package nom.tam.fits;
   
   /*-
    * #%L
    * nom.tam.fits
    * %%
    * Copyright (C) 1996 - 2024 nom-tam-fits
    * %%
    * This is free and unencumbered software released into the public domain.
   * 
   * Anyone is free to copy, modify, publish, use, compile, sell, or
   * distribute this software, either in source code form or as a compiled
   * binary, for any purpose, commercial or non-commercial, and by any
   * means.
   * 
   * In jurisdictions that recognize copyright laws, the author or authors
   * of this software dedicate any and all copyright interest in the
   * software to the public domain. We make this dedication for the benefit
   * of the public at large and to the detriment of our heirs and
   * successors. We intend this dedication to be an overt act of
   * relinquishment in perpetuity of all present and future rights to this
   * software under copyright law.
   * 
   * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
   * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
   * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
   * IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR
   * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
   * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
   * OTHER DEALINGS IN THE SOFTWARE.
   * #L%
   */
  
  import java.io.IOException;
  import java.lang.reflect.Array;
  import java.util.Arrays;
  
  import nom.tam.fits.header.Bitpix;
  import nom.tam.fits.header.IFitsHeader;
  import nom.tam.fits.header.Standard;
  import nom.tam.util.ArrayDataInput;
  import nom.tam.util.ArrayDataOutput;
  import nom.tam.util.ArrayFuncs;
  import nom.tam.util.ByteFormatter;
  import nom.tam.util.ByteParser;
  import nom.tam.util.Cursor;
  import nom.tam.util.FormatException;
  
  import static nom.tam.fits.header.Standard.NAXIS1;
  import static nom.tam.fits.header.Standard.NAXIS2;
  import static nom.tam.fits.header.Standard.TBCOLn;
  import static nom.tam.fits.header.Standard.TDMAXn;
  import static nom.tam.fits.header.Standard.TDMINn;
  import static nom.tam.fits.header.Standard.TFIELDS;
  import static nom.tam.fits.header.Standard.TFORMn;
  import static nom.tam.fits.header.Standard.TLMAXn;
  import static nom.tam.fits.header.Standard.TLMINn;
  import static nom.tam.fits.header.Standard.TNULLn;
  
  /**
   * ASCII table data. ASCII tables are meant for human readability without any special tools. However, they are far less
   * flexible or compact than {@link BinaryTable}. As such, users are generally discouraged from using this type of table
   * to represent FITS table data. This class only supports scalar entries of type <code>int</code>, <code>long</code>,
   * <code>float</code>, <code>double</code>, or else <code>String</code> types.
   * 
   * @see AsciiTableHDU
   * @see BinaryTable
   */
  @SuppressWarnings("deprecation")
  public class AsciiTable extends AbstractTableData {
  
      private static final int MAX_INTEGER_LENGTH = 10;
  
      private static final int FLOAT_MAX_LENGTH = 16;
  
      private static final int LONG_MAX_LENGTH = 20;
  
      private static final int INT_MAX_LENGTH = 10;
  
      private static final int DOUBLE_MAX_LENGTH = 24;
  
      /** Whether I10 columns should be treated as <code>int</code> provided that defined limits allow for it. */
      private static boolean isI10PreferInt = true;
  
      // private static final Logger LOG = Logger.getLogger(AsciiTable.class.getName());
  
      /** The number of rows in the table */
      private int nRows;
  
      /** The number of fields in the table */
      private int nFields;
  
      /** The number of bytes in a row */
      private int rowLen;
  
      /** The null string for the field */
      private String[] nulls;
  
      /** The type of data in the field */
     private Class<?>[] types;
 
     /** The offset from the beginning of the row at which the field starts */
     private int[] offsets;
 
     /** The number of bytes in the field */
     private int[] lengths;
 
     /** The byte buffer used to read/write the ASCII table */
     private byte[] buffer;
 
     /** Markers indicating fields that are null */
     private boolean[] isNull;
 
     /** Column names */
     private String[] names;
 
     /**
      * An array of arrays giving the data in the table in binary numbers
      */
     private Object[] data;
 
     /**
      * The parser used to convert from buffer to data.
      */
     private ByteParser bp;
 
     /** The actual stream used to input data */
     private ArrayDataInput currInput;
 
     /** Create an empty ASCII table */
     public AsciiTable() {
         data = new Object[0];
         buffer = null;
         nFields = 0;
         nRows = 0;
         rowLen = 0;
         types = new Class[0];
         lengths = new int[0];
         offsets = new int[0];
         nulls = new String[0];
         names = new String[0];
     }
 
