carray 1.4.0 → 1.5.1

data/ext/calculus/lib/math/interp/adapter_interp1d.rb

@@ -1,31 +0,0 @@
-# ----------------------------------------------------------------------------
-#
-#  carray/math/interp/adapter_interp1d.rb
-#
-#  This file is part of Ruby/CArray extension library.
-#  You can redistribute it and/or modify it under the terms of
-#  the Ruby Licence.
-#
-#  Copyright (C) 2005 Hiroki Motoyoshi
-#
-# ----------------------------------------------------------------------------
-
-require "carray/math/interp"
-
-class CA::Interp::CAInterp1D < CA::Interp::Adapter
-  
-  install_adapter "interp1d"
-  
-  def initialize (scales, value, options={})
-    @y  = value
-    @x  = scales
-  end
-  
-  def evaluate (x0)
-    @y.interpolate(@x, x0)
-  end
-
-  alias grid evaluate
-
-end
-

data/ext/mathfunc/extconf.rb

@@ -1,18 +0,0 @@
-require "mkmf"
-
-dir_config("carray", "../..", "../..")
-
-have_header("carray.h")
-
-if /cygwin|mingw/ =~ RUBY_PLATFORM
-  have_library("carray")
-end
-
-have_header("tgmath.h")
-
-have_func("atan2",  "math.h")
-have_func("hypot",  "math.h")
-have_func("lgamma", "math.h")
-have_func("expm1",  "math.h")
-
-create_makefile("carray/carray_mathfunc")

data/ext/mathfunc/test/test_hypot.rb

@@ -1,5 +0,0 @@
-require 'carray'
-
-p x = CArray.object(101).span(0..1)
-
-p CAMath.hypot(x[:*,nil], x[nil,:*])

data/ext/mathfunc/test/test_j0.rb

@@ -1,22 +0,0 @@
-require 'carray'
-
-x = CArray.float(101).span(0..10)
-
-CA.gnuplot { |g|
-  g.plot2d(
-           [x, x.erf],
-           [x, x.erfc]
-           )
-  g.plot2d(
-           [x, x.j0],
-           [x, x.j1],
-           [x, x.jn(2)],
-           [x, x.jn(3)],
-           [x, x.jn(4)],
-           [x, x.jn(5)]
-           )
-  g.plot2d(
-           [x, x.y0],
-           [x, x.y1]
-           )
-}

data/ext/mathfunc/test/test_jn.rb

@@ -1,8 +0,0 @@
-require "carray"
-
-p x = CAMath.yn(1, CA_FLOAT(0..100))
-
-CA.gnuplot { |g|
-  g.plot2d([x], :with=>"lines")
-}
-

data/ext/mathfunc/test/test_sph.rb

@@ -1,9 +0,0 @@
-require "carray"
-
-r = CA_DOUBLE(1)
-theta = CA_DOUBLE(30).rad
-phi = CA_DOUBLE(45).rad
-
-x,y,z = CAMath.spherical_to_xyz(r, theta, phi)
-p CAMath.xyz_to_spherical(x,y,z)
-p [r, theta, phi]

data/lib/carray/autoload/autoload_io_table.rb

@@ -1 +0,0 @@
-autoload_method "CArray.extend_as_table", "carray/io/table"

data/lib/carray/autoload/autoload_math_interp.rb

@@ -1,4 +0,0 @@
-module CA
-  autoload :Interp, "carray/math/interp"
-  autoload_method "self.interp", "carray/math/interp"
-end

