There should be far fewer hatched objects than other
objects, so we're spending too much effort finding
all the possible damage when we probably should
just be auto-regenerating all the hatching.
This also moves it out of being done during redraw,
which was proving problematic. Plus the refill
zones architecture does it during commit, and has
a lot more miles behind it.
Make the VIEW* parameter const. Since PCB_TEXT does a null-check,
it's not very clear if this can ever be null (and if it is,
why don't the other VIEW_ITEMs check?), so don't make them
all references too at this time.
Also dereference a few pointers a bit earlier to make non-null
promises sooner rather than later.
Saves a mishmash of local HIDE/SHOW defs along with various literals.
Also provide a function that computes the scale at which a given
IU size becomes (notionally) a certain size on screen. While this
is a simple division, it's a bit opaque in terms of meaning.
Also it means the divide by zero case can be more universally
defended-against, which has traditionally been a bug opportunity.
Recommendation is to avoid using the year nomenclature as this
information is already encoded in the git repo. Avoids needing to
repeatly update.
Also updates AUTHORS.txt from current repo with contributor names
PCB_SHAPE::Normalize() normalized a graphic segment by (sometimes) swapping
start and end points. But it creates an issue when flipping a segment: the
reference point can be changed.
This normalization was made for DRC purpose when comparing 2 footprints.
Now, PCB_SHAPE::Normalize() does not change end points of a segment, and a
specific PCB_SHAPE::NormalizeForCompare() normalize fct is used for DRC compare.
There was a gentle mish-mash of booleans, some with
true being left/right and some up/down, and some functions that
can flip in both axes (which is never actually done, and doesn't
really make geometric sense).
Replace all this with the FLIP_DIRECTION enum class, which makes
the intention completely unambiguous.
This also then allows a small scattering of simplifications,
because everything takes the same type and you don't have to
fiddle booleans to fit.
We are derived from std::vector now. We don't need our own prima donna
increment and dereference overload as these just create bloat and
non-standard coding practices
Add direct handling of quadratic beziers to save compute time and number
of points. Update cubic interpolation to reduce number of points
generated for a given smoothness
Cache data on open and used cached data to avoid multiple re-calcs
Remove minimum line length and number of segments and replace with
standard max error level. Allows us to specify the tolerance of bezier
interpolation
(Copper shapes, on the other hand, can. Since we use
GetItemDescription() to report DRC errors and the like,
it's better to have it report the net even when it's
<no net>.)
(This prevents rectangle from not getting combined into pad if
proxy flag is turned off as Clipper collapses the polygon union
because it's too small.)
Jeff Young