     /**
      * Creates an ASCII table given a header. For tables that contain integer-valued columns of format <code>I10</code>,
      * the {@link #setI10PreferInt(boolean)} mayb be used to control whether to treat them as <code>int</code> or as
      * <code>long</code> values (the latter is the default).
      *
      * @param      hdr           The header describing the table
      *
      * @throws     FitsException if the operation failed
      * 
      * @deprecated               (<i>for internal use</i>) Visibility may be reduced to the package level in the future.
      */
     @Deprecated
     public AsciiTable(Header hdr) throws FitsException {
         this(hdr, isI10PreferInt);
     }
 
     /**
      * <p>
      * Create an ASCII table given a header, with custom integer handling support.
      * </p>
      * <p>
      * The <code>preferInt</code> parameter controls how columns with format "<code>I10</code>" are handled; this is
      * tricky because some, but not all, integers that can be represented in 10 characters can be represented as 32-bit
      * integers. Setting it <code>true</code> may make it more likely to avoid unexpected type changes during
      * round-tripping, but it also means that some (large number) data in I10 columns may be impossible to read.
      * </p>
      * 
      * @param      hdr           The header describing the table
      * @param      preferInt     if <code>true</code>, format "I10" columns will be assumed <code>int.class</code>,
      *                               provided TLMINn/TLMAXn or TDMINn/TDMAXn limits (if defined) allow it. if
      *                               <code>false</code>, I10 columns that have no clear indication of data range will be
      *                               assumed <code>long.class</code>.
      *
      * @throws     FitsException if the operation failed
      * 
      * @deprecated               Use {@link #setI10PreferInt(boolean)} instead prior to reading ASCII tables.
      */
     @Deprecated
     public AsciiTable(Header hdr, boolean preferInt) throws FitsException {
         String ext = hdr.getStringValue(Standard.XTENSION, Standard.XTENSION_IMAGE);
 
         if (!ext.equalsIgnoreCase(Standard.XTENSION_ASCIITABLE)) {
             throw new FitsException("Not an ASCII table header (XTENSION = " + hdr.getStringValue(Standard.XTENSION) + ")");
         }
 
         nRows = hdr.getIntValue(NAXIS2);
         nFields = hdr.getIntValue(TFIELDS);
         rowLen = hdr.getIntValue(NAXIS1);
 
         types = new Class[nFields];
         offsets = new int[nFields];
         lengths = new int[nFields];
         nulls = new String[nFields];
         names = new String[nFields];
 
         for (int i = 0; i < nFields; i++) {
             names[i] = hdr.getStringValue(Standard.TTYPEn.n(i + 1), TableHDU.getDefaultColumnName(i));
             offsets[i] = hdr.getIntValue(TBCOLn.n(i + 1)) - 1;
             String s = hdr.getStringValue(TFORMn.n(i + 1));
             if (offsets[i] < 0 || s == null) {
                 throw new FitsException("Invalid Specification for column:" + (i + 1));
             }
             s = s.trim();
             char c = s.charAt(0);
             s = s.substring(1);
             if (s.indexOf('.') > 0) {
                 s = s.substring(0, s.indexOf('.'));
             }
 
             try {
                 lengths[i] = Integer.parseInt(s);
             } catch (NumberFormatException e) {
                 throw new HeaderCardException("Invalid " + TFORMn.n(i + 1).key() + " value: '" + s + "'");
             }
 
             if (lengths[i] < 0) {
                 throw new HeaderCardException("Invalid " + TFORMn.n(i + 1).key() + " value: '" + s + "'");
             }
 
             switch (c) {
             case 'A':
                 types[i] = String.class;
                 break;
             case 'I':
                 if (lengths[i] == MAX_INTEGER_LENGTH) {
                     types[i] = guessI10Type(i, hdr, preferInt);
                 } else {
                     types[i] = lengths[i] > MAX_INTEGER_LENGTH ? long.class : int.class;
                 }
                 break;
             case 'F':
             case 'E':
                 types[i] = float.class;
                 break;
             case 'D':
                 types[i] = double.class;
                 break;
             default:
                 throw new FitsException("could not parse column type of ascii table");
             }
 
             nulls[i] = hdr.getStringValue(TNULLn.n(i + 1));
             if (nulls[i] != null) {
                 nulls[i] = nulls[i].trim();
             }
         }
     }
 
