method readSlice.
/**
* Reads a two-dimensional slice of the data cube from the given input channel.
*
* @param offsets position in the channel where tile data begins, one value per bank.
* @param byteCounts number of bytes for the compressed tile data, one value per bank.
* @param lower (<var>x</var>, <var>y</var>) coordinates of the first pixel to read relative to the tile.
* @param upper (<var>x</var>, <var>y</var>) coordinates after the last pixel to read relative to the tile.
* @param subsampling (<var>sx</var>, <var>sy</var>) subsampling factors.
* @param location pixel coordinates in the upper-left corner of the tile to return.
* @return a single tile decoded from the GeoTIFF file.
*/
@Override
Raster readSlice(final long[] offsets, final long[] byteCounts, final long[] lower, final long[] upper, final int[] subsampling, final Point location) throws IOException, DataStoreException {
final DataType dataType = getDataType();
final int width = pixelCount(lower, upper, subsampling, X_DIMENSION);
final int height = pixelCount(lower, upper, subsampling, Y_DIMENSION);
final int chunksPerRow = width * (targetPixelStride / samplesPerChunk);
final int betweenRows = subsampling[1] - 1;
final long head = beforeFirstBand + sourcePixelStride * (lower[X_DIMENSION]);
final long tail = afterLastBand - sourcePixelStride * (lower[X_DIMENSION] + (width - 1) * subsampling[X_DIMENSION]);
/*
* `head` and `tail` are the number of sample values to skip at the beginning and end of each row.
* `betweenPixels` is the number of sample values to skip between each pixel, but the actual skips
* are more complicated if only a subset of the bands are read. The actual number of sample values
* to skip between "chunks" is detailed by `skipAfterChunks`.
*
* `pixelsPerElement` below is a factor for converting a count of pixels to a count of primitive elements
* in the bank. The `pixelsPerElement` factor is usually 1, except when more than one pixel is packed in
* each single primitive type (e.g. 8 bits per byte in bilevel image). The `head` needs to be a multiple
* of `pixelsPerElement`; this restriction is documented in `Inflater.skip(long)` and should have been
* verified by `TiledGridResource`.
*/
// Always ≥ 1 and usually = 1.
final int pixelsPerElement = getPixelsPerElement();
assert (head % pixelsPerElement) == 0 : head;
if (inflater == null) {
inflater = Inflater.create(this, input(), source.getCompression(), source.getPredictor(), sourcePixelStride, getTileSize(X_DIMENSION), chunksPerRow, samplesPerChunk, skipAfterChunks, pixelsPerElement, dataType);
}
final Inflater inflater = this.inflater;
final int capacity = getBankCapacity(pixelsPerElement);
final Buffer[] banks = new Buffer[numBanks];
for (int b = 0; b < numBanks; b++) {
/*
* Prepare the object which will perform the actual decompression row-by-row,
* optionally skipping chunks if a subsampling is applied.
*/
final Buffer bank = RasterFactory.createBuffer(dataType, capacity);
inflater.setInputOutput(offsets[b], byteCounts[b], bank);
/*
* At this point, `inflater` is a data input doing decompression eventually followed
* by the mathematical operation identified by `predictor`.
*/
for (long y = lower[1]; --y >= 0; ) {
// `skip(…)` may round to next element boundary.
inflater.skip(scanlineStride);
}
for (int y = height; --y > 0; ) {
// (height - 1) iterations.
inflater.skip(head);
inflater.uncompressRow();
inflater.skip(tail);
for (int j = betweenRows; --j >= 0; ) {
inflater.skip(scanlineStride);
}
}
// Last iteration without the trailing `skip(…)` calls.
inflater.skip(head);
inflater.uncompressRow();
fillRemainingRows(bank.flip());
banks[b] = bank;
}
return Raster.createWritableRaster(model, RasterFactory.wrap(dataType, banks), location);
}
Aggregations