data/lib/carray/base/basic.rb

@@ -1,1146 +0,0 @@
-# ----------------------------------------------------------------------------
-#
-#  carray/base/base.rb
-#
-#  This file is part of Ruby/CArray extension library.
-#  You can redistribute it and/or modify it under the terms of
-#  the Ruby Licence.
-#
-#  Copyright (C) 2005 Hiroki Motoyoshi
-#
-# ----------------------------------------------------------------------------
-
-module CAMath
-  include Math
-end
-
-def nan
-  0.0/0.0
-end
-
-class Array                          # :nodoc:
-  def +@
-    return CA_SIZE(self)
-  end
-  def to_ca
-    return CA_OBJECT(self)
-  end
-end
-
-class Range                          # :nodoc: 
-  def +@
-    return CA_SIZE(self)    
-  end
-  def to_ca
-    return CA_OBJECT(self)
-  end
-end
-
-# Create new CArray object from the return value of the block
-# with data type +type+. The dimensional size and the initialization value
-# are guessed from the return value of the block.
-# The block should return one of the following objects.
-#
-# * Numeric
-# * Array
-# * CArray
-# * an object that has either method +to_ca+ or +to_a+ or +map+
-#
-# When the return value of the block is a Numeric or CScalar object,
-# CScalar object is returned.
-#
-def CArray.__new__ (type, *args) # :nodoc:
-  case v = args.first
-  when CArray
-    return ( v.data_type == type ) ? v.to_ca : v.to_type(type)
-  when Array
-    return CArray.new(type, CArray.guess_array_shape(v)) { v }
-  when Range
-    return CArray.span(type, *args)
-  when String
-    if type == CA_OBJECT
-      return CScalar.new(CA_OBJECT) { v }
-    elsif type == CA_BOOLEAN
-        v = v.dup
-        v.tr!('^01',"1")
-        v.tr!('01',"\x0\x1")
-        return CArray.boolean(v.length).load_binary(v)
-    else
-      case v
-      when /;/
-        v = v.strip.split(/\s*;\s*/).
-                         map{|s| s.split(/\s+|\s*,\s*/).map{|x| x=='_' ? UNDEF : x} }
-      else
-        v = v.strip.split(/\s+|\s*,\s*/).map{|x| x=='_' ? UNDEF : x}
-      end
-      return CArray.new(type, CArray.guess_array_shape(v)) { v }
-    end
-  when NilClass
-    return CArray.new(type, [0])
-  else
-    if v.respond_to?(:to_ca)
-      ca = v.to_ca
-      return ( ca.data_type == type ) ? ca : ca.to_type(type)
-    else
-      return CScalar.new(type) { v }
-    end
-  end
-end
-
-def CArray.__new_fixlen__ (bytes, v) # :nodoc:
-  case v
-  when CArray
-    return ( v.data_type == :fixlen ) ? v.to_ca : v.to_type(:fixlen, :bytes=>bytes)
-  when Array
-    unless bytes
-      bytes = v.map{|s| s.length}.max
-    end
-    return CArray.new(:fixlen, CArray.guess_array_shape(v), :bytes=>bytes) { v }
-  when NilClass
-    return CArray.new(type, [0])
-  else
-    if v.respond_to?(:to_ca)
-      ca = v.to_ca
-      return ( ca.data_type == :fixlen ) ? ca : ca.to_type(:fixlen, :bytes=>bytes)
-    else
-      return CScalar.new(:fixlen, :bytes=>bytes) { v }
-    end
-  end
-end
-
-#
-# CA_INT8(data)
-#   :
-# CA_CMPLX256(data)
-#
-# Create new CArray object from +data+ with data type
-# which is guessed from the method name. +data+ should be one of the following
-# objects.
-#
-# * Numeric
-# * Array
-# * CArray
-# * an object that has either method +to_ca+ or +to_a+ or +map+
-#
-# When the block returns a Numeric or CScalar object,
-# the resulted array is a CScalar object.
-
-[
- "CA_BOOLEAN",
- "CA_INT8",
- "CA_UINT8",
- "CA_INT16",
- "CA_UINT16",
- "CA_INT32",
- "CA_UINT32",
- "CA_INT64",
- "CA_UINT64",
- "CA_FLOAT32",
- "CA_FLOAT64",
- "CA_FLOAT128",
- "CA_CMPLX64",
- "CA_CMPLX128",
- "CA_CMPLX256",
- "CA_OBJECT",
- "CA_BYTE",
- "CA_SHORT",