     /**
      * Creates an ASCII table from existing data in column-major format order.
      *
      * @param  columns       The data for scalar-valued columns. Each column must be an array of <code>int[]</code>,
      *                           <code>long[]</code>, <code>float[]</code>, <code>double[]</code>, or else
      *                           <code>String[]</code>, containing the same number of elements in each column (the
      *                           number of rows).
      * 
      * @return               a new ASCII table with the data. The tables data may be partially independent from the
      *                           argument. Modifications to the table data, or that to the argument have undefined
      *                           effect on the other object. If it is important to decouple them, you can use a
      *                           {@link ArrayFuncs#deepClone(Object)} of your original data as an argument.
      * 
      * @throws FitsException if the argument is not a suitable representation of FITS data in columns
      * 
      * @see                  BinaryTable#fromColumnMajor(Object[])
      * 
      * @since                1.19
      */
     public static AsciiTable fromColumnMajor(Object[] columns) throws FitsException {
         AsciiTable t = new AsciiTable();
         for (int i = 0; i < columns.length; i++) {
             try {
                 t.addColumn(columns[i]);
             } catch (Exception e) {
                 throw new FitsException("col[" + i + "]: " + e.getMessage(), e);
             }
         }
         return t;
     }
 
     void setColumnName(int col, String value)
             throws IllegalArgumentException, IndexOutOfBoundsException, HeaderCardException {
         HeaderCard.validateChars(value);
         names[col] = value;
     }
 
     /**
      * Checks if the integer value of a specific key requires <code>long</code> value type to store.
      *
      * @param  h   the header
      * @param  key the keyword to check
      *
      * @return     <code>true</code> if the keyword exists and has an integer value that is outside the range of
      *                 <code>int</code>. Otherwise <code>false</code>
      *
      * @see        #guessI10Type(int, Header, boolean)
      */
     private boolean requiresLong(Header h, IFitsHeader key, Long dft) {
         long l = h.getLongValue(key, dft);
         if (l == dft) {
             return false;
         }
 
         return (l < Integer.MIN_VALUE || l > Integer.MAX_VALUE);
     }
 
     /**
      * Guesses what type of values to use to return I10 type table values. Depending on the range of represented values
      * I10 may fit into <code>int</code> types, or else require <code>long</code> type arrays. Therefore, the method
      * checks for the presence of standard column limit keywords TLMINn/TLMAXn and TDMINn/TDMAXn and if these exist and
      * are outside of the range of an <code>int</code> then the call will return <code>long.class</code>. If the header
      * does not define the data limits (fully), it will return the class the caller prefers. Otherwise (data limits were
      * defined and fit into the <code>int</code> range) <code>int.class</code> will be returned.
      *
      * @param  col       the 0-based table column index
      * @param  h         the header
      * @param  preferInt whether we prefer <code>int.class</code> over <code>long.class</code> in case the header does
      *                       not provide us with a clue.
      *
      * @return           <code>long.class</code> if the data requires long or we prefer it. Othwerwise
      *                       <code>int.class</code>
      *
      * @see              #AsciiTable(Header, boolean)
      */
     private Class<?> guessI10Type(int col, Header h, boolean preferInt) {
         col++;
 
         if (requiresLong(h, TLMINn.n(col), Long.MAX_VALUE) || requiresLong(h, TLMAXn.n(col), Long.MIN_VALUE)
                 || requiresLong(h, TDMINn.n(col), Long.MAX_VALUE) || requiresLong(h, TDMAXn.n(col), Long.MIN_VALUE)) {
             return long.class;
         }
 
         if ((h.containsKey(TLMINn.n(col)) || h.containsKey(TDMINn.n(col))) //
                 && (h.containsKey(TLMAXn.n(col)) || h.containsKey(TDMAXn.n(col)))) {
             // There are keywords defining both min/max values, and none of them require long types...
             return int.class;
         }
 
         return preferInt ? int.class : long.class;
     }
 
     /**
      * Return the data type in the specified column, such as <code>int.class</code> or <code>String.class</code>.
      *
      * @param  col The 0-based column index
      *
      * @return     the class of data in the specified column.
      *
      * @since      1.16
      */
     public final Class<?> getColumnType(int col) {
         return types[col];
     }
 
     int addColInfo(int col, Cursor<String, HeaderCard> iter) {
         String tform = null;
         if (types[col] == String.class) {
             tform = "A" + lengths[col];
         } else if (types[col] == int.class || types[col] == long.class) {
             tform = "I" + lengths[col];
         } else if (types[col] == float.class) {
             tform = "E" + lengths[col] + ".0";
         } else if (types[col] == double.class) {
             tform = "D" + lengths[col] + ".0";
         }
 
         Standard.context(AsciiTable.class);
         if (names[col] != null) {
             iter.add(HeaderCard.create(Standard.TTYPEn.n(col + 1), names[col]));
         }
         iter.add(HeaderCard.create(Standard.TFORMn.n(col + 1), tform));
         iter.add(HeaderCard.create(Standard.TBCOLn.n(col + 1), offsets[col] + 1));
         Standard.context(null);
         return lengths[col];
     }
 
