
The Python Imaging Library
$Id: CHANGES$

*** Changes from release 0.2b4 to 0.3b2 ***

ACKNOWLEDGEMENTS: David Ascher, Phil Austin, Anthony Baxter, Jan Blom,
Greg Coats, Fred Drake, Tim Docker, Robin Friedrich, Jack Jansen,
Richard Jones, Skip Montanaro, Fredrik Nehr, Mark Hammond, Hkan
Karlsson, Bertil Reinhammar, Eric S. Raymond, Joel Shprentz, Perry
Stoll, Paul Svensson, Ivan Tkatchev, and many others have contributed
bug reports, code snippets, ideas, and inspiration for new features
(many of which are yet to come!).

Here's what's new this far in PIL 0.3:

(0.3b2 released)

The test suite includes 825 individual tests.

+ An Image "getbands" method has been added.  It returns a tuple
  containing the individual band names for this image.  To figure
  out how many bands an image has, use "len(im.getbands())".

+ An Image "putpixel" method has been added.

+ The Image "point" method can now be used to convert "L" images
  to any other format, via a lookup table.  That table should
  contain 256 values for each band in the output image.

+ Some file drivers (including FLI/FLC, GIF, and IM) accidently
  overwrote the offset method with an internal attribute.  All
  drivers have been updated to use private attributes where
  possible.

+ The Image "histogram" method now works for "I" and "F" images.
  For these modes, PIL divides the range between the min and
  max values used in the image into 256 bins.  You can also
  pass in your own min and max values via the "extrema" option:

	h = im.histogram(extrema=(0, 255))

+ An Image "getextrema" method has been added.  It returns the
  min and max values used in the image. In this release, this
  works for single band images only.

+ Changed the PNG driver to load and save mode "I" images as
  16-bit images.  When saving, values outside the range 0..65535
  are clipped.

+ Fixed ImageFont.py to work with the new "pilfont" compiler.

+ Added JPEG "save" and "draft" support for mode "YCbCr" images.
  Note that if you save an "YCbCr" image as a JPEG file and read
  it back, it is read as an RGB file.  To get around this, you
  can use the "draft" method:

	im = Image.open("color.jpg")
	im.draft("YCbCr", im.size)

+ Read "RGBA" TGA images.  Also fixed the orientation bug; all
  images should now come out the right way.

+ Changed mode name (and internal representation) from "YCrCb"
  to "YCbCr" (!)
  *** WARNING: MAY BREAK EXISTING CODE ***

(0.3b1 released)

The test suite includes 750 individual tests.

+ The "pilfont" package is now included in the standard PIL
  distribution.  The pilfont utility can be used to convert
  X BDF and PCF raster font files to a format understood by
  the ImageFont module.

+ GIF files are now interlaced by default.  To write a
  non-interlaced file, pass interlace=0 to the "save"
  method.

+ The default string format has changed for the "fromstring"
  and "tostring" methods.
  *** WARNING: MAY BREAK EXISTING CODE ***

  Notes: If no extra arguments are given, the first line in
  the string buffer is the top line of the image, instead of
  the bottom line.  For RGB images, the string now contains
  3 bytes per pixel instead of 4.  These changes were made
  to make the methods compatible with the "fromstring"
  factory function.

  To get the old behaviour, use the following syntax:

	data = im.tostring("raw", "RGBX", 0, -1)
	im.fromstring(data, "raw", "RGBX", 0, -1)

+ "new" no longer gives a MemoryError if the width or height
  is zero (this only happened on platforms where malloc(0)
  returns NULL).

+ "new" now adds a default palette object to "P" images.

+ You can now convert directly between all modes supported by
  PIL.  When converting colour images to "P", PIL defaults to
  a "web" palette and dithering.  When  converting greyscale
  images to "1", PIL uses a thresholding and dithering.

