revert the precalculation of the sum of the coefficients - it

[imager.git] / TODO

                         *** TODO ***

Release Plans (subject to change)
-------------

brown-bag bugs may add intermediate releases.  The dates are goals,
not commitments.

0.44: (december 2004)
- set i_format for every file type on read and test for it:
  - bmp - done
  - jpeg - done
  - tiff - done
  - tga - done
  - rgb
  - png - done
  - gif - done
  - pnm - done
  - raw - done
- check each file reader for possible integer overflows
  - bmp - done
  - tiff
  - tga
  - rgb
  - png
  - gif
  - pnm
- check bmp code uses image data offset correctly - done
- check quant code for integer overflows - done
- check for old URLs (umich and imager.perl.org/~addi/...) - done

0.45: (march 2005?)
- audit pnm.c
- audit tga.c
- audit rgb.c and add tests
- capture TIFF read warnings (i_warnings tag?) (done)
- add Imager::Cookbook with at least 5 recipes
- implement i_incomplete for bmp files
- store floating point tags in an appropriate precision (done)
- allow image creation to fail on malloc() failure for the image data
  rather than aborting.  The current behaviour could be a denial of
  service attack.  This has limited utility because of the way memory
  over commitment works on Linux (reverted)
- eliminate unused variable warning (especially the RETVAL ones in Imager.xs)
- add META.yml (bypass EU::MM's limited mechanism)
- check if freetype 2 is faster/slower than freetype 1 with Imager's glyph
  caching.  If FT1 is faster, add caching as a TODO for FT2.
- handle probing for libraries better (use pkg-config or I<library>-config
  where possible (freetype2 and libpng)) (done)
- add concept index to Imager.pm pod
- replace testimg/penguin-base.ppm with a smaller image
- add a sample CGI HTML and image generation scripts that work together,
  with appropriate security management
- examples for fountain filter in Imager::Filters

0.46: (june 2005)
- allow limits to be set on the size of an image read from a file.  This is
  to prevent an attacker supplying huge images that consume all of 
  memory causing a denial of service attack.
- add Imager::Tutorial (see Tk::UserGuide for a structure)
  don't cover installation - belongs in README or INSTALL
  it doesn't need to cover everything - read/write/create/simple drawing
  and simple text - other stuff belongs in the cookbook or in sample code
- add 5 more recipes to Imager::Cookbook
- implement gsamp()/gsampf()/plin() etc methods for those low level image
  interfaces which don't yet have methods.
- implement i_incomplete for png and gif files.

Speculation
-----------

Some of what's listed below has been implemented.

- should i_incomplete be optional?  ie. reading a short image results in
  an error unless the user requests incomplete images.

Iolayer:
- Add scalar/mmap to iolayer
- Add close() code to iolayer for fakeseek sources.
- Merge callback interface into iolayer
- and raw
- Implment the maxread threshold (Indicates how much can
  be read from a source before it stops reading from its
  underlying source.  For making sure only the relevant
  data is read from a socket for example. 

Enhanched internal structure:

MultiImage & metadata support:
- Figure what interface should be between C and perl?
- How to store data in the C interface for tags/metadata?
  Support tiff's madness, i.e. tags and options.
- Error handling, make a general i_errstr object and arrange
  such that each image has an i_errstr object to hold its
  local errors?
- SEE design/represent.txt for proposed new structure and
  interface design that takes these factors into account.
- define common i_* tags for specifying attribute common among images
  like spatial resolution (implement for other image types, especially 
  TIFF)  (Spatial resolution is supported for all types that support 
  it - are there any other common properties we can add?)
- implement the i_incomplete tag for other formats  

New Features:
- Add mng support, pcx and aalib support.
  - Windows icon files (.ico)
  - ILBM (Amiga) images
  - photoshop files (I think I've seen docs)
  - XBM
  - FITS
  - WMF (extract bitmap data on read)
  - gzip or bzip2 compressed raw
  - postscript for output
  - PAM file support (man pam in recent PNM tools)

- Transforms, interpolated multidimensional lookup tables.
  Usefull for CMYK <-> RGB table lookup.

- advanced font layout (spacing, kerning, alignment) (Artur?)

