Added more documentation, some minor IFD tags, removed test for LZW

Author: fietenoer

src/main/java/org/apache/sysds/runtime/io/ReaderCOG.java

@@ -59,6 +59,9 @@ private MatrixBlock readCOG(BufferedInputStream bis, long estnnz) throws IOExcep
         // TODO: Currently only reads the first image which is the full resolution image
         // In the future, this could be extended to read the overviews as well
         // But keep in mind that we are only returning a single MatrixBlock, so there needs to be some special handling
+        // TODO: Is the metadata (e.g. the coordinates) necessary in SystemDS? Currently not possible as we only return a MatrixBlock
+        // However, this could possibly be changed in the future to somehow also store relevant metadata if desired.
+        // Currently this implementation reads the most important data from the header.
         COGProperties cogP = new COGProperties(cogHeader.getIFD());
 
         // number of tiles for Width and Length
@@ -97,8 +100,6 @@ private MatrixBlock readCOG(BufferedInputStream bis, long estnnz) throws IOExcep
                 byteReader.reset();
             }
 
-            // TODO: If the tile is compressed, decompress the currentTileData here
-
             if (cogP.getCompression() == 8) {
                 currentTileData = COGCompressionUtils.decompressDeflate(currentTileData);
             }

src/main/java/org/apache/sysds/runtime/io/cog/COGByteReader.java

@@ -5,6 +5,11 @@
 import java.io.BufferedInputStream;
 import java.io.IOException;
 
+/**
+ * This class is used by the COGReader to read bytes from a BufferedInputStream.
+ * It is wrapper that keeps track of the bytes read and can therefore be used to
+ * easily go to specific offsets.
+ */
 public class COGByteReader {
     private long totalBytesRead;
     private BufferedInputStream bis;
@@ -45,18 +50,38 @@ public byte[] readBytes(int length) {
         return header;
     }
 
+    /**
+     * Reads a given number of bytes from the BufferedInputStream.
+     * Increments the totalBytesRead counter by the number of bytes read.
+     * @param length
+     * @return
+     */
     public byte[] readBytes(long length) {
+        // TODO: When properly implementing BigTIFF, this could be a problem when not being able to skip bytes
+        // In BigTIFF the offset can be larger than maxInt which isn't a problem for skipping bytes
+        // but could be a problem when the tiles are not sequential in the file and we need to jump back
+        // to a previous position (where we can't use skip).
         if (length > Integer.MAX_VALUE) {
             throw new DMLRuntimeException("Cannot read more than Integer.MAX_VALUE bytes at once");
         }
         return readBytes((int) length);
     }
 
+    /**
+     * Offers the same functionality as BufferedInputStream.mark.
+     * Allows for returning to a previous point if the readlimit is not exceeded.
+     * @param readlimit
+     */
     public void mark(long readlimit) {
         this.readlimit = readlimit;
         bis.mark((int) readlimit + 1);
     }
 
+    /**
+     * Offers the same functionality as BufferedInputStream.reset.
+     * Resets the stream to the last marked position.
+     * @throws DMLRuntimeException
+     */
     public void reset() throws DMLRuntimeException {
         try {
             bis.reset();
@@ -66,6 +91,12 @@ public void reset() throws DMLRuntimeException {
         }
     }
 
+    /**
+     * Skips a given number of bytes without reading them.
+     * Useful for jumping to specific offsets
+     * @param n Number of bytes to skip
+     * @throws DMLRuntimeException
+     */
     public void skipBytes(long n) throws DMLRuntimeException {
         try {
             long skipped = bis.skip(n);

src/main/java/org/apache/sysds/runtime/io/cog/COGCompressionUtils.java

@@ -14,6 +14,7 @@ public class COGCompressionUtils {
      * @throws DMLRuntimeException
      */
     public static byte[] decompressDeflate(byte[] compressedData) throws DMLRuntimeException {
+        // Use the native Java implementation of deflate to decompress the data
         Inflater inflater = new Inflater();
         inflater.setInput(compressedData);
 

src/main/java/org/apache/sysds/runtime/io/cog/COGHeader.java

@@ -2,12 +2,13 @@
 
 import org.apache.sysds.runtime.DMLRuntimeException;
 