+ Added a "dither" option to "convert".  By default, "convert"
  uses floyd-steinberg error diffusion for "P" and "1" targets,
  so this option is only used to *disable* dithering. Allowed
  values are NONE (no dithering) or FLOYDSTEINBERG (default).

	imOut = im.convert("P", dither=Image.NONE)

+ Added a full set of "I" decoders.  You can use "fromstring"
  (and file decoders) to read any standard integer type as an
  "I" image.

+ Added some support for "YCbCr" images (creation, conversion
  from/to "L" and "RGB", IM YCC load/save)

+ "getpixel" now works properly with fractional coordinates.

+ ImageDraw "setink" now works with "I", "F", "RGB", "RGBA",
  "RGBX", "CMYK", and "YCbCr" images.

+ ImImagePlugin no longer attaches palettes to "RGB" images.

+ Various minor fixes.

(0.3a4 released)

+ Added IPTC/NAA support (experimental).

+ Eliminated AttributeError exceptions after "crop" (thanks
  to Skip Montanaro)

+ Reads some uncompressed formats via memory mapping (this
  is currently supported on Win32 only)

+ Fixed some last minute glitches in the last alpha release
  (Types instead of types in Image.py, version numbers, etc.)

+ Eliminated some more bogus compiler warnings.

+ Various fixes to make PIL compile and run smoother on Macs
  (from Jack Jansen).

+ Fixed "fromstring" and "tostring" for mode "I" images.

(0.3a3 released)

The test suite includes 530 individual tests.

+ Eliminated unexpected side-effect in "paste" with matte.  "paste"
  now works properly also if compiled with "gcc".

+ Adapted to Python 1.5 (build issues only)

+ Fixed the ImageDraw "point" method to draw also the last
  point (!).

+ Added "I" and "RGBX" support to Image.new.

+ The plugin path is now properly prepended to the module search
  path when a plugin module is imported.

+ Added "draw" method to the ImageWin.Dib class.  This is used by
  Topaz to print images on Windows printers.

+ "convert" now supports conversions from "P" to "1" and "F".

+ "paste" can now take a colour instead of an image as the first argument.
  The colour must match the colour argument given to the new function, and
  match the mode of the target image.

+ Fixed "paste" to allow a mask also for mode "F" images.

+ The BMP driver now saves mode "1" images.  When loading images, the mode
  is set to "L" for 8-bit files with greyscale palettes, and to "P" for
  other 8-bit files.

+ The IM driver now reads and saves "1" images (file modes "0 1" or "L 1").

+ The JPEG and GIF drivers now saves "1" images.  For JPEG, the image
  is saved as 8-bit greyscale (it will load as mode "L").  For GIF, the
  image will be loaded as a "P" image.

+ Fixed a potential buffer overrun in the GIF encoder.

(0.3a2 released)

The test suite includes 400 individual tests.

+ Improvements to the test suite revealed a number of minor bugs, which
  are all fixed.  Note that crop/paste, 32-bit ImageDraw, and ImageFont
  are still weak spots in this release.

+ Added "putpalette" method to the Image class.  You can use this
  to add or modify the palette for "P" and "L" images.  If a palette
  is added to an "L" image, it is automatically converted to a "P"
  image.

+ Fixed ImageDraw to properly handle 32-bit image memories
  ("RGB", "RGBA", "CMYK", "F")

+ Fixed "fromstring" and "tostring" not to mess up the mode attribute
  in default mode.

+ Changed ImPlatform.h to work on CRAY's (don't have one at home, so I
  haven't tried it).  The previous version assumed that either "short"
  or "int" were 16-bit wide. PIL still won't compile on platforms where
  neither "short", "int" nor "long" are 32-bit wide.

+ Added file= and data= keyword arguments to PhotoImage and BitmapImage.
  This allows you to use them as drop-in replacements for the corre-
  sponding Tkinter classes.

+ Removed bogus references to the crack coder (ImagingCrack).

(0.3a1 released)

+ Make sure image is loaded in "tostring".

+ Added floating point packer (native 32-bit floats only).