- "CA_INT",
- "CA_FLOAT",
- "CA_DOUBLE",
- "CA_COMPLEX",
- "CA_DCOMPLEX",
- "CA_SIZE",
-].each do |name|
-  eval %{
-    def #{name} (*val)
-      CArray.__new__(#{name}, *val)
-    end
-  }
-end
-
-def CA_FIXLEN (val, options = {})
-  CArray.__new_fixlen__(options[:bytes], val)
-end
-
-class CArray
-
-  def has_attribute?
-    if ( not @attribute ) or @attribute.empty?
-      return false
-    else
-      return true
-    end
-  end
-
-  def attribute= (obj)
-    unless obj.is_a?(Hash)
-      raise "attribute should be a hash object"
-    end
-    @attribute = obj
-  end
-
-  def attribute
-    @attribute ||= {}
-    return @attribute
-  end
-  
-  def first
-    self[0]
-  end
-  
-  def last
-    self[-1]
-  end
-
-  def range
-    return (self.min)..(self.max)
-  end
-
-  def asign (*idx)
-    self[*idx] = yield
-    return self
-  end
-
-  # mask
-  
-  #
-  # Returns the number of masked elements.
-  #
-
-  def count_masked (*argv)
-    if has_mask?
-      return mask.int32.accumulate(*argv)
-    else
-      return 0
-    end
-  end
-
-  #
-  # Returns the number of not-masked elements.
-  #
-  def count_not_masked (*argv)
-    if has_mask?
-      return mask.not.int32.accumulate(*argv)
-    else
-      return elements
-    end
-  end
-
-  def maskout! (*argv)
-    if argv.size == 1
-      val = argv.first
-      case val
-      when CArray, Symbol
-        self[val] = UNDEF
-      else
-        self[:eq, val] = UNDEF
-      end
-    else
-      self[*argv] = UNDEF          
-    end
-    return self
-  end
-
-  def maskout (*argv)
-    obj = self.to_ca
-    if argv.size == 1
-      val = argv.first
-      case val
-      when CArray, Symbol
-        obj[val] = UNDEF
-      else
-        obj[:eq, val] = UNDEF
-      end
-    else
-      obj[*argv] = UNDEF      
-    end
-    return obj
-  end
-
-  # matchup
-
-  def matchup (ref)
-    ri = ref.sort_addr
-    rs = ref[ri].to_ca
-    si = rs.bsearch(self)
-    return ri.project(si)
-  end
-
-  def matchup_nearest (ref)
-    ri = ref.sort_addr
-    rs = ref[ri].to_ca
-    si = rs.section(self).round.int32
-    si.trim!(0,si.size)
-    return ri[si].to_ca
-  end
-
-  # replace
-  #
-  #
-  def replace (from, to)
-    self[:eq, from] = to
-    return self
-  end
-
-  # reshape
-
-  def reshape (*newdim)
-    ifalse = nil
-    i = 0
-    0.upto(newdim.size-1) do |i|
-      if newdim[i].nil?
-        newdim[i] = dim[i]
-      elsif newdim[i] == false 
-        ifalse = i
-        break
-      end
-    end
-    k = 0
-    (newdim.size-1).downto(i+1) do |j|
-      if newdim[j].nil?
-        newdim[j] = dim[rank-1-k]
-      end
-      k += 1
-    end
-    if ifalse
-      newdim[ifalse] = 
-          elements/newdim.select{|x| x!=false}.inject(1){|s,x| s*x}
-    end
-    return refer(data_type, newdim, :bytes=>bytes)
-  end
-
-  # flatten
-
-  def flattened
-    return reshape(elements)
-  end
-
-  def flatten
-    return reshape(elements).to_ca
-  end
-
-  # pulled
-
-  def pulled (*args)
-    idx = args.map{|s| s.nil? ? :% : s}
-    return self[*idx]
-  end
-  
-  def pull (*args)
-    idx = args.map{|s| s.nil? ? :% : s}
-    return self[*idx].to_ca
-  end
-
-  # reversed
-
-  def reversed
-    return self[*([-1..0]*rank)]
-  end
-
-  # roll / shift
-
-  def shift! (*argv, &block)
-    self[] = self.shifted(*argv, &block)
-    return self
-  end
-
-  def shift (*argv, &block)
-    return self.shifted(*argv, &block).to_ca
-  end
-
-  def rolled (*argv)
-    argv.push({:roll => Array.new(rank){1} })
-    return shifted(*argv)
-  end
-
-  def roll! (*argv)
-    self[] = self.rolled(*argv)
-    return self
-  end