     @Override
     public int addColumn(Object newCol) throws FitsException, IllegalArgumentException {
         if (newCol == null) {
             throw new FitsException("data is null");
         }
 
         if (!newCol.getClass().isArray()) {
             throw new IllegalArgumentException("Not an array: " + newCol.getClass().getName());
         }
 
         int maxLen = 1;
         if (newCol instanceof String[]) {
             String[] sa = (String[]) newCol;
             for (String element : sa) {
                 if (element != null && element.length() > maxLen) {
                     maxLen = element.length();
                 }
             }
         } else if (newCol instanceof double[]) {
             maxLen = DOUBLE_MAX_LENGTH;
         } else if (newCol instanceof int[]) {
             maxLen = INT_MAX_LENGTH;
         } else if (newCol instanceof long[]) {
             maxLen = LONG_MAX_LENGTH;
         } else if (newCol instanceof float[]) {
             maxLen = FLOAT_MAX_LENGTH;
         } else {
             throw new FitsException(
                     "No AsciiTable support for elements of " + newCol.getClass().getComponentType().getName());
         }
         addColumn(newCol, maxLen);
 
         // Invalidate the buffer
         buffer = null;
 
         return nFields;
     }
 
     /**
      * Adds an ASCII table column with the specified ASCII text width for storing its elements.
      *
      * @param  newCol                   The new column data, which must be one of: <code>int[]</code>,
      *                                      <code>long[]</code>, <code>float[]</code>, <code>double[]</code>, or else
      *                                      <code>String[]</code>. If the table already contains data, the length of the
      *                                      array must match the number of rows already contained in the table.
      * @param  width                    the ASCII text width of the for the column entries (without the string
      *                                      termination).
      *
      * @return                          the number of columns after this one is added.
      *
      * @throws IllegalArgumentException if the column data is not an array or the specified text <code>width</code> is
      *                                      &le;1.
      * @throws FitsException            if the column us of an unsupported data type or if the number of entries does
      *                                      not match the number of rows already contained in the table.
      * 
      * @see                             #addColumn(Object)
      */
     public int addColumn(Object newCol, int width) throws FitsException, IllegalArgumentException {
         if (width < 1) {
             throw new IllegalArgumentException("Illegal ASCII column width: " + width);
         }
 
         if (!newCol.getClass().isArray()) {
             throw new IllegalArgumentException("Not an array: " + newCol.getClass().getName());
         }
 
         if (nFields > 0 && Array.getLength(newCol) != nRows) {
             throw new FitsException(
                     "Mismatched number of rows: expected " + nRows + ", got " + Array.getLength(newCol) + "rows.");
         }
 
         if (nFields == 0) {
             nRows = Array.getLength(newCol);
         }
 
         Class<?> type = ArrayFuncs.getBaseClass(newCol);
         if (type != int.class && type != long.class && type != float.class && type != double.class
                 && type != String.class) {
             throw new FitsException("No AsciiTable support for elements of " + type.getName());
         }
 
         data = Arrays.copyOf(data, nFields + 1);
         offsets = Arrays.copyOf(offsets, nFields + 1);
         lengths = Arrays.copyOf(lengths, nFields + 1);
         types = Arrays.copyOf(types, nFields + 1);
         nulls = Arrays.copyOf(nulls, nFields + 1);
         names = Arrays.copyOf(names, nFields + 1);
 
         data[nFields] = newCol;
         offsets[nFields] = rowLen + 1;
         lengths[nFields] = width;
         types[nFields] = ArrayFuncs.getBaseClass(newCol);
         names[nFields] = TableHDU.getDefaultColumnName(nFields);
 
         rowLen += width + 1;
         if (isNull != null) {
             boolean[] newIsNull = new boolean[nRows * (nFields + 1)];
             // Fix the null pointers.
             int add = 0;
             for (int i = 0; i < isNull.length; i++) {
                 if (i % nFields == 0) {
                     add++;
                 }
                 if (isNull[i]) {
                     newIsNull[i + add] = true;
                 }
             }
             isNull = newIsNull;
         }
         nFields++;
 
         // Invalidate the buffer
         buffer = null;
 
         return nFields;
     }
 
     /**
      * Beware that adding rows to ASCII tables may be very inefficient. Avoid addding more than a few rows if you can.
      */
     @Override
     public int addRow(Object[] newRow) throws FitsException {
         try {
             // If there are no fields, then this is the
             // first row. We need to add in each of the columns
             // to get the descriptors set up.
             if (nFields == 0) {
                 for (Object element : newRow) {
                     addColumn(element);
                 }
             } else {
                 for (int i = 0; i < nFields; i++) {
                     Object o = ArrayFuncs.newInstance(types[i], nRows + 1);
                     System.arraycopy(data[i], 0, o, 0, nRows);
                     System.arraycopy(newRow[i], 0, o, nRows, 1);
                     data[i] = o;
                 }
                 nRows++;
             }
             // Invalidate the buffer
             buffer = null;
             return nRows;
         } catch (Exception e) {
             throw new FitsException("Error adding row:" + e.getMessage(), e);
         }
     }
 
