# HG changeset patch
# User Jaroslav Hajek <highegg@gmail.com>
# Date 1254204250 -7200
# Node ID 5627b5865d99c3107e1bc51e8f44510f552cdab1
# Parent  63249224f78d0f25a40fcbfbd5191f44b5a42123
implement reinterpret, bitpack & bitunpack

diff -r 63249224f78d -r 5627b5865d99 src/ChangeLog
--- a/src/ChangeLog	Mon Sep 28 21:52:27 2009 -0400
+++ b/src/ChangeLog	Tue Sep 29 08:04:10 2009 +0200
@@ -1,3 +1,8 @@
+2009-09-29  Jaroslav Hajek  <highegg@gmail.com>
+
+	* DLD-FUNCTIONS/reinterpret.cc: New source.
+	* Makefile.in: Add it here.
+
 2009-09-28  Shai Ayal  <shaiay@users.sourceforge.net>
 
 	* DLD-FUNCTIONS/fltk_backend.cc (plot_window::handle): Redraw the
diff -r 63249224f78d -r 5627b5865d99 src/DLD-FUNCTIONS/reinterpret.cc
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/src/DLD-FUNCTIONS/reinterpret.cc	Tue Sep 29 08:04:10 2009 +0200
@@ -0,0 +1,424 @@
+/*
+
+Copyright (C) 2009 VZLU Prague
+
+This file is part of Octave.
+
+Octave is free software; you can redistribute it and/or modify it
+under the terms of the GNU General Public License as published by the
+Free Software Foundation; either version 3 of the License, or (at your
+option) any later version.
+
+Octave is distributed in the hope that it will be useful, but WITHOUT
+ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+for more details.
+
+You should have received a copy of the GNU General Public License
+along with Octave; see the file COPYING.  If not, see
+<http://www.gnu.org/licenses/>.
+
+*/
+
+#ifdef HAVE_CONFIG_H
+#include <config.h>
+#endif
+
+#include "mx-base.h"
+
+#include "defun-dld.h"
+#include "error.h"
+#include "gripes.h"
+#include "oct-obj.h"
+#include "unwind-prot.h"
+
+static dim_vector
+get_vec_dims (const dim_vector& old_dims, octave_idx_type n)
+{
+  if (old_dims.length () == 2 && old_dims(0) == 1)
+    return dim_vector (1, n);
+  else if (old_dims.length () == 2 && old_dims (0) == 0 && old_dims (1) == 0)
+    return dim_vector ();
+  else
+    return dim_vector (n, 1);
+}
+
+template <class ArrayType>
+static void 
+get_data_and_bytesize (const ArrayType& array,
+                       const void *& data,
+                       octave_idx_type& byte_size,
+                       dim_vector& old_dims)
+{
+  // The array given may be a temporary, constructed from a scalar or sparse
+  // array. This will ensure the data will be deallocated after we exit.
+  unwind_protect::add_delete (new ArrayType (array));
+
+  data = reinterpret_cast<const void *> (array.data ());
+  byte_size = array.byte_size ();
+
+  old_dims = array.dims ();
+}
+
+template <class ArrayType>
+static ArrayType
+reinterpret_copy (const void *data, octave_idx_type byte_size,
+                  const dim_vector& old_dims)
+{
+  typedef typename ArrayType::element_type T;
+  octave_idx_type n = byte_size / sizeof (T);
+
+  if (n * static_cast<int> (sizeof (T)) < byte_size)
+    warning ("reinterpret: discarding excess data");
+
+  ArrayType retval (get_vec_dims (old_dims, n));
+  T *dest = retval.fortran_vec ();
+  std::memcpy (dest, data, n * sizeof (T));
+
+  return retval;
+}
+
+DEFUN_DLD (reinterpret, args, ,
+  "-*- texinfo -*-\n\
+@deftypefn {Loadable Function} {@var{y} =} reinterpret (@var{x}, @var{class})\n\
+Returns a new array @var{y} resulting from interpreting the data of\n\
+@var{x} in memory as data of the numeric class @var{class}. Both the class\n\
+of @var{x} and @var{class} must be one of the built-in numeric classes:\n\
+\n\
+@example\n\
+  \"logical\"\n\
+  \"char\"\n\
+  \"int8\"\n\
+  \"int16\"\n\
+  \"int32\"\n\
+  \"int64\"\n\
+  \"uint8\"\n\
+  \"uint16\"\n\
+  \"uint32\"\n\
+  \"uint64\"\n\
+  \"double\"\n\
+  \"single\"\n\
+  \"double complex\"\n\
+  \"single complex\"\n\
+@end example\n\
+\n\
+the last two are reserved for @var{class}; they indicate that a complex-valued result\n\
+is requested. Complex arrays are stored in memory as consecutive pairs of real numbers.\n\
+The sizes of integer types are given by their bit counts. Both logical and char are typically\n\