-
-  def roll (*argv)
-    return self.rolled(*argv).to_ca
-  end
-
-  def transpose! (*argv)
-    self[] = self.transposed(*argv)
-    return self
-  end
-
-  def transpose (*argv)
-    return self.transposed(*argv).to_ca
-  end
-
-  # Reutrns the reference which rank is reduced 
-  # by eliminating the dimensions which size == 1 
-  def compacted
-    if rank == 1
-      return self[]
-    else
-      newdim = dim.reject{|x| x == 1 }
-      return ( rank != newdim.size ) ? reshape(*newdim) : self[]
-    end
-  end
-
-  # Returns the array which rank is reduced 
-  # by eliminating the dimensions which size == 1 
-  def compact
-    if rank == 1
-      return self.to_ca
-    else
-      newdim = dim.reject{|x| x == 1 }
-      return ( rank != newdim.size ) ? reshape(*newdim).to_ca : self.to_ca
-    end
-  end
-
-  # Returns the reference which elements are sorted by the comparison method
-  # given as block
-  def sorted_by (type=nil, opt={}, &block)
-    type, bytes =
-      CArray.guess_type_and_bytes(type||data_type, opt[:bytes]||bytes)
-    cmpary = convert(type, :bytes=>bytes, &block)
-    return self[cmpary.sort_addr]
-  end
-
-  # Returns the array which elements are sorted by the comparison method
-  # given as block
-  def sort_by (type=nil, opt={}, &block)
-    type, bytes =
-      CArray.guess_type_and_bytes(type||data_type, opt[:bytes]||bytes)
-    cmpary = convert(type, :bytes=>bytes, &block)
-    return self[cmpary.sort_addr].to_ca
-  end
-
-  def max_by (&block)
-    if empty?
-      return UNDEF
-    else
-      addr = convert(:object, &block).max_addr
-      return self[addr]
-    end
-  end
-
-  def min_by (&block)
-    if empty?
-      return UNDEF
-    else
-      addr = convert(:object, &block).min_addr
-      return self[addr]
-    end
-  end
-
-  def nlargest (n)
-    obj = self.to_ca
-    list = []
-    n.times do |i|
-      k = obj.max_addr
-      list << obj[k]
-      obj[k] = UNDEF
-    end
-    list.to_ca.to_type(data_type)
-  end
-
-  def nlargest_addr (n)
-    obj = self.to_ca
-    list = []
-    n.times do |i|
-      k = obj.max_addr
-      list << k
-      obj[k] = UNDEF
-    end
-    CA_INT64(list)
-  end
-
-  def nsmallest (n)
-    obj = self.to_ca
-    list = []
-    n.times do |i|
-      k = obj.min_addr
-      list << obj[k]
-      obj[k] = UNDEF
-    end
-    list.to_ca.to_type(data_type)
-  end
-
-  def nsmallest_addr (n)
-    obj = self.to_ca
-    list = []
-    n.times do |i|
-      k = obj.min_addr
-      list << k
-      obj[k] = UNDEF
-    end
-    CA_INT64(list)
-  end
-
-  # Returns (1,n) array from 1-dimensional array 
-  def to_row 
-    if rank != 1
-      raise "rank should be 1"
-    end
-    return self[1,:%]
-  end
-  
-  # Returns (n,1) array from 1-dimensional array 
-  def to_column
-    if rank != 1
-      raise "rank should be 1"
-    end
-    return self[:%,1]
-  end
-
-  # Returns the array resized to the dimension given as `newdim`.
-  # The new area is filled by the value returned by the block.
-  def resize (*newdim, &block)
-    if newdim.size != rank
-      raise "rank mismatch"
-    end
-    offset = Array.new(rank){0}
-    rank.times do |i|
-      d = newdim[i]
-      case d
-      when nil
-        newdim[i] = dim[i]
-      when Integer
-        if d < 0
-          newdim[i] *= -1
-          offset[i] = newdim[i] - dim[i]
-        end
-      else
-        raise "invalid dimension size"
-      end
-    end
-    out = CArray.new(data_type, newdim, &block)
-    if out.has_mask?
-      out.mask.paste(offset, self.false)
-    end
-    out.paste(offset, self)
-    return out
-  end
-
-  # insert
