This is really intended to be a list of ideas that came up so they don't get lost. We are working towards making a real TODO with scheduling for the 1.4 and 2.0 release out of this. Please add things to this file when the need to do them is discovered. Explanations of why or how this would be useful are even better. Insight into possible ways to implement are even better than that. =================================================================== GUI / Functionality Separation Both file scope wise and in use, particulary so that the GUI version of the tool uses the pdb whenever possible (for finding bugs, and for macro recording) The core and the GUI have to reside in separate subdirectories. Macro recording and better scripting support: This pretty much is going to require the aforementioned gui/func seperation. More stuff needs to be triggered via pdb to make this a possibilty. Named undo Each pushing of an undo operation should include a textual string for human consumption. That way, the "undo" menu option can list exactly what will be undone. Its sorta there, but could be more specific. NOTE: With eyerything going thru the PDB, we'll get a much better undo system and macro recording more or less for free. This is schedulded for 2.0. Fix stuff so that the tile size could actually be changed eventually Currently gimp core is mostly setup to use a potentially variable tilesize, but lots of plugins and some internal stuff are hardcoded to expect 64x64 tiles. This is good for 8-bit images on x86, but with potential of deep images and other platforms, having this variable could be a real gain in performance tweaking. This will be hard because the XCF file format assumes 64x64 tiles. XCF load/save will have to be written to retile images, or the XCF file format redesigned. --sg Automagically guess whats a good tile_cache size: Not sure how to do this. Suggestions are offered in the install. Optimize transform_core (special cases, fft stuff?, optimizations only raph understands, etc...) Some sort of image/drawable locking: so we don't munge images by doing >1 ops on them at once. For plugins in particular, may also simplifie some of the tool redesign. Integrate gimp-16 stuff? Probably won't happen for 1.4 but will definitely be in 2.0 as GEGL. see http://film.gimp.org/ Action/active/Procedural Layers People seem to want them. Planned for 2.0. Previews in file save (for jpeg compression, etc): Currently there is no easy way to see the effects of compression or other image saving effects before actually saving and reloading an image. NOTE: The JPEG plug-in uses an awful hack to acomplish this goal. There needs to be an API for save plug-ins that provides this functionality. We should not allow another plug-in to follow the bad example of the JPEG plug-in. Generic preview code in libgimp for use by plugins All plugins should have a preview. it would be nice if there were code in libgimp to make this easier. Possibly just generalize a good exmaple of preview code (quartics?) and slap it in libgimp. Maybe throw some of the gck stuff in there too. Or possibly make a new plugin helper lib combining gck, megawidget, preview code, etc. "All" plugins might be too extreme. Some plugins (autostretch contrast) have little need for an interface, and exist IMHO better w/o one. Also, there are different needs for previews depending on the plug-in type, so a set of preview widgets would be a better solution. Should be done in 1.4. Indexed/color reducing to arbitrary number of colors: Current convert.c is limited to indexing to 8-bit palettes or less. Dont know what would be involved in making it work for larger palettes. convert.c has some deep magic involved in it, so who knows. (Related: indexed images with more than 8 bits depth. --sg) Active brushes: Use module system to dynamically load new brush types, eg Raph's(?) watercolour brushes, pixmap brushes, or image hose style brushes. Need to define a proper API for hooking. Maybe really want to define new tool types, rather than new brush types. Various Tool Ideas: Make tools listen for gimage dirty signals to fix munging. When a tool caches image data privately, you need to reset when the image gets dirty. Some degree of drawing tools, straight line, etc: Perhaps large parts of gfig could be salvaged for this. Its not something "paint" programs traditionaly do, but there usefulnes is obvious. square/circle/ellipse are already there basically (just make a wrapper to select and stroke). need a better straigh line drawing ui though. More Xinput stuff ( gradient brushes, pen "strokes" ?): This is very important for "artist" types. The more value-added utility we can make for Xinput stuff the better. BETTER FONT SUPPORT! Text should be redone entirely, with a native T1/TTF interpreter. Using X to render text is cheap but nasty. NOTE: Actually we should use FreeType (and probably Pango) for this. The gimp-freetype plug-in should be used as a test-bed to develop the next-generation text-tool.. (There's a good argument for making all text operations run thru an extension. --sg) See gimp-freetype for exciting work on going around