     @Override
     public void deleteColumns(int start, int len) throws FitsException {
         ensureData();
 
         Object[] newData = new Object[nFields - len];
         int[] newOffsets = new int[nFields - len];
         int[] newLengths = new int[nFields - len];
         Class<?>[] newTypes = new Class[nFields - len];
         String[] newNulls = new String[nFields - len];
 
         // Copy in the initial stuff...
         System.arraycopy(data, 0, newData, 0, start);
         // Don't do the offsets here.
         System.arraycopy(lengths, 0, newLengths, 0, start);
         System.arraycopy(types, 0, newTypes, 0, start);
         System.arraycopy(nulls, 0, newNulls, 0, start);
 
         // Copy in the final
         System.arraycopy(data, start + len, newData, start, nFields - start - len);
         // Don't do the offsets here.
         System.arraycopy(lengths, start + len, newLengths, start, nFields - start - len);
         System.arraycopy(types, start + len, newTypes, start, nFields - start - len);
         System.arraycopy(nulls, start + len, newNulls, start, nFields - start - len);
 
         for (int i = start; i < start + len; i++) {
             rowLen -= lengths[i] + 1;
         }
 
         data = newData;
         offsets = newOffsets;
         lengths = newLengths;
         types = newTypes;
         nulls = newNulls;
 
         if (isNull != null) {
             boolean found = false;
 
             boolean[] newIsNull = new boolean[nRows * (nFields - len)];
             for (int i = 0; i < nRows; i++) {
                 int oldOff = nFields * i;
                 int newOff = (nFields - len) * i;
                 for (int col = 0; col < start; col++) {
                     newIsNull[newOff + col] = isNull[oldOff + col];
                     found = found || isNull[oldOff + col];
                 }
                 for (int col = start + len; col < nFields; col++) {
                     newIsNull[newOff + col - len] = isNull[oldOff + col];
                     found = found || isNull[oldOff + col];
                 }
             }
             if (found) {
                 isNull = newIsNull;
             } else {
                 isNull = null;
             }
         }
 
         // Invalidate the buffer
         buffer = null;
 
         nFields -= len;
     }
 
     /**
      * Beware that repeatedly deleting rows from ASCII tables may be very inefficient. Avoid calling this more than once
      * (or a few times) if you can.
      */
     @Override
     public void deleteRows(int start, int len) throws FitsException {
         if (nRows == 0 || start < 0 || start >= nRows || len <= 0) {
             return;
         }
         if (start + len > nRows) {
             len = nRows - start;
         }
 
         ensureData();
 
         for (int i = 0; i < nFields; i++) {
             try {
                 Object o = ArrayFuncs.newInstance(types[i], nRows - len);
                 System.arraycopy(data[i], 0, o, 0, start);
                 System.arraycopy(data[i], start + len, o, start, nRows - len - start);
                 data[i] = o;
             } catch (Exception e) {
                 throw new FitsException("Error deleting row: " + e.getMessage(), e);
             }
 
         }
         nRows -= len;
     }
 
     @Override
     protected void loadData(ArrayDataInput in) throws IOException, FitsException {
         currInput = in;
 
         if (buffer == null) {
             getBuffer((long) nRows * rowLen, 0);
         }
 
         data = new Object[nFields];
         for (int i = 0; i < nFields; i++) {
             data[i] = ArrayFuncs.newInstance(types[i], nRows);
         }
 
         bp.setOffset(0);
 
         int rowOffset;
         for (int i = 0; i < nRows; i++) {
             rowOffset = rowLen * i;
             for (int j = 0; j < nFields; j++) {
                 try {
                     if (!extractElement(rowOffset + offsets[j], lengths[j], data, j, i, nulls[j])) {
                         if (isNull == null) {
                             isNull = new boolean[nRows * nFields];
                         }
 
                         isNull[j + i * nFields] = true;
                     }
                 } catch (ArrayIndexOutOfBoundsException e) {
                     throw new FitsException("not enough data: " + e, e);
                 }
             }
         }
     }
 
     @Override
     public void read(ArrayDataInput in) throws FitsException {
         currInput = in;
         super.read(in);
     }
 