-  def insert_block (offset, bsize, &block)
-    if offset.size != rank or
-        bsize.size != rank
-      raise "rank mismatch"
-    end
-    newdim = dim
-    grids = dim.map{|d| CArray.int32(d) }
-    rank.times do |i|
-      if offset[i] < 0
-        offset[i] += dim[i]
-      end
-      if offset[i] < 0 or offset[i] >= dim[i] or bsize[i] < 0
-        raise "invalid offset or size"
-      end
-      if bsize[i] > 0
-        newdim[i] += bsize[i]
-      end
-      grids[i][0...offset[i]].seq!
-      grids[i][offset[i]..-1].seq!(offset[i]+bsize[i])
-    end
-    out = CArray.new(data_type, newdim)
-    if block_given?
-      sel = out.true
-      sel[*grids] = 0
-      out[sel] = block.call
-    end
-    out[*grids] = self
-    return out
-  end
-
-  def delete_block (offset, bsize)
-    if offset.size != rank or
-      bsize.size != rank
-      raise "rank mismatch"
-    end
-    newdim = dim
-    grids  = []
-    rank.times do |i|
-      if offset[i] < 0
-        offset[i] += dim[i]
-      end
-      if bsize[i] >= 0
-        if offset[i] < 0 or offset[i] >= dim[i]
-          raise "invalid offset or size"
-        end
-        newdim[i] -= bsize[i]
-      else
-        if offset[i] + bsize[i] + 1 < 0 or offset[i] + bsize[i] > dim[i]
-          raise "invalid offset or size"
-        end
-        newdim[i] += bsize[i]
-      end
-      grids[i] = CArray.int32(newdim[i])
-      if bsize[i] >= 0
-        if offset[i] > 0
-          grids[i][0...offset[i]].seq!
-        end
-        if offset[i] + bsize[i] < dim[i]
-          grids[i][offset[i]..-1].seq!(offset[i]+bsize[i])
-        end
-      else
-        if offset[i]+bsize[i] > 0
-          grids[i][0..offset[i]+bsize[i]].seq!
-        end
-        if offset[i]+bsize[i]+1 < dim[i]-1
-          grids[i][offset[i]+bsize[i]+1..-1].seq!(offset[i]+1)
-        end
-      end
-    end
-    return self[*grids].to_ca
-  end
-
-  def where_range
-    w = where
-    x = (w - w.shifted(1){-2}).sub!(1).where
-    y = (w - w.shifted(-1){-2}).add!(1).where
-    list = []
-    x.each_addr do |i|
-      list.push(w[x[i]]..w[y[i]])
-    end
-    return list
-  end
-  
-  def order (dir = 1)
-    if dir >= 0   ### ascending order
-      if has_mask?
-        obj = template(:int32) { UNDEF }
-        sel = is_not_masked
-        obj[sel][self[sel].sort_addr].seq!
-        return obj
-      else
-        obj = template(:int32)
-        obj[sort_addr].seq!
-        return obj
-      end
-    else           ### descending order
-      if has_mask?
-        obj = template(:int32) { UNDEF}
-        sel = is_not_masked
-        obj[sel][self[sel].sort_addr.reversed].seq!
-        return obj
-      else  
-        obj = template(:int32)
-        obj[sort_addr.reversed].seq!
-        return obj
-      end
-    end
-  end
-
-  # Returns the array eliminated all the duplicated elements.
-  def uniq
-    ary = to_a.uniq
-    if has_mask?
-      ary.delete(UNDEF)
-    end
-    if has_data_class?
-      return CArray.new(data_class, [ary.length]) { ary }
-    else
-      return CArray.new(data_type, [ary.length], :bytes=>bytes) { ary }
-    end
-  end
-
-  # Returns the array eliminated all the duplicated elements.
-  def duplicated_values
-    if uniq.size == size
-      return []
-    else
-      hash = {}
-      list = []
-      each_with_addr do |v, addr|
-        if v == UNDEF
-          next
-        elsif hash[v]
-          list << [v, addr, hash[v]]
-          hash[v] += 1
-        else
-          hash[v] = 0
-        end
-      end
-      return list
-    end
-  end
-
-  #
-  # CArray.span(data_type, range[, step])