+one byte wide; however, this is not guaranteed by C++. If your system is IEEE conformant,\n\
+single and double should be 4 bytes and 8 bytes wide, respectively.\n\
+\"logical\" is not allowed for @var{class}.\n\
+If the input is a row vector, the return value is a row vector, otherwise it is a column vector.\n\
+If the bit length of @var{x} is not divisible by that of @var{class}, excess bits are discarded.\n\
+@seealso{bitunpack,bitpack}\n\
+@end deftypefn")
+{
+  octave_value retval;
+
+  if (args.length () == 2)
+    {
+      const void *data = 0;
+      octave_idx_type byte_size = 0;
+      dim_vector old_dims;
+
+      octave_value array = args(0);
+
+      if (array.is_bool_type ())
+        get_data_and_bytesize (array.bool_array_value (), data, byte_size, old_dims);
+      else if (array.is_string ())
+        get_data_and_bytesize (array.char_array_value (), data, byte_size, old_dims);
+      else if (array.is_integer_type ())
+        {
+          if (array.is_int8_type ())
+            get_data_and_bytesize (array.int8_array_value (), data, byte_size, old_dims);
+          else if (array.is_int16_type ())
+            get_data_and_bytesize (array.int16_array_value (), data, byte_size, old_dims);
+          else if (array.is_int32_type ())
+            get_data_and_bytesize (array.int32_array_value (), data, byte_size, old_dims);
+          else if (array.is_int64_type ())
+            get_data_and_bytesize (array.int64_array_value (), data, byte_size, old_dims);
+          else if (array.is_uint8_type ())
+            get_data_and_bytesize (array.uint8_array_value (), data, byte_size, old_dims);
+          else if (array.is_uint16_type ())
+            get_data_and_bytesize (array.uint16_array_value (), data, byte_size, old_dims);
+          else if (array.is_uint32_type ())
+            get_data_and_bytesize (array.uint32_array_value (), data, byte_size, old_dims);
+          else if (array.is_uint64_type ())
+            get_data_and_bytesize (array.uint64_array_value (), data, byte_size, old_dims);
+          else
+            assert (0);
+        }
+      else if (array.is_complex_type ())
+        {
+          if (array.is_single_type ())
+            get_data_and_bytesize (array.float_complex_array_value (), data, byte_size, old_dims);
+          else
+            get_data_and_bytesize (array.complex_array_value (), data, byte_size, old_dims);
+        }
+      else if (array.is_real_type ())
+        {
+          if (array.is_single_type ())
+            get_data_and_bytesize (array.float_array_value (), data, byte_size, old_dims);
+          else
+            get_data_and_bytesize (array.array_value (), data, byte_size, old_dims);
+        }
+      else
+        error ("reinterpret: invalid input class: %s", array.class_name ().c_str ());
+
+      std::string numclass = args(1).string_value ();
+
+      if (error_state || numclass.size () == 0)
+        ;
+      else if (numclass == "char")
+        retval = octave_value (reinterpret_copy<charNDArray> (data, byte_size, old_dims), true);
+      else if (numclass[0] == 'i')
+        {
+          if (numclass == "int8")
+            retval = reinterpret_copy<int8NDArray> (data, byte_size, old_dims);
+          else if (numclass == "int16")
+            retval = reinterpret_copy<int16NDArray> (data, byte_size, old_dims);
+          else if (numclass == "int32")
+            retval = reinterpret_copy<int32NDArray> (data, byte_size, old_dims);
+          else if (numclass == "int64")
+            retval = reinterpret_copy<int64NDArray> (data, byte_size, old_dims);
+        }
+      else if (numclass[0] == 'u')
+        {
+          if (numclass == "uint8")
+            retval = reinterpret_copy<uint8NDArray> (data, byte_size, old_dims);
+          else if (numclass == "uint16")
+            retval = reinterpret_copy<uint16NDArray> (data, byte_size, old_dims);
+          else if (numclass == "uint32")
+            retval = reinterpret_copy<uint32NDArray> (data, byte_size, old_dims);
+          else if (numclass == "uint64")