*** Changes from release 0.1b1 to 0.2 (b5) ***

+ Modified "fromstring" and "tostring" methods to use file codecs.
  Also added "fromstring" factory method to create an image directly
  from data in a string.

+ Added support for 32-bit floating point images (mode "F").  You
  can convert between "L" and "F" images, and apply a subset of the
  available image processing methods on the "F" image.  You can also
  read virtually any data format into a floating point image memory;
  see the section on "Decoding Floating Point Data" in the handbook
  for more information.

(0.2b5 released; on windows only)

+ Fixed the tobitmap() method to work properly for small bitmaps.

+ Added RMS and standard deviation to the ImageStat.Stat class.  Also
  modified the constructor to take an optional feature mask, and also
  to accept either an image or a list containing the histogram data.

+ The BitmapImage code in ImageTk can now use a special bitmap
  decoder, which has to be patched into Tk.  See the "Tk/pilbitmap.txt"
  file for details.  If not installed, bitmaps are transferred to Tk as
  XBM strings.

+ The PhotoImage code in ImageTk now uses a Tcl command ("PyImagingPaste")
  instead of a special image type.  This gives somewhat better performance,
  and also allows PIL to support transparency.
  *** WARNING: TKAPPINIT MUST BE MODIFIED ***

+ ImageTk now honours the alpha layer in RGBA images.  Only fully
  transparent pixels are made transparent (that is, the alpha layer
  is treated as a mask).  To treat the alpha laters as a matte, you
  must paste the image on the background before handing it over to
  ImageTk.

+ Added McIdas reader (supports 8-bit images only).

+ PIL now preloads drivers for BMP, GIF, JPEG, PPM, and TIFF.  As
  long as you only load and save these formats, you don't have to
  wait for a full scan for drivers.  To force scanning, call the
  Image.init() function.

+ The "seek" and "tell" methods are now always available, also for
  single-frame images.

+ Added optional mask argument to histogram method.  The mask may
  be an "1" or "L" image with the same size as the original image.
  Only pixels where the mask is non-zero are included in the
  histogram.

+ The paste method now allows you to specify only the lower left
  corner (a 2-tuple), instead of the full region (a 4-tuple).

+ Reverted to old plugin scanning model; now scans all directory
  names in the path when looking for plugins.

+ Added PIXAR raster support.  Only uncompressed ("dumped") RGB
  images can currently be read.

+ Added FlashPix (FPX) read support.  Reads all pixel formats, but
  only the highest resolution is read, and the viewing transform is
  currently ignored.

+ Made PNG encoding somewhat more efficient in "optimize" mode; a
  bug in 0.2b4 didn't enable all predictor filters when optimized
  storage were requested.

+ Added Microsoft Image Composer (MIC) read support.  When opened,
  the first sprite in the file is loaded.  You can use the seek method
  to load additional sprites from the file.

+ Properly reads "P" and "CMYK" PSD images.

+ "pilconvert" no longer optimizes by default; use the -o option to
  make the file as small as possible (at the expense of speed); use
  the -q option to set the quality when compressing to JPEG.

+ Fixed "crop" not to drop the palette for "P" images.

+ Added and verified FLC support.

+ Paste with "L" or "RGBA" alpha is now several times faster on most
  platforms.

+ Changed Image.new() to initialize the image to black, as described
  in the handbook.  To get an uninitialized image, use None as the
  colour.

+ Fixed the PDF encoder to produce a valid header; Acrobat no longer
  complains when you load PDF images created by PIL.

+ PIL only scans fully-qualified directory names in the path when
  looking for plugins.
  *** WARNING: MAY BREAK EXISTING CODE ***

+ Faster implementation of "save" used when filename is given,
  or when file object has "fileno" and "flush" methods.

+ Fixed the "crop" method not to crash if the region extends
  outside the source image.

+ Eliminated a massive memory leak in the "save" function.

+ The GIF decoder doesn't crash if the code size is set to an illegal
  value.  This could happen since another bug didn't handle local
  palettes properly if they didn't have the same size as the
  global palette (not very common).