X! "Font brush" a brush that could use chars rendered from fonts as the brush bitmap. limited utility for char's, but a couple of nice sets of winding fonts could make it interestings. And you could use words or strings. Very nice idea, indeed. Freetype would even allow to resize and transform the font based on pressure, speed, tilt, ... A clone tool that lets you rotate before writing to destination. A few GUI things suggestions... More configurabilty: I would like to be able to let the user stick arbitrary buttons into the toolbox and attach arbitrary functions to them. --sg What about a second button-bar? We dont have icons to add to the main toolbar anyway, so it would be hard to distinguish what button does what (they are small -> text wont fit). Look at my mockup: http://tigert.gimp.org/files/temp/wilber_palette.gif It might work in such way that 1. you click add -> new button 2. then right click on it -> you would get the normal -menu (like when clicking right mousebutton over an image) from which you could select the plugin/tool to assign to this button. 3. To apply the thing to an image you first click on the button and the statusbar would show something like 'click on an image to apply function xxxx' -> click an image to assign the tool an image and apply it. Now this is just an idea, but it might solve the 'which image do we want to apply things to? -problem. Also see http://www.home.unix-ag.org/simon/gimp/gimpbuttons.html Simon Budig has an experiment similar to this. Maybe it would make sense to implement something like this? --tigert Brush-shaped cursors: Possible solutions, generate cursors on the fly, use shaped pixmaps, possibly just draw directly on the preview, any other ideas? Raphs canvas might offer the oppurtunity to do this in color and anti-aliased. Crop tool could have an optional shader for the outside area so when you crop, the area outside would show in a darker color / user definable color / black / white etc.. (maybe a selector for the outside area: ____________ Outside fill: [|_none_______| = ] |.black......| |.white......| |.user def...| |.shade 50%..| `------------' That would be very handy for cropping scanned photos etc - one would see the result much easier apart from the rest. Suggestion from Ville Hautamaki (CW): Pattern groups Especially once patterns are demand-loaded. It would be very conveint to say "load all the wood patterns", or even just "set 1" etc. Same for brushes, gradients, fonts, ... For the pixmap/hose stuff: add a "pipe edit" option to the brush dialog so we can call up a hose editor (to reorder images, scale them, etc) NOTE: Probably overkill and difficult to implement in the current brush model. Users can use the brush and pipe load/save plug-ins if they really want to edit brushes. Grid: A gridding, and a snap to gridding. Basically a way of turning on and off a grid, and setting the grid size. Then also a way of starting to draw at a grid intersection and then end drawing at another. Have a "Preview" button too that performs the transformation without interpolation: As it stands, transforms can be very slow and feedback to the user is small. A preiview would save much time and make the action more accururate. Selections should be transformable: (e.g. rotate an elliptical selection; the selection, not it's content!!). This shouldnt be too hard. Perhaps just need to add the hooks for the selection info (usually in TileManager structs) to be passed to the existent image ops. Have a possibility to add a text as selection: If selections would be editable as described above, we'd then have editable vector-text. See the bezier option in gimp-freetype! Option to create New Indexed image (with the choice of palette foo). Big cad style cross-hairs cursor. Curve deal in the gradient editor?: It could be useful to have a curve widget for each of the color channels or hsv. For example, you have a custom palette you like, but you want it to get dark from left to right. You pop up the "Value" curve and make a curve getting darker. (Note: this would require a complete rewrite of the gradient _engine_, not just the editor. --sg) Redesign of the Blend Tool dialog? (it's rather large...) Let PDB stuff register under the layers_dialog menu: Lots of potential in scripts to do layer/channel manip. Might as well let them register on those menus. Some sort of mdi for dialogs so you could tab them together or pull them apart: Would be nice to be able to pile all the dialogs into one big notebook. Make color picker able to choose from any color on the screen. This sounds really simple. GTK-2.0 will does this for us. Code to set built in icons or icon path? Is this the wm's job or soemthing internal? Could be nice for plug-ins to register an icon for internal use. Progress bars on long proccesses: Currently, long-term ops like blend run from a recursive event loop, not the idle loop like they should. This should be cleaned up a little. -- Austin. "Open into Layer" might be a good idea. Ability to get an exact count of the total number of colors in the image: Xpm plugin has an algo to do this. Should be trivial to add this to the core.