The objective of these changes is to create a more testable doc/logic API.
Included changes:
- Added doc::Context, doc::Documents (the old app::Documents),
and doc::Sprites (and observers for each)
- Added raster::Sprite::createBasicSprite()
- Added doc::ColorMode (equal to raster::PixelFormat)
- Added some methods to doc::Document: context(), sprites(), sprite()
(to replace app::Document::getSprite()), width/height(), colorMode(),
name(), and close()
- Moved app::DocumentObserver/Event to doc::DocumentObserver/Event
- Replaced app::ContextObserver with doc::DocumentsObserver and a couple
of signals.
- Renamed app::Context::getActiveDocument() with
doc::Context::activeDocument()
- Renamed app::Context::getActiveLocation() with
app::Context::activeLocation()
- Removed app::ContextObserverList
- Removed app::DocumentId (now we can use doc::ObjectId)
- Removed app::Context::getSettings()
This is a work-in-progress, dropRange() must be reimplemented. Now that
the Timeline uses LayerIndex, we could change Timeline::Range to
DocumentRange and create unit tests for the logic side.
There are cases where we need to modify the list of observers of certain
entity when we are in a notification loop (i.e. iterating its observers).
E.g. A general update notification about the current document to all its
observers could create the mini editor, which is a DocumentObserver, so
a new observer is added to the list in the same notification loop.
Anyway, as we cannot modify the observer list (std::vector) when we are
notifying them (any modification in the std::vector invalidates
its iterators), the fix is quite easy (but not optimal): we can create
a copy of the observers list so we can iterate the list.
Note: If we have performance issues about this, we could try a std::list,
but at the moment this fix is quite enough.
On WinXP classic theme we receive a WM_NCHITTEST messages for the
scrollbars, and as they overlap the resize borders (see WM_NCCALCSIZE
handler) we have to return a proper value as if they really were the
borders of the window.
Note: Scrollbars are still visible in classic theme, that is just ugly,
but at least the user can resize the window from bottom and right edges.
This is an huge refactor to avoid handling Allegro FONT directly. Some
changes:
* Add she::System::defaultDisplay/Font, createRgbaSurface, loadSurface,
and loadRgbaSurface.
* Rename she::CreateSystem/Instance to she::create_system/instance.
* Remove ui/font.cpp and move ui/fontbmp.cpp to she library.
* ui::IButtonIcon uses she::Surface instead of BITMAP.
* Rename she::LockedSurface::drawAlphaSurface -> drawRgbaSurface
* Rename ui::SetDisplay -> set_display
* Rename _ji_font_text_len -> ui::Graphics::measureUIStringLength
If the transparent/mask color has alpha = 0, we cannot filter by
RGB/Grayscale values, we have to blend all colors anyway. Only when
the mask has alpha > 0 we can check if the RGB values are different to
the mask RGB values. In other words, comparing the source color with the
mask, makes sense only when the mask has alpha > 0.
The new default location is %AppData%/Aseprite folder. This will be useful
for a future setup program. So if aseprite.ini is located in aseprite.exe
directory, it acts like a portable program, in other case it acts like
an installed program.
When we create a new image for the sprite, we've to adjust the image mask
color to the sprite mask color (as the image isn't in the Stock yet).
This problem appears in v1.0.1 (9dfec919e4)
because now we don't change the mask color of sprite images in the
rendering process. This kind of bug can be detected running on debug
mode with an assert in the render procedure.
This bug is because the mask color of cel images were fixed when they
were used in the rendering process. Now, the mask color is fixed when the
image is added to the raster::Stock structure.
Simplified the color selector UI. Now RGB/HSV colors are shown correctly
and a new warning icon/button is visible when the color is not part of
the current palette. This button can be used to quickly add the new entry
to the palette.
- Add ContextBarObserver (and MovingPixelsState implements this interface)
- PixelMovements::discardImage() receives a "commit" flag to do a rollback
of the operation if the user needs it.
- Add red/blue onion skin type
- Add app::calculate_next_frame() function to calculate the next frame
depending of the timeline configuration (loop range, animation direction)
- Add app::finder() to simplify the access to widgets loaded from xml files
- If we use a RGB color in a background layer (indexed image),
and the first palette entry (transparent color) matches that RGB color,
we can use that entry, because this is the background (the transparent
color isn't transparent at all in the background layer).
- If we use the same RGB color in a transparent color, we've to avoid the
transparent color, because the user want to paint with that specific
solid color.
- Palette::findBestfit() receives a mask_index now, so we can find what
color best matches a specific RGB avoiding that specific mask color,
or we can use -1 to use any color (e.g. for background layers).
- Added app::ColorTarget() to simplify and fix color_utils::color_for_layer
and app_get_color_to_clear_layer(), so now we can use the new findBestfit
to return the transparent color for background layers (if a RGB color
matches the transparent color RGB values).