+ Added predictor support (TIFF 6.0 section 14) to the TIFF decoder.

+ Fixed palette and padding problems in BMP driver.  Now properly
  writes "1", "L", "P" and "RGB" images.

+ Fixed getpixel()/getdata() to return correct pixel values.

+ Added PSD (PhotoShop) read support.  Reads both uncompressed
  and compressed images of most types.

+ Added GIF write support (writes "uncompressed" GIF files only,
  due to unresolvable licensing issues).  The "gifmaker.py" script
  can be used to create GIF animations.

+ Reads 8-bit "L" and "P" TGA images.  Also reads 16-bit "RGB"
  images.

+ Added FLI read support.  This driver has only been tested
  on a few FLI samples.

+ Reads 2-bit and 4-bit PCX images.

+ Added MSP read and write support.  Both version 1 and 2 can be
  read, but only version 1 (uncompressed) files are written.

+ Fixed a bug in the FLI/FLC identification code that caused the
  driver to raise an exception when parsing valid FLI/FLC files.

+ Improved performance when loading file format plugins, and when
  opening files.

+ Added GIF animation support, via the "seek" and "tell" methods.
  You can use "player.py" to play an animated GIF file.

+ Removed MNG support, since the spec is changing faster than I
  can change the code.  I've added support for the experimental
  ARG format instead.  Contact me for more information on this
  format.

+ Added keyword options to the "save" method.  The following options
  are currently supported:

      format	option		description
      --------------------------------------------------------
      JPEG	optimize	minimize output file at the
				expense of compression speed.

      JPEG	progressive	enable progressive output. the
				option value is ignored.

      JPEG	quality		set compression quality (1-100).
				the default value is 75.

      JPEG	smooth		smooth dithered images.  value
				is strengh (1-100).  default is
				off (0).

      PNG	optimize	minimize output file at the
				expense of compression speed.

  Expect more options in future releases.  Also note that
  file writers silently ignore unknown options.

+ Plugged memory leaks in the PNG and TIFF decoders.

+ Added PNG write support.

+ (internal) RGB unpackers and converters now set the pad byte
  to 255 (full opacity).

+ Properly handles the "transparency" property for GIF, PNG
  and XPM files.

+ Added a "putalpha" method, allowing you to attach a "1" or "L"
  image as the alpha layer to an "RGBA" image.

+ Various improvements to the sample scripts:

  "pilconvert"  Carries out some extra tricks in order to make
		the resulting file as small as possible.

  "explode"	(NEW) Split an image sequence into individual frames.

  "gifmaker"	(NEW) Convert a sequence file into a GIF animation.
		Note that the GIF encoder create "uncompressed" GIF
		files, so animations created by this script are
		rather large (typically 2-5 times the compressed
		sizes).

  "image2py"	(NEW) Convert a single image to a python module.  See
		comments in this script for details.

  "player"	If multiple images are given on the command line,
		they are interpreted as frames in a sequence.  The
		script assumes that they all have the same size.
		Also note that this script now can play FLI/FLC
		and GIF animations.

		This player can also execute embedded Python
		animation applets (ARG format only).

  "viewer"	Transparent images ("P" with transparency property,
		and "RGBA") are superimposed on the standard Tk back-
		ground.

+ Fixed colour argument to "new".  For multilayer images, pass a
  tuple: (Red, Green, Blue), (Red, Green, Blue, Alpha), or (Cyan,
  Magenta, Yellow, Black).

+ Added XPM (X pixmap) read support.

(0.2b3 released)

+ Added MNG (multi-image network graphics) read support.  "Ming"
  is a proposed animation standard, based on the PNG file format.

  You can use the "player" sample script to display some flavours
  of this format.  The MNG standard is still under development, 
  as is this driver.  More information, including sample files,
  can be found at <ftp://swrinde.nde.swri.edu/pub/mng>

+ Added a "verify" method to images loaded from file.  This method
  scans the file for errors, without actually decoding the image
  data, and raises a suitable exception if it finds any problems.
  Currently implemented for PNG and MNG files only.