     /**
      * Move an element from the buffer into a data array.
      *
      * @param  offset        The offset within buffer at which the element starts.
      * @param  length        The number of bytes in the buffer for the element.
      * @param  array         An array of objects, each of which is a simple array.
      * @param  col           Which element of array is to be modified?
      * @param  row           Which index into that element is to be modified?
      * @param  nullFld       What string signifies a null element?
      *
      * @throws FitsException if the operation failed
      */
     private boolean extractElement(int offset, int length, Object[] array, int col, int row, String nullFld)
             throws FitsException {
 
         bp.setOffset(offset);
 
         if (nullFld != null) {
             String s = bp.getString(length);
             if (s.trim().equals(nullFld)) {
                 return false;
             }
             bp.skip(-length);
         }
         try {
             if (array[col] instanceof String[]) {
                 ((String[]) array[col])[row] = bp.getString(length);
             } else if (array[col] instanceof int[]) {
                 ((int[]) array[col])[row] = bp.getInt(length);
             } else if (array[col] instanceof float[]) {
                 ((float[]) array[col])[row] = bp.getFloat(length);
             } else if (array[col] instanceof double[]) {
                 ((double[]) array[col])[row] = bp.getDouble(length);
             } else if (array[col] instanceof long[]) {
                 ((long[]) array[col])[row] = bp.getLong(length);
             } else {
                 throw new FitsException("Invalid type for ASCII table conversion:" + array[col]);
             }
         } catch (FormatException e) {
             throw new FitsException("Error parsing data at row,col:" + row + "," + col + "  ", e);
         }
         return true;
     }
 
     @Override
     protected void fillHeader(Header h) {
         h.deleteKey(Standard.SIMPLE);
         h.deleteKey(Standard.EXTEND);
 
         Standard.context(AsciiTable.class);
 
         Cursor<String, HeaderCard> c = h.iterator();
         c.add(HeaderCard.create(Standard.XTENSION, Standard.XTENSION_ASCIITABLE));
         c.add(HeaderCard.create(Standard.BITPIX, Bitpix.BYTE.getHeaderValue()));
         c.add(HeaderCard.create(Standard.NAXIS, 2));
         c.add(HeaderCard.create(Standard.NAXIS1, rowLen));
         c.add(HeaderCard.create(Standard.NAXIS2, nRows));
         c.add(HeaderCard.create(Standard.PCOUNT, 0));
         c.add(HeaderCard.create(Standard.GCOUNT, 1));
         c.add(HeaderCard.create(Standard.TFIELDS, nFields));
 
         for (int i = 0; i < nFields; i++) {
             addColInfo(i, c);
         }
 
         Standard.context(null);
     }
 
     /**
      * Read some data into the buffer.
      */
     private void getBuffer(long size, long offset) throws IOException, FitsException {
 
         if (currInput == null) {
             throw new IOException("No stream open to read");
         }
 
         if (size > Integer.MAX_VALUE) {
             throw new FitsException("Cannot read ASCII table > 2 GB");
         }
 
         buffer = new byte[(int) size];
         if (offset != 0) {
             FitsUtil.reposition(currInput, offset);
         }
         currInput.readFully(buffer);
         bp = new ByteParser(buffer);
     }
 
     /**
      * <p>
      * Returns the data for a particular column in as a flattened 1D array of elements. See {@link #addColumn(Object)}
      * for more information about the format of data elements in general.
      * </p>
      * 
      * @param  col           The 0-based column index.
      * 
      * @return               an array of primitives (for scalar columns), or else an <code>Object[]</code> array.
      * 
      * @throws FitsException if the table could not be accessed
      *
      * @see                  #setColumn(int, Object)
      * @see                  #getElement(int, int)
      * @see                  #getNCols()
      */
     @Override
     public Object getColumn(int col) throws FitsException {
         ensureData();
         return data[col];
     }
 
     @Override
     protected Object[] getCurrentData() {
         return data;
     }
 
     @Override
     public Object[] getData() throws FitsException {
         return (Object[]) super.getData();
     }
 
     @Override
     public Object getElement(int row, int col) throws FitsException {
         if (data != null) {
             return singleElement(row, col);
         }
         return parseSingleElement(row, col);
     }
 
     @Override
     public int getNCols() {
         return nFields;
     }
 
     @Override
     public int getNRows() {
         return nRows;
     }
 
     @Override
     public Object[] getRow(int row) throws FitsException {
 
         if (data != null) {
             return singleRow(row);
         }
         return parseSingleRow(row);
     }
 
     /**
      * Get the number of bytes in a row
      *
      * @return The number of bytes for a single row in the table.
      */
     public int getRowLen() {
         return rowLen;
     }
 
     @Override
     protected long getTrueSize() {
         return (long) nRows * rowLen;
     }
 