-import java.io.BufferedInputStream;
-import java.io.IOException;
 import java.util.ArrayList;
 import java.nio.ByteBuffer;
 import java.nio.ByteOrder;
 
+/**
+ * Represents a header for a COG file. This includes IFDs, endianess etc.
+ */
 public class COGHeader {
     private boolean isLittleEndian;
     private String GDALMetadata;
@@ -137,6 +138,12 @@ public Number parseByteArray(byte[] bytes, int length, int offset, boolean isDec
         }
     }
 
+    /**
+     * Prepares the COG header by reading the first 4 bytes and determining the byte order.
+     * Needs to be called before anything else is done with the COG header.
+     * @param byteReader
+     * @return
+     */
     private static COGHeader prepareHeader(COGByteReader byteReader) {
         // Read first 4 bytes to determine byte order and make sure it is a valid TIFF
         byte[] header = byteReader.readBytes(4);
@@ -388,6 +395,8 @@ public static String isCompatible(IFDTag[] IFD) {
                     break;
                 case Compression:
                     // After implementing additional decompression methods, this can be extended
+                    // TODO: LZW would be a great addition as it is widely used
+                    // Furthermore, JPEG support would also be a good addition
                     // 1: none, 8: deflate
                     if (tag.getData()[0].intValue() != 1 && tag.getData()[0].intValue() != 8) {
                         return "Unsupported compression: " + tag.getData()[0];

src/main/java/org/apache/sysds/runtime/io/cog/COGProperties.java

@@ -2,6 +2,9 @@
 
 import java.util.Arrays;
 
+/**
+ * Properties of a COG file that are useful for reading the file.
+ */
 public class COGProperties {
 
     private int rows;

src/main/java/org/apache/sysds/runtime/io/cog/IFDTag.java

@@ -1,5 +1,8 @@
 package org.apache.sysds.runtime.io.cog;
 
+/**
+ * Represents a single tag in the IFD of a TIFF file
+ */
 public class IFDTag {
     private IFDTagDictionary tagId;
     private short dataType;

src/main/java/org/apache/sysds/runtime/io/cog/IFDTagDictionary.java

@@ -1,5 +1,8 @@
 package org.apache.sysds.runtime.io.cog;
 
+/**
+ * Enum for mapping IFD tag values to their corresponding tag names
+ */
 public enum IFDTagDictionary {
     Unknown(-1),
     // Right now we will only support baseline TIFF
@@ -8,7 +11,6 @@ public enum IFDTagDictionary {
     ImageWidth(256),
     ImageLength(257),
     BitsPerSample(258),
-    // TODO: LZW Compression
     Compression(259),
     PhotometricInterpretation(262),
     Threshholding(263),
@@ -53,9 +55,13 @@ public enum IFDTagDictionary {
     TileLength(323),
     TileOffsets(324),
     TileByteCounts(325),
-    // TODO: Support nodata values
-    GDALNoData(42113);
-
+    GDALNoData(42113),
+    GeoKeyDirectoryTag(34735),
+    GeoDoubleParamsTag(34736),
+    GeoAsciiParamsTag(34737),
+    ModelPixelScaleTag(33550),
+    ModelTiepointTag(33922),
+    ModelTransformationTag(34264);
 
 
     private final int value;

src/main/java/org/apache/sysds/runtime/io/cog/SampleFormatDataTypes.java

@@ -1,5 +1,8 @@
 package org.apache.sysds.runtime.io.cog;
 
+/**
+ * Enum for mapping sample formats of TIFF image data to names
+ */
 public enum SampleFormatDataTypes {
     UNSIGNED_INTEGER(1),
     SIGNED_INTEGER(2),

src/main/java/org/apache/sysds/runtime/io/cog/TIFFDataTypes.java

@@ -1,5 +1,8 @@
 package org.apache.sysds.runtime.io.cog;
 
+/**
+ * Enum for mapping data types of IFD tags in TIFF to readable names
+ */
 public enum TIFFDataTypes {
     BYTE(1),
     ASCII(2),