This includes both block drags and workspace scrolls and fixes#404. I
used isDragging rather than checking the origin of click so that it can
easily be extended to the case where the flyout is scrolled by dragging
a block, a la LLK/scratch-blocks#206.
* Add a new method to be called when the contents of the workspace change and
the scrollbars need to be adjusted but the the chrome (trash, toolbox, etc)
are expected to stay in the same place.
Change a bunch of calls to svgResize to either be removed or call the new
method instead. This is a nice performance win since the offsetHeight/Width
call in svgResize can be expensive, especially when called as often as we do -
there was some layout thrashing.
This also paves the way for moving calls to recordDeleteAreas
(which is also expensive) to a more cacheable spot than on every
mouse down/touch event.
of things (namely the scrollbars)
* Fix size of graph demo when it first loads by calling svgResize.
The graph starts with fixed width and was relying on a resize event
to fire (which I believe was removed in commit
217c681b86).
* Fix the resizing of the code demo. The demo's tab min-width used to
match the toolbox's width was only being set on a resize event, but
commit 217c681b86 changed how that worked.
* Fix up some comments.
* Use specific workspaces rather than Blockly.getMainWorkspace().
* Make workspace required for resizeSvgContents and update
some calls to send real workspaces rather than ones that are
null.
Remove the private tag on terminateDrag_ because it is only
actually called from outside the BlockSvg object.
* Remove a rogue period.
* Recategorize BlockSvg.terminateDrag_ to @package instead of @private so that
other developers don't use it, but it still can be used by other Blockly classes.
* Add a TODO to fix issue #307.
* Add @package to workspace resizeContents.
from getting selected. Use e.preventDefault instead of adding and removing
classes on mousedown and mouseup. This keeps the browser from having to
potentially recacluate style on mousedown and mouseup events.
css to mark text as unselectable instead. I can't prove this is
significantly faster with numbers since the call to setup the time out
costs ~.1ms and the call when the timer fires is ~.05. This happens on almost
every mouse event though. Plus, it seems to remove a flicker of highlighting when
frames are skipped and keeps scrolling via the scrollbar from highlighting.
with a manual calculation. getBBox() can cause the browser to
re-layout the whole page and we have much of the information
(e.g. individual block height and width) we need already cached.
Note that getBoundingRectangle's calculation is slightly different
than the one returned by getBBox. It is off by 3 in the y direction
due to how one of the curves is drawn. This new calculation is
technically more accurate.