+ Added support for interlaced GIF images.

+ Added PNG read support -- if linked with the ZLIB compression library,
  PIL reads all kinds of PNG images, except interlaced files.

+ Improved PNG identification support -- doesn't mess up on unknown
  chunks, identifies all possible PNG modes, and verifies checksum
  on PNG header chunks.

+ Added an experimental reader for placable Windows Meta Files (WMF).
  This reader is still very incomplete, but it illustrates how PIL's
  drawing capabilities can be used to render vector and metafile
  formats.

+ Added restricted drivers for images from Image Tools (greyscale
  only) and LabEye/IFUNC (common interchange modes only).

+ Some minor improvements to the sample scripts provided in the
  "Scripts" directory.

+ The test images have been moved to the "Images" directory.

(0.2b2 released)
(0.2b1 released; Windows only)

+ Fixed filling of complex polygons.  The ImageDraw "line" and
  "polygon" methods also accept Path objects.

+ The ImageTk "PhotoImage" object can now be constructed directly
  from an image.  You can also pass the object itself to Tkinter,
  instead of using the "image" attribute.  Finally, using "paste"
  on a displayed image automatically updates the display.

+ The ImageTk "BitmapImage" object allows you to create transparent
  overlays from 1-bit images.  You can pass the object itself to
  Tkinter.  The constructor takes the same arguments as the Tkinter
  BitmapImage class; use the "foreground" option to set the colour
  of the overlay.

+ Added a "putdata" method to the Image class.  This can be used to
  load a 1-layer image with data from a sequence object or a string.
  An optional floating point scale and offset can be used to adjust
  the data to fit into the 8-bit pixel range.  Also see the "getdata"
  method.

+ Added the EXTENT method to the Image "transform" method.  This can
  be used to quickly crop, stretch, shrink, or mirror a subregion
  from another image.

+ Adapted to Python 1.4.

+ Added a project makefile for Visual C++ 4.x.  This allows you to
  easily build a dynamically linked version of PIL for Windows 95
  and NT.

+ A Tk "booster" patch for Windows is available.  It gives dramatic
  performance improvements for some displays.  Has been tested with
  Tk 4.2 only, but is likely to work with Tk 4.1 as well.  See the Tk
  subdirectory for details.

+ You can now save 1-bit images in the XBM format.  In addition, the
  Image class now provides a "tobitmap" method which returns a string
  containing an XBM representation of the image.  Quite handy to use
  with Tk.

+ More conversions, including "RGB" to "1" and more.

(0.2a1 released)

+ Where earlier versions accepted lists, this version accepts arbitrary
  Python sequences (including strings, in some cases).  A few resource
  leaks were plugged in the process.

+ The Image "paste" method now allows the box to extend outside
  the target image.  The size of the box, the image to be pasted,
  and the optional mask must still match.

+ The ImageDraw module now supports filled polygons, outlined and
  filled ellipses, and text.  Font support is rudimentary, though.

+ The Image "point" method now takes an optional mode argument,
  allowing you to convert the image while translating it.  Currently,
  this can only be used to convert "L" or "P" images to "1" images
  (creating thresholded images or "matte" masks).

+ An Image "getpixel" method has been added.  For single band images,
  it returns the pixel value at a given position as an integer.
  For n-band images, it returns an n-tuple of integers.

+ An Image "getdata" method has been added.  It returns a sequence
  object representing the image as a 1-dimensional array.  Only len()
  and [] can be used with this sequence.  This method returns a
  reference to the existing image data, so changes in the image
  will be immediately reflected in the sequence object.

+ Fixed alignment problems in the Windows BMP writer.

+ If converting an "RGB" image to "RGB" or "L", you can give a second
  argument containing a colour conversion matrix.

+ An Image "getbbox" method has been added.  It returns the bounding
  box of data in an image, considering the value 0 as background.