-  # CArray.span(range[, step]) -> data_type guessed by range.first type
-  #
-
-  def self.span (*argv)
-    if argv.first.is_a?(Range)
-      type = nil
-    else
-      type, = *CArray.guess_type_and_bytes(argv.shift, nil)
-    end
-    range, step = argv[0], argv[1]
-    start, stop = range.begin, range.end
-    if step == 0
-      raise "step should not be 0"
-    end
-    if not type
-      case start
-      when Integer
-        type = CA_INT32
-      when Float
-        type = CA_FLOAT64
-      else
-        type = CA_OBJECT
-      end
-    end
-    if type == CA_OBJECT and not step
-      return CA_OBJECT(range.to_a)
-    else
-      step ||= 1
-      if range.exclude_end?
-        n = ((stop - start).abs/step).floor
-      else
-        n = ((stop - start).abs/step).floor + 1
-      end
-      if start <= stop
-        return CArray.new(type, [n]).seq(start, step)
-      else
-        return CArray.new(type, [n]).seq(start, -step.abs)
-      end
-    end
-  end
-
-  def span! (range)
-    first = range.begin.to_r
-    last  = range.end.to_r
-    if range.exclude_end?
-      seq!(first, (last-first)/elements)
-    else
-      seq!(first, (last-first)/(elements-1))
-    end
-    return self
-  end
-
-  def span (range)
-    return template.span!(range)
-  end
-
-  #
-  #
-  #
-
-  def scale! (xa, xb)
-    xa = xa.to_f
-    xb = xb.to_f
-    seq!(xa, (xb-xa)/(elements-1))
-  end
-
-  def scale (xa, xb)
-    template.scale!(xa, xb)
-  end
-
-  # index / indices / axes
-
-  def address ()
-    return CArray.int32(*dim).seq!
-  end
-
-  def index (n = 0)
-    unless n.is_a?(Integer)
-      raise ArgumentError, "argument should be an integer"
-    end
-    if n.between?(0, rank-1)
-      return CArray.int32(dim[n]).seq!
-    else
-      raise ArgumentError,
-            "invalid dimension specifier #{n} (0..#{self.rank-1})"
-    end
-  end
-
-  #
-  #
-  #
-
-  def indices
-    list = Array.new(rank) {|i|
-      rpt = self.dim
-      rpt[i] = :%
-      index(i)[*rpt]
-    }
-    if block_given?
-      return yield(*list)
-    else
-      return list
-    end
-  end
-
-  #
-  # Returns the 8-bit integer CArray object filled with 0 which
-  # dimension size is same as +self+. The resulted array represents
-  # the logical array which has +false+ for its all elements.
-  #
-  def false ()
-    return template(:boolean)
-  end
-
-  #
-  # Returns the 8-bit integer CArray object filled with 1 which
-  # dimension size is same as +self+. The resulted array represents
-  # the logical array which has +true+ for its all elements.
-  #
-  def true ()
-    return template(:boolean) { 1 }
-  end
-
-  def contains (*list)
-    result = self.false()
-    list.each do |item|
-      result = result | self.eq(item)
-    end
-    return result 
-  end
-
-  def between (a, b)
-    return (self >= a) & (self <= b)
-  end
-
-  # Returns map
-  def map (&block)
-    return self.convert(CA_OBJECT, &block).to_a
-  end
-
-  # Array#join like method
-  #
-  # > a = CArray.object(3,3).seq("a",:succ)
-  # => <CArray.object(3,3): elem=9 mem=72b
-  # [ [ "a", "b", "c" ],
-  #   [ "d", "e", "f" ],
-  #   [ "g", "h", "i" ] ]>
-  # 
-  # > a.join("\n",",")
-  # => "a,b,c\nd,e,f\ng,h,i"
-  #
-
-  def join (*argv)
-    case argv.size
-    when 0
-      return to_a.join()
-    when 1
-      sep = argv.shift
-      return to_a.join(sep)
-    else
-      sep = argv.shift
-      return self[:i, false].map { |s|
-          s[0, false].join(*argv)
-      }.join(sep)
-    end
-  end
-    
-  # 
-  def to_bit_string (nb)
-    hex = CArray.uint8(((nb*elements)/8.0).ceil)