     /**
      * Checks if an element is <code>null</code>.
      *
      * @param  row The 0-based row
      * @param  col The 0-based column
      *
      * @return     if the given element has been nulled.
      */
     public boolean isNull(int row, int col) {
         if (isNull != null) {
             return isNull[row * nFields + col];
         }
         return false;
     }
 
     /**
      * Read a single element from the table. This returns an array of dimension 1.
      *
      * @throws FitsException if the operation failed
      */
     private Object parseSingleElement(int row, int col) throws FitsException {
 
         Object[] res = new Object[1];
         try {
             getBuffer(lengths[col], getFileOffset() + (long) row * (long) rowLen + offsets[col]);
         } catch (IOException e) {
             buffer = null;
             throw new FitsException("Unable to read element", e);
         }
         res[0] = ArrayFuncs.newInstance(types[col], 1);
 
         boolean success = extractElement(0, lengths[col], res, 0, 0, nulls[col]);
         buffer = null;
 
         return success ? res[0] : null;
     }
 
     /**
      * Read a single row from the table. This returns a set of arrays of dimension 1.
      *
      * @throws FitsException if the operation failed
      */
     private Object[] parseSingleRow(int row) throws FitsException {
 
         Object[] res = new Object[nFields];
 
         try {
             getBuffer(rowLen, getFileOffset() + (long) row * (long) rowLen);
         } catch (IOException e) {
             throw new FitsException("Unable to read row", e);
         }
 
         for (int i = 0; i < nFields; i++) {
             res[i] = ArrayFuncs.newInstance(types[i], 1);
             if (!extractElement(offsets[i], lengths[i], res, i, 0, nulls[i])) {
                 res[i] = null;
             }
         }
 
         // Invalidate buffer for future use.
         buffer = null;
         return res;
     }
 
     @Override
     public void setColumn(int col, Object newData) throws FitsException {
         ensureData();
         if (col < 0 || col >= nFields || newData.getClass() != data[col].getClass()
                 || Array.getLength(newData) != Array.getLength(data[col])) {
             throw new FitsException("Invalid column/column mismatch:" + col);
         }
         data[col] = newData;
 
         // Invalidate the buffer.
         buffer = null;
     }
 
     @Override
     public void setElement(int row, int col, Object newData) throws FitsException {
         ensureData();
         try {
             System.arraycopy(newData, 0, data[col], row, 1);
         } catch (Exception e) {
             throw new FitsException("Incompatible element:" + row + "," + col, e);
         }
         setNull(row, col, false);
 
         // Invalidate the buffer
         buffer = null;
 
     }
 
     /**
      * Mark (or unmark) an element as null. Note that if this FITS file is latter written out, a TNULL keyword needs to
      * be defined in the corresponding header. This routine does not add an element for String columns.
      *
      * @param row  The 0-based row.
      * @param col  The 0-based column.
      * @param flag True if the element is to be set to null.
      */
     public void setNull(int row, int col, boolean flag) {
         if (flag) {
             if (isNull == null) {
                 isNull = new boolean[nRows * nFields];
             }
             isNull[col + row * nFields] = true;
         } else if (isNull != null) {
             isNull[col + row * nFields] = false;
         }
 
         // Invalidate the buffer
         buffer = null;
     }
 
     /**
      * Set the null string for a columns. This is not a public method since we want users to call the method in
      * AsciiTableHDU and update the header also.
      */
     void setNullString(int col, String newNull) {
         if (col >= 0 && col < nulls.length) {
             nulls[col] = newNull;
         }
     }
 
     @Override
     public void setRow(int row, Object[] newData) throws FitsException {
         if (row < 0 || row > nRows) {
             throw new FitsException("Invalid row in setRow");
         }
         ensureData();
         for (int i = 0; i < nFields; i++) {
             try {
                 System.arraycopy(newData[i], 0, data[i], row, 1);
             } catch (Exception e) {
                 throw new FitsException("Unable to modify row: incompatible data:" + row, e);
             }
             setNull(row, i, false);
         }
 
         // Invalidate the buffer
         buffer = null;
 
     }
 
     /**
      * Extract a single element from a table. This returns an array of length 1.
      */
     private Object singleElement(int row, int col) {
 
         Object res = null;
         if (isNull == null || !isNull[row * nFields + col]) {
             res = ArrayFuncs.newInstance(types[col], 1);
             System.arraycopy(data[col], row, res, 0, 1);
         }
         return res;
     }
 
     /**
      * Extract a single row from a table. This returns an array of Objects each of which is an array of length 1.
      */
     private Object[] singleRow(int row) {
 
         Object[] res = new Object[nFields];
         for (int i = 0; i < nFields; i++) {
             if (isNull == null || !isNull[row * nFields + i]) {
                 res[i] = ArrayFuncs.newInstance(types[i], 1);
                 System.arraycopy(data[i], row, res[i], 0, 1);
             }
         }
         return res;
     }
 