+ An Image "offset" method has been added.  It returns a new image
  where the contents of the image have been offset the given distance
  in X and/or Y direction.  Data wraps between edges.

+ Saves PDF images.  The driver creates a binary PDF 1.1 files, using
  JPEG compression for "L", "RGB", and "CMYK" images, and hex encoding
  (same as for PostScript) for other formats.

+ The paste method now accepts "1" masks.  Zero means transparent,
  any other pixel value means opaque.  This is faster than using an
  "L" transparency mask.

+ Properly writes EPS files (and properly prints images to postscript
  printers as well).

+ Reads 4-bit BMP files, as well as 4 and 8-bit Windows ICO and CUR
  files.  Cursor animations are not supported.

+ Fixed alignment problems in the Sun raster loader.

+ Added "draft" and "thumbnail" methods.  The draft method is used
  to optimize loading of JPEG and PCD files, the thumbnail method is
  used to create a thumbnail representation of an image.

+ Added Windows display support, via the ImageWin class (see the
  handbook for details).

+ Added raster conversion for EPS files.  This requires GNU or Aladdin
  Ghostscript, and probably works on UNIX only.

+ Reads PhotoCD (PCD) images.  The base resolution (768x512) can be
  read from a PhotoCD file.

+ Eliminated some compiler warnings.  Bindings now compile cleanly in C++
  mode.  Note that the Imaging library itself must be compiled in C mode.

+ Added "bdf2pil.py", which converts BDF fonts into images with associated
  metrics.  This is definitely work in progress.  For info, see description
  in script for details.

+ Fixed a bug in the "ImageEnhance.py" module.

+ Fixed a bug in the netpbm save hack in "GifImagePlugin.py"

+ Fixed 90 and 270 degree rotation of rectangular images.

+ Properly reads 8-bit TIFF palette-color images.

+ Reads plane separated RGB and CMYK TIFF images.

+ Added driver debug mode.  This is enabled by setting Image.DEBUG
  to a non-zero value.  Try the -D option to "pilfile.py" and see what
  happens.

+ Don't crash on "atend" constructs in PostScript files.

+ Only the Image module imports _imaging directly.  Other modules
  should refer to the binding module as "Image.core".

*** Changes from release 0.0 to 0.1 (b1) ***

+ A handbook is available (distributed separately).

+ The coordinate system is changed so that (0,0) is now located
  in the upper left corner.  This is in compliancy with ISO 12087
  and 90% of all other image processing and graphics libraries.

+ Modes "1" (bilevel) and "P" (palette) have been introduced.  Note
  that bilevel images are stored with one byte per pixel.

+ The Image "crop" and "paste" methods now accepts None as the
  box argument, to refer to the full image (self, that is).

+ The Image "crop" method now works properly.

+ The Image "point" method is now available.  You can use either a
  lookup table or a function taking one argument.

+ The Image join function has been renamed to "merge".

+ An Image "composite" function has been added.  It is identical
  to copy() followed by paste(mask).

+ An Image "eval" function has been added.  It is currently identical
  to point(function); that is, only a single image can be processed.

+ A set of channel operations has been added.  See the "ImageChops"
  module, test_chops.py, and the handbook for details.

+ Added the "pilconvert" utility, which converts image files.  Note
  that the number of output formats are still quite restricted.

+ Added the "pilfile" utility, which quickly identifies image files
  (without loading them, in most cases).

+ Added the "pilprint" utility, which prints image files to Postscript
  printers.

+ Added a rudimentary version of the "pilview" utility, which is
  simple image viewer based on Tk.  Only File/Exit and Image/Next
  works properly.

+ An interface to Tk has been added.  See "Lib/ImageTk.py" and README
  for details.

+ An interface to Jack Jansen's Img library has been added (thanks to
  Jack).  This allows you to read images through the Img extensions file
  format handlers.  See the file "Lib/ImgExtImagePlugin.py" for details.

+ Postscript printing is provided through the PSDraw module.  See the
  handbook for details.