-    hex.bits[nil].paste([0], self.bits[false,[(nb-1)..0]].flatten)
-    hex.bits[] = hex.bits[nil,[-1..0]]
-    return hex.to_s
-  end
-
-  def from_bit_string (bstr, nb)
-    hex = CArray.uint8(bstr.length).load_binary(bstr)
-    hex.bits[] = hex.bits[nil,[-1..0]]
-    bits = hex.bits.flatten
-    self.bits[false,[(nb-1)..0]][nil].paste([0], bits)
-    return self
-  end
-
-  def self.from_bit_string (bstr, nb, data_type=CA_INT32, dim=nil)
-    if dim
-      obj = CArray.new(data_type, dim)
-    else
-      dim0 = ((bstr.length*8)/nb.to_f).floor
-      obj = CArray.new(data_type, [dim0])
-    end
-    obj.from_bit_string(bstr, nb)
-    return obj
-  end
-
-
-end
-
-class CArray
-
-  #
-  #  ref = CA_INT([[0,1,2],[1,2,0],[2,0,1]])
-  #  a = CArray.int(3,3).seq(1)
-  #  b = CArray.int(3,3).seq(11)
-  #  c = CArray.int(3,3).seq(21)
-  #
-  #  CArray.pickup(CA_OBJECT, ref, [a,b,c])
-  #  => <CArray.object(3,3): elem=9 mem=72b
-  #  [ [ 1, 12, 23 ],
-  #    [ 14, 25, 6 ],
-  #    [ 27, 8, 19 ] ]>
-  #
-  #  CArray.pickup(CA_OBJECT, ref, ["a","b","c"])
-  #  => <CArray.object(3,3): elem=9 mem=36b
-  #  [ [ "a", "b", "c" ],
-  #    [ "b", "c", "a" ],
-  #    [ "c", "a", "b" ] ]>
-  #
-  def self.pickup (data_type, ref, args)
-    out = ref.template(data_type)
-    args.each_with_index do |v, i|
-      s = ref.eq(i)
-      case v
-      when CArray
-        out[s] = v[s]
-      else
-        out[s] = v
-      end
-    end
-    return out
-  end
-
-  def self.combine (data_type, tdim, list, at = 0)
-    has_fill_value = false
-    if block_given?
-      fill_value = yield
-      has_fill_value = true
-    end
-    if not tdim.is_a?(Array) or tdim.size == 0
-      raise "invalid binding dimension"
-    end
-    if not list.is_a?(Array) or list.size == 0
-      raise "invalid list"
-    end
-    list = list.map{|x| CArray.wrap_readonly(x, data_type) }
-    ref  = list.detect{|x| x.is_a?(CArray) or not x.scalar? }
-    unless ref
-      raise "at least one element in list should be a carray"
-    end
-    dim   = ref.dim
-    rank  = ref.rank
-    trank = tdim.size
-    if at < 0
-      at += rank - trank + 1
-    end
-    unless at.between?(0, rank - trank)
-      raise "concatnating position out of range"
-    end
-    list.map! do |x|
-      if x.scalar?
-        rdim = dim.clone
-        rdim[at] = :%
-        x = x[*rdim]        # convert CScalar to CARepeat
-      end
-      x
-    end
-    block = CArray.object(*tdim){ list }
-    edim = tdim.clone
-    idx = Array.new(tdim)
-    offset = Array.new(tdim.size) { [] }
-    tdim.each_with_index do |td, i|
-      edim[i] = 0
-      idx.map!{0}
-      idx[i] = nil
-      block[*idx].each do |e|
-        offset[i] << edim[i]
-        edim[i] += e.dim[at+i]  # extended dimension size
-      end
-    end
-    newdim = dim.clone
-    newdim[at,trank] = edim     # extended dimension size
-    if has_fill_value
-      obj = CArray.new(data_type, newdim) { fill_value }
-    else      
-      obj = CArray.new(data_type, newdim)
-    end
-    idx = newdim.map{0}
-    block.each_with_index do |item, tidx|
-      (at...at+trank).each_with_index do |d,i|
-        idx[d] = offset[i][tidx[i]]
-      end
-      obj.paste(idx, item)
-    end
-    obj
-  end
-
-  def self.bind (data_type, list, at = 0)
-    return CArray.combine(data_type, [list.size], list, at)
-  end
-
-  def self.composite (data_type, tdim, list, at = 0)
-    if not tdim.is_a?(Array) or tdim.size == 0
-      raise "invalid tiling dimension"
-    end
-    if not list.is_a?(Array) or list.size == 0
-      raise "invalid carray list"
-    end
-    list = list.map{|x| CArray.wrap_readonly(x, data_type) }