     /**
      * @deprecated It is not entirely foolproof for keeping the header in sync -- it is better to (re)wrap tables in a
      *                 new HDU and editing the header as necessary to incorporate custom entries. May be removed from
      *                 the API in the future.
      */
     @Deprecated
     @Override
     public void updateAfterDelete(int oldNCol, Header hdr) throws FitsException {
 
         int offset = 0;
         for (int i = 0; i < nFields; i++) {
             offsets[i] = offset;
             hdr.addValue(TBCOLn.n(i + 1), offset + 1);
             offset += lengths[i] + 1;
         }
         for (int i = nFields; i < oldNCol; i++) {
             hdr.deleteKey(TBCOLn.n(i + 1));
         }
 
         hdr.addValue(NAXIS1, rowLen);
     }
 
     @Override
     public void write(ArrayDataOutput str) throws FitsException {
         // Make sure we have the data in hand.
         if (str != currInput) {
             ensureData();
         }
 
         // If buffer is still around we can just reuse it,
         // since nothing we've done has invalidated it.
         if (data == null) {
             throw new FitsException("Attempt to write undefined ASCII Table");
         }
 
         if ((long) nRows * rowLen > Integer.MAX_VALUE) {
             throw new FitsException("Cannot write ASCII table > 2 GB");
         }
 
         buffer = new byte[nRows * rowLen];
 
         bp = new ByteParser(buffer);
         for (int i = 0; i < buffer.length; i++) {
             buffer[i] = (byte) ' ';
         }
 
         ByteFormatter bf = new ByteFormatter();
 
         for (int i = 0; i < nRows; i++) {
 
             for (int j = 0; j < nFields; j++) {
                 int offset = i * rowLen + offsets[j];
                 int len = lengths[j];
                 if (isNull != null && isNull[i * nFields + j]) {
                     if (nulls[j] == null) {
                         throw new FitsException("No null value set when needed");
                     }
                     bf.format(nulls[j], buffer, offset, len);
                 } else if (types[j] == String.class) {
                     String[] s = (String[]) data[j];
                     bf.format(s[i], buffer, offset, len);
                 } else if (types[j] == int.class) {
                     int[] ia = (int[]) data[j];
                     bf.format(ia[i], buffer, offset, len);
                 } else if (types[j] == float.class) {
                     float[] fa = (float[]) data[j];
                     bf.format(fa[i], buffer, offset, len);
                 } else if (types[j] == double.class) {
                     double[] da = (double[]) data[j];
                     bf.format(da[i], buffer, offset, len);
                 } else if (types[j] == long.class) {
                     long[] la = (long[]) data[j];
                     bf.format(la[i], buffer, offset, len);
                 }
             }
         }
 
         // Now write the buffer.
         try {
             str.write(buffer);
             FitsUtil.pad(str, buffer.length, (byte) ' ');
         } catch (IOException e) {
             throw new FitsException("Error writing ASCII Table data", e);
         }
     }
 
     @Override
     public AsciiTableHDU toHDU() {
         Header h = new Header();
         fillHeader(h);
         return new AsciiTableHDU(h, this);
     }
 
     /**
      * <p>
      * Controls how columns with format "<code>I10</code>" are handled; this is tricky because some, but not all,
      * integers that can be represented in 10 characters form 32-bit integers. Setting it <code>true</code> may make it
      * more likely to avoid unexpected type changes during round-tripping, but it also means that some values in I10
      * columns may be impossible to read. The default behavior is to assume <code>true</code>, and thus to treat I10
      * columns as <code>int</code> values.
      * </p>
      * 
      * @param value if <code>true</code>, format "I10" columns will be assumed <code>int.class</code>, provided
      *                  TLMINn/TLMAXn or TDMINn/TDMAXn limits (if defined) allow it. if <code>false</code>, I10 columns
      *                  that have no clear indication of data range will be assumed <code>long.class</code>.
      *
      * @since       1.19
      * 
      * @see         AsciiTable#isI10PreferInt()
      */
     public static void setI10PreferInt(boolean value) {
         isI10PreferInt = value;
     }
 
     /**
      * Checks if I10 columns should be treated as containing 32-bit <code>int</code> values, rather than 64-bit
      * <code>long</code> values, when possible.
      * 
      * @return <code>true</code> if I10 columns should be treated as containing 32-bit <code>int</code> values,
      *             otherwise <code>false</code>.
      * 
      * @since  1.19
      * 
      * @see    #setI10PreferInt(boolean)
      */
     public static boolean isI10PreferInt() {
         return isI10PreferInt;
     }
 }