- ways to check if characters are present in a font, eg. checking if
  ligatures are present (done for FreeType2)

- font synthesis - synthesize a bold or slanted font from a normal font
  (or even from an existing bold or slanted font)

- utf8 support for text output
  (available for FT1, freetype2, should be easy for Win32)

- easy interfaces for text output:
  - align text around point, including:
    - left, center, right horizontal alignment
    - top, center, baseline, bottom alignment
    - choose whether to base alignment upon character cells or font 
      metrics to avoid spacing problems between lines
  - fill a box from text, with newline handling
    - an option to just return the information that would be used to 
      output the text, so it can be used for sizing calculations
    - how much text fit in the box/how much space is left in the box
  - fill a box with rich text (size and font changes)
    - as above
    - wrapping around boxes/images
  - pod2Imager :)

- tools to get more information about fonts/characters:
  - name of font "Arial Bold Italic" vs "arialbi.ttf"
  - ABC widths for characters (currently the only way to get the C width 
    for a character is to get the width with a following space and without
    and then compare)
  - character names

- encodings for text output (maybe using the Encode module to work 
  between encodings)

- image rotation, 3 ways of doing rotation:
  - rotation by shearing, which produces makes lengths in the image larger,
    but could be useful
  (2 of the 3 have been done)

- read_multi() needs to handle other multi-image types,
   (handles TIFF and GIF)

- write_multi() to save other multi-image types, (handles TIFF and GIF)
  - pnm binary formats support multiple images per file

- compose channels - build a new image based on channels from several
  images

- arc outlines

- complex lines/curves:
  - thick lines:
    - dotted/dashed lines
    - doubled lines?
    - end-point controls (butt-end, square-end, round-end, pointy, arrows)
  - brush-lines (as with the Gimp, or Photoshop)
    - more complex with multiple strokes and various randomness as with
      Fractal Design Painter (or later versions of Corel Photo Paint)
  - complex line shapes:
    - lines
    - general curves (cubic or bezier, whichever is more general)
    - easy access to general curves:
      - points to make a nice curve from
      - various circular curves (centre, radii, start/end angles, 
        direction)
  - Complex thick lines are probably more useful than brush-lines for 
    the things I expect Imager to be used for.
  - http://www.develop-help.com/imager/strokes.png (.jpg for smaller version)

- have Imager::Color{::Float}? use Graphics::ColorNames if present.

Clean up:
- Make sure everything is doable with the OO interface
  i_flood_fill() for example.
  - include low-level functions like gsamp() and so on

- Compile with memory debugging enabled and fix all leaks

- Add mycalloc() memory allocation wrappers.

- should we overload <=> or cmp for Imager::Color objects?
  too many worms, leave it out

- should the interlace option for reading raw images have 
  symbolic names in the OO interface?

- improve consistency in the interfaces, both in trying to use named
  parameters where possible but also in using consistent names, eg.
  (tx, ty) vs (left, top) in rubthrough() vs paste().  Do we want a (simple)
  standard for this (suggested names)?  Remember to preserve the old 
  interface where possible.

- try to clean up the inconsistencies between font types:
  - utf8 (even if we just treat characters over 0xFF as missing for T1)
    (done for FT2, FT1, T1)
  - transformations (done for FT2)
  - has_char() method (done for FT2, FT1, T1)

Format specific issues:
- provide patches for libgif and libungif that fix their bugs
  and give a useful extension interface.  Probe for the 
  installation of the patches in Makefile.PL to let gif.c
  know what features it can use.

- Add options for pnm writer to save in any of the p1..P6
  formats.  Even if the input has 1 channel, write 3 and such
  cases.  Also allow ascii mode.  Need to be able to write 
  pbm images which needs ties to the quantization code.

- save paletted images as paletted where that's supported.  Done
  for gif/tga/bmp/tiff.  Not done for png yet.

- read other format paletted images as paletted images.  This has 
  been done for gif/tga/bmp/tiff formats but not for png.

- read/write 16-bit/sample images as such for tiff

- read more metadata from images, esp tiff tags, EXIF format information 
  from TIFF and JPEG.

- handle 16-bit/sample pgm/ppm files

Documentation:
- Add to the documentation
- Write a tutorial?
- sample code and Imager/Samples.pod describing them
- Imager/Cookbook.pod
- Write a guide to installing the helper libraries
- Go through the entire project and add comments in pod
  so doco.perl can be used to read them.
- see the IRC log in design/doclayout.txt
- method index?
- concept index?

Build/configure:
- use prompt() (from ExtUtils::MakeMaker) for questions in Makefile.PL
- use the library's mechanism's to find the library if needed, for
  example, freetype-config for FT2
- the freetype2 library is built as "freetype.lib" on Win32 and needs
  to be renamed to be found by the configuration process and included
  with -lfreetype.  Try to work around that.
- freetype2 and freetype1 tend to be installed in /usr/include/freetype2
  and /usr/include/freetype directories, or similarly named (somedir/freetype
  for example), instead of putting these directly into the search path
  search for the freetype{2}?/someheader and then add the subdirectory if
  we find the file in that directory.
- switch to Test::More
- change headers to use more unique re-include protection macros