+            retval = reinterpret_copy<uint64NDArray> (data, byte_size, old_dims);
+        }
+      else if (numclass == "single")
+        retval = reinterpret_copy<FloatNDArray> (data, byte_size, old_dims);
+      else if (numclass == "double")
+        retval = reinterpret_copy<NDArray> (data, byte_size, old_dims);
+      else if (numclass == "single complex")
+        retval = reinterpret_copy<FloatComplexNDArray> (data, byte_size, old_dims);
+      else if (numclass == "double complex")
+        retval = reinterpret_copy<ComplexNDArray> (data, byte_size, old_dims);
+
+      if (! error_state && retval.is_undefined ())
+        error ("reinterpret: cannot convert to %s class", numclass.c_str ());
+    }
+  else
+    print_usage ();
+
+  return retval;
+}
+
+template <class ArrayType>
+ArrayType
+do_bitpack (const boolNDArray& bitp)
+{
+  typedef typename ArrayType::element_type T;
+  octave_idx_type n = bitp.numel () / (sizeof (T) * CHAR_BIT);
+
+  if (n * static_cast<int> (sizeof (T)) * CHAR_BIT < bitp.numel ())
+    warning ("bitpack: discarding excess data");
+
+  ArrayType retval (get_vec_dims (bitp.dims (), n));
+
+  const bool *bits = bitp.fortran_vec ();
+  char *packed = reinterpret_cast<char *> (retval.fortran_vec ());
+
+  octave_idx_type m = n * sizeof (T);
+
+  for (octave_idx_type i = 0; i < m; i++)
+    {
+      char c = bits[0];
+      for (int j = 1; j < CHAR_BIT; j++)
+        c |= bits[j] << j;
+
+      packed[i] = c;
+      bits += CHAR_BIT;
+    }
+
+  return retval;
+}
+
+DEFUN_DLD (bitpack, args, ,
+  "-*- texinfo -*-\n\
+@deftypefn {Loadable Function} {@var{y} =} bitpack (@var{x}, @var{class})\n\
+Returns a new array @var{y} resulting from interpreting a logical array @var{x}\n\
+as raw bit pattern for data of the numeric class @var{class}. @var{class} must be\n\
+one of the built-in numeric classes:\n\
+\n\
+@example\n\
+  \"char\"\n\
+  \"int8\"\n\
+  \"int16\"\n\
+  \"int32\"\n\
+  \"int64\"\n\
+  \"uint8\"\n\
+  \"uint16\"\n\
+  \"uint32\"\n\
+  \"uint64\"\n\
+  \"double\"\n\
+  \"single\"\n\
+@end example\n\
+\n\
+The number of elements of @var{x} should be divisible by the bit length of @var{class}.\n\
+If it is not, excess bits are discarded. Bits come in increasing order of significance, i.e.\n\
+@code{x(1)} is bit 0, @code{x(2)} is bit 1, etc.\n\
+The result is a row vector if @var{x} is a row vector, otherwise it is a column vector.\n\
+@seealso{bitunpack,reinterpret,bitget,bitset}\n\
+@end deftypefn")
+{
+  octave_value retval;
+
+  if (args.length () == 2 && args(0).is_bool_type ())
+    {
+      boolNDArray bitp = args(0).bool_array_value ();
+
+      std::string numclass = args(1).string_value ();
+
+      if (error_state || numclass.size () == 0)
+        ;
+      else if (numclass == "char")
+        retval = octave_value (do_bitpack<charNDArray> (bitp), true);
+      else if (numclass[0] == 'i')
+        {
+          if (numclass == "int8")
+            retval = do_bitpack<int8NDArray> (bitp);
+          else if (numclass == "int16")
+            retval = do_bitpack<int16NDArray> (bitp);
+          else if (numclass == "int32")
+            retval = do_bitpack<int32NDArray> (bitp);
+          else if (numclass == "int64")
+            retval = do_bitpack<int64NDArray> (bitp);
+        }
+      else if (numclass[0] == 'u')
+        {
+          if (numclass == "uint8")
+            retval = do_bitpack<uint8NDArray> (bitp);
+          else if (numclass == "uint16")
+            retval = do_bitpack<uint16NDArray> (bitp);
+          else if (numclass == "uint32")
+            retval = do_bitpack<uint32NDArray> (bitp);
+          else if (numclass == "uint64")
+            retval = do_bitpack<uint64NDArray> (bitp);
+        }
+      else if (numclass == "single")
+        retval = do_bitpack<FloatNDArray> (bitp);
+      else if (numclass == "double")
+        retval = do_bitpack<NDArray> (bitp);
+      else if (numclass == "single complex")
+        retval = do_bitpack<FloatComplexNDArray> (bitp);
+      else if (numclass == "double complex")