-    ref  = list.detect{|x| x.is_a?(CArray) or not x.scalar? }
-    unless ref
-      raise "at least one element in list should be a carray"
-    end
-    dim   = ref.dim
-    rank  = ref.rank
-    if at < 0
-      at += rank + 1 #  "+ 1" is needed here
-    end
-    unless at.between?(0,rank)
-      raise "tiling position is out of range"
-    end
-    trank = tdim.size
-    list.map! do |x|
-      if x.scalar?
-        rdim = dim.clone
-        rdim[at] = :%
-        x = x[*rdim]     # convert CScalar to CARepeat
-      end
-      x
-    end
-    newdim = dim.clone
-    newdim[at,0] = tdim
-    obj = CArray.new(data_type, newdim)
-    idx = Array.new(rank+trank) { nil }
-    CArray.each_index(*tdim) do |*tidx|
-      idx[at,trank] = tidx
-      obj[*idx] = list.shift
-    end
-    obj
-  end
-
-  def self.merge (data_type, list, at = -1)
-    return CArray.composite(data_type, [list.size], list, at)
-  end
-  
-  def self.join (*argv)
-    # get options
-    case argv.first
-    when Integer, Symbol, String
-      type, = *CArray.guess_type_and_bytes(argv.shift, 0)
-    else
-      type = argv.flatten.first.data_type
-    end
-    # process
-    conc = argv.map do |list|
-      case list
-      when CArray
-        if list.rank == 1
-          list[:%,1]
-        else
-          list
-        end
-      when Array
-        x0 = list.first
-        if list.size == 1 and
-            x0.is_a?(CArray) and
-            x0.rank == 1
-          list = [x0[:%,1]]
-        else
-        list = list.map { |x|
-          case x
-          when CArray
-            if x.rank == 1
-              x[:%,1]
-            else
-              x
-            end
-          when Array
-            y = x.first
-            if x.size == 1 and
-                y.is_a?(CArray) and
-                y.rank == 1
-              y[1,:%]
-            else
-              CArray.join(*x)
-            end
-          else
-            x
-          end
-        }
-        end
-        if block_given?
-          CArray.bind(type, list, 1, &block)
-        else
-          CArray.bind(type, list, 1)
-        end 
-      else
-        list
-      end
-    end
-    if conc.size > 1
-      return CArray.bind(type, conc)
-    else
-      return conc.first
-    end
-  end
-  
-  # Returns object carray has elements of splitted carray at dimensions 
-  #      which is given by arguments
-  #
-  #    a = CA_INT([[1,2,3], [4,5,6], [7,8,9]])
-  #
-  #    a.split(0) 
-  #      [1,2,3], [4,5,6], [7,8,9]
-  #
-  #    a.split(1)
-  #      [1,4,7], [2,5,8], [3,6,9]
-  #
-
-  def split (*argv)
-    odim = dim.values_at(*argv)
-    out  = CArray.object(*odim)
-    idx  = [nil] * rank
-    attach {
-      out.map_with_index! do |o, v|
-        argv.each_with_index do |r, i|
-          idx[r] = v[i]
-        end
-        self[*idx].to_ca
-      end
-    }
-    return out
-  end
-
-end
-
-class CAUnboundRepeat
-
-  def template (*argv, &block)
-    return parent.template(*argv,&block)[*spec.map{|x| x != :* ? nil : x}]
-  end
-
-end
-
-class Numeric
-  
-  def eq (other)
-    case other
-    when CArray
-      return other.eq(self)
-    else
-      return send(:eq, *other.coerce(self))
-    end
-  end
-
-  def ne (other)
-    case other
-    when CArray
-      return other.ne(self)
-    else
-      return send(:ne, *other.coerce(self))
-    end
-  end
-  
-end
-
-module CA
-  
-  def self.meshgrid (*args)
-    dim = args.map(&:size)
-    out = []
-    args.each_with_index do |arg, i|
-      newdim = dim.dup
-      newdim[i] = :%
-      out[i] = arg[*newdim].to_ca
-    end
-    return *out
-  end
-  
-end
-
-