- Removed fixup_color_for_layer/background() functions in color_utils
- Fix NewImageFromMask() to use the mask color to clear the image
- Improve the Editor pen preview (cursor.cpp) to draw the bounds of the
pen when it will paint with the transparent color.
This field is disabled, and the program isn't ready to support less than
256 colors in Indexed mode, so it's better if we remove this field to avoid
confusion.
We have to wait to validate not-scrolled regions before we can use
Widget::scrollRegion() or Window::moveWindow() again. This is an issue
to see in the future (to avoid lossing mouse move messages). At the moment,
one solution is to use the last received mouse move message.
There were problems calling a pure virtual function (IFileOpProgress
implemented by OpenFileJob) when we are already in ~Job() dtor. So we've
to wait the background thread (added Job::waitJob() function) to join
the thread so it can use IFileOpProgress safely.
Also the save process of .ase files now can be cancelled (it wasn't
possible before).
This is a common issue when the save process fails, we get an .ase file
without header. So basically with this change we save the most important
information of the header first, and then save the rest of frames/layers
info. At the end we can save the file size.
- Other change: removed "current_frame_header" static variable.
The problem was in DocumentApi::moveCel(). We cannot move a Cel between
layers and frames at the same time easily (it's not possible to do it
without being in a temporal invalid state, e.g. where two cels are in the
same frame position). It's better if we create completely new cel for
the target and remove the previous one in all cases (undoers are already
prepared to do those operations correctly).
- Add Sprite::getImageRefs() member function.
- Fix double-click behavior when double-clicks are generated from
Manager::generateMouseMessages().
- Add Widget::setDoubleClickeable/isDoubleClickeable() member functions
for widgets that accept double-clicks.
We planned to migrate everything to ui::Graphics to port the program to
other back-end (e.g. SDL or Allegro5). This is one step forward to that
direction.
Changes:
- Delete a lot of deprecated functions from src/ui/draw.h/cpp and SkinTheme
that used BITMAP/ji_screen.
- Rename ui::ji_move_region to ui::_move_region
- Move ui::jwidget_get_texticon_info to ui::Widget::getTextIconInfo.
Now the member function returns client coordinate instead of absolute
ones.
- ui::drawTextBox() now receives a ui::Graphics instead of a BITMAP
- Add ui::Graphics::drawChar/measureChar() member functions.
- Replace ui::jrectexclude() with with ui::Graphics::fillAreaBetweenRects()
This fix some strange scenario/bug where slots iterators where invalidated
in the middle of the signaling process. (The whole signal/slot impl needs
a review/simplification.)
On Windows, instead of polling Allegro mouse position/buttons we can use
the mouse messages (WM_*) directly to generate she::Events. Those events
are received by the ui::Manager and converted to ui::Messages.
Maybe this is a possible fix for issue #133 (Wacom tablets don't work
properly).
Changes:
- Don't use jmouse_z() directly (new ui::MouseMessage::wheelDelta() member)
- Add ui::_internal_set_mouse_position() to change the jmouse_x/y(0)
from the new mouse position received in she::Events. The same for
ui::_internal_set_mouse_buttons().
- Modify the ui::Manager to generate mouse events in any case: using old
Allegro 4 polling method, or from she::Events.
At the moment the program doesn't support palettes with less than 256
colors. But at the future the idea is to load the number of colors that
the palette says.
This is an old bug in the ui::Manager. We don't need to "free" (release)
the mouse widget when we re-stack windows order (e.g. because the user
clicked the non top-most window).
- Merge Timeline::STATE_MOVING_LAYER/CEL/FRAME to STATE_MOVING_RANGE.
- Remove celmove.h/cpp and move_cel/copy_cel functions. Now they are
in the Timeline as dropRange/Cels/Frames/Layers member functions.
- Add DocumentApi::copyFrame/moveCel/copyCel member functions.
- Add timeline_drop_layer_deco and timeline_drop_frame_deco skin parts.
- Move code from DuplicateLayerCommand::onExecute() to new member
function DocumentApi::duplicateLayer().
- Fix a bug changing Cel's frame number: we weren't updating the Cel
position properly inside the LayerImage::m_cels collection. Now we use
LayerImage::moveCel() to change the Cel frame number.
- Other fixes to DocumentApi: Change bgcolor from int to color_t.
With this we can avoid confusion in the execution of commands like
Clear, that depending on the enable/disable state of the timeline range
it affects the timeline (clearing the whole range) or the sprite editor
(clearing the selected area). So now, when a sprite editor is clicked
we disable the range in the timeline, it's a way to say "as you clicked
this editor, commands will affect the editor, if you click the timeline
again, commands will affect the timeline".