+        retval = do_bitpack<ComplexNDArray> (bitp);
+
+      if (! error_state && retval.is_undefined ())
+        error ("bitpack: cannot pack to %s class", numclass.c_str ());
+    }
+  else
+    print_usage ();
+
+  return retval;
+}
+
+template <class ArrayType>
+boolNDArray
+do_bitunpack (const ArrayType& array)
+{
+  typedef typename ArrayType::element_type T;
+  octave_idx_type n = array.numel () * sizeof (T) * CHAR_BIT;
+
+  boolNDArray retval (get_vec_dims (array.dims (), n));
+
+  const char *packed = reinterpret_cast<const char *> (array.fortran_vec ());
+  bool *bits = retval.fortran_vec ();
+
+  octave_idx_type m = n / CHAR_BIT;
+
+  for (octave_idx_type i = 0; i < m; i++)
+    {
+      char c = packed[i];
+      bits[0] = c & 1;
+      for (int j = 1; j < CHAR_BIT; j++)
+        bits[j] = (c >>= 1) & 1;
+      bits += CHAR_BIT;
+    }
+
+  return retval;
+}
+
+DEFUN_DLD (bitunpack, args, ,
+  "-*- texinfo -*-\n\
+@deftypefn {Loadable Function} {@var{y} =} bitpack (@var{x})\n\
+Returns a logical array @var{y} corresponding to the raw bit pattern of @var{x}.\n\
+@var{x} must belong to one of the built-in numeric classes:\n\
+\n\
+@example\n\
+  \"char\"\n\
+  \"int8\"\n\
+  \"int16\"\n\
+  \"int32\"\n\
+  \"int64\"\n\
+  \"uint8\"\n\
+  \"uint16\"\n\
+  \"uint32\"\n\
+  \"uint64\"\n\
+  \"double\"\n\
+  \"single\"\n\
+@end example\n\
+\n\
+The result is a row vector if @var{x} is a row vector, otherwise it is a column vector.\n\
+@seealso{bitpack,reinterpret,bitget,bitset}\n\
+@end deftypefn")
+{
+  octave_value retval;
+
+  if (args.length () == 1 && (args(0).is_numeric_type () || args(0).is_string ()))
+    {
+      octave_value array = args(0);
+
+      if (array.is_string ())
+        retval = do_bitunpack (array.char_array_value ());
+      else if (array.is_integer_type ())
+        {
+          if (array.is_int8_type ())
+            retval = do_bitunpack (array.int8_array_value ());
+          else if (array.is_int16_type ())
+            retval = do_bitunpack (array.int16_array_value ());
+          else if (array.is_int32_type ())
+            retval = do_bitunpack (array.int32_array_value ());
+          else if (array.is_int64_type ())
+            retval = do_bitunpack (array.int64_array_value ());
+          else if (array.is_uint8_type ())
+            retval = do_bitunpack (array.uint8_array_value ());
+          else if (array.is_uint16_type ())
+            retval = do_bitunpack (array.uint16_array_value ());
+          else if (array.is_uint32_type ())
+            retval = do_bitunpack (array.uint32_array_value ());
+          else if (array.is_uint64_type ())
+            retval = do_bitunpack (array.uint64_array_value ());
+          else
+            assert (0);
+        }
+      else if (array.is_complex_type ())
+        {
+          if (array.is_single_type ())
+            retval = do_bitunpack (array.float_complex_array_value ());
+          else
+            retval = do_bitunpack (array.complex_array_value ());
+        }
+      else if (array.is_real_type ())
+        {
+          if (array.is_single_type ())
+            retval = do_bitunpack (array.float_array_value ());
+          else
+            retval = do_bitunpack (array.array_value ());
+        }
+      else
+        error ("bitunpack: invalid input class: %s", array.class_name ().c_str ());
+    }
+  else
+    print_usage ();
+
+  return retval;
+}
diff -r 63249224f78d -r 5627b5865d99 src/Makefile.in
--- a/src/Makefile.in	Mon Sep 28 21:52:27 2009 -0400
+++ b/src/Makefile.in	Tue Sep 29 08:04:10 2009 +0200
@@ -71,7 +71,7 @@
 	gammainc.cc gcd.cc getgrent.cc getpwent.cc getrusage.cc \
 	givens.cc hess.cc hex2num.cc inv.cc kron.cc lookup.cc \
 	lsode.cc lu.cc luinc.cc matrix_type.cc max.cc md5sum.cc \
-	pinv.cc qr.cc quad.cc qz.cc rand.cc rcond.cc regexp.cc \
+	pinv.cc qr.cc quad.cc qz.cc rand.cc rcond.cc regexp.cc reinterpret.cc \
 	schur.cc sparse.cc spparms.cc sqrtm.cc sub2ind.cc svd.cc syl.cc \
 	symrcm.cc symbfact.cc time.cc tsearch.cc typecast.cc \
 	urlwrite.cc __contourc__.cc __delaunayn__.cc \