If the root item is a Window, then we use the default-font-* properties as the basis
for all unspecified font properties in text rendering.
This patch centralizes the code for retrieving that in the runtime window.
If this becomes a performance bottleneck, then we could cache it in a
box-pinned property in the run-time window.
- Have the cursor at the right place when clicking on end of line
- make sure not to render new lines if there are some when in single-line mode
- Enter should make a new line
since there is only one function in it, just put that function in the Window
then there is no need to heap alocate a dyn FontMetrics just to call this function.
In preparation of having mutiple-lines TextInput, we will need to give more
data to this function so it no longer belong in FontMetrics
Also remove the unused FontMetric::line_height()
The RenderCache (slab and generation) is always in a refcell, so we can
just pass that through. This also eliminates the ItemGraphicsCache
wrapper in the GL backend.
This will allow more fine grained borrowing in the future.
Fix one vtable_address_comparisons clippy link.
It is fine to only compare the data pointer.
Two different model with the same data pointer in a Rc could only be obtained
with tricks like transmute because each model is allocated on a different Rc.
* Change SharedVector::as_mut_slice() to SharedVector::make_mut_slice()
* SharedPixelBuffer::as_mut_slice and SharedPixelBuffer::make_mut_slice()
* SharedPixelBuffer::as_bytes_mut() to SharedPixelBuffer::make_mut_bytes()
This is consistent with Rc::make_mut() and is supposed to emphasize
that, unlike the usual as_* operations, it is not a free operation but
may involve work (detaching the data by cloning it).
Amends commit 7384ebdc32
There's only one place where we need this ourselves and that's also
easily done by hand. Otherwise the `as_bytes(_mut)()` accessor provides the
functionality along with `width()` and `height()`.
Other than that the API is mostly consuming, given that `sixtyfps::Image`
has no way of extracting the data again.
This adds an ImageBuffer and PixelBuffer type for SharedVector
backed images. The documentation explains how to use this
with low-level rendering functions and the popular image crate.
Fixes#387
* In capacity_for_grow, don't compare the number of elements
(current_cap) with the size of an element in bytes, but
with the required capacity instead (to detect that we don't need
to grow). Not all call sites check for that (i.e. push), so there's
a new test for that.
* When re-allocating due to growth and copying elements from the old
inner to the new inner, make sure to copy all old elements from the beginning,
not only the last element repeatedly.
