Path2D cannot be generated dynamically

Allow using a clip mask to Plotter improve fill/compositing performance.

Allow all stroke parameters to vary while stroking. Use a second iterator which is zipped with Iterator to specify them.

For reusing a Plotter to reduce number of allocations, it's necessary to clear the contained Raster before each composite. The way this must be done currently is:

let mut pr = p.raster();
pr.clear();
p = Plotter::new(pr);
p.set_transform(Transform::{...}); // Reset the transform

It would be more elegant to have an API to do either:

p.raster_mut().clear();

or

p.clear();

Currently, Raster and Mask widths are assumed to be divisible by 4.

Example triggers panic: #20

use footile::{Plotter, PathOp::*};

mod png;

fn main() -> Result<(), std::io::Error> {
    let path = vec![
        Move(8.0, 4.0),
        Line(8.0, 3.0),
        Cubic(8.0, 3.0, 8.0, 3.0, 9.0, 3.75),
        Line(8.0, 3.75),
        Line(8.5, 3.75),
        Line(8.5, 3.5),
        Close(),
    ];
    let mut p = Plotter::new(64, 64);

    png::write_mask(p.fill(&path, footile::FillRule::NonZero), "./overlapping.png")
}

thread 'main' panicked at 'assertion failed: dy > Fixed::ZERO', src/fig.rs:121:9

  12:     0x560a45c0b21b - footile::fig::Edge::new::ha025b804eef6ddc7
                               at src/fig.rs:121
  13:     0x560a45c0e249 - footile::fig::Scanner::edge_regular::hded8104148dc4b5a
                               at src/fig.rs:636
  14:     0x560a45c0dc4a - footile::fig::Scanner::update_edges::h8eb063b97b6c4e32
                               at src/fig.rs:594
  15:     0x560a45c0d058 - footile::fig::Scanner::scan_vertex::hb7e7c4706bd86814
                               at src/fig.rs:483
  16:     0x560a45c0cd71 - footile::fig::Fig::fill::h494ccc1f0bc5b08e
                               at src/fig.rs:439
  17:     0x560a45b93381 - footile::plotter::Plotter::fill::hf6b17fb11e1691d2
                               at /home/aldaron/lordaldaron/footile/src/plotter.rs:331
  18:     0x560a45b90b7a - overlapping::main::h878a2bc11f333978
                               at examples/overlapping.rs:17

The following code generates two images that don't render correctly.

Code

use footile::{FillRule, Plotter, PathBuilder};

fn main() -> Result<(), std::io::Error> {
    let mut p = Plotter::new(64, 64);

    let path = PathBuilder::new()
        .absolute()
        .move_to(15.562022, 10.953022)
        .line_to(15.562022, 38.120804)
        .line_to(11.504303, 28.536913) // extra pixel here
        .line_to(11.504303, 10.953022)
        .build();

    p.fill(path.iter(), FillRule::NonZero).write_png("test_a.png")?;

    let path = PathBuilder::new()
        .absolute()
        .move_to(37.02108, 10.953022)
        .line_to(41.546997, 10.953022)
        .line_to(41.546997, 28.536913)
        .line_to(41.0788, 38.120804)
        .line_to(37.48928, 38.120804)
        .line_to(37.02108, 28.536913)
        .line_to(37.02108, 10.953022)
        .build();

    p.fill(path.iter(), FillRule::NonZero).write_png("test_b.png")
}

Results

Here they are:

• test_a.png
  
• test_b.png
  

test_a has a pixel with a greater fill value than it should and test_b has a pixel with a lower fill value than it should.

This matches Cairo, Skia and most other graphics libraries. It also allows for much better performance because we avoid the division on each blend.

It would be helpful if PathOp derived Debug, Copy, Clone, etc.

In the documentation it says:

pub fn absolute(self) -> Self

Use absolute coordinates for subsequent operations.

This is the default setting.

But testing out the code, relative is the default.

When creating a larger Raster, it should be possible to composite small masks. This will greatly reduce compositing time.

Hello,

I found out that leaving enabled the use-simd feature, every example fails to run when compiled in normal unoptimized mode. But only heptagram and drop fails to run when compiled in --release mode.

But disabling the use-simd feature makes them all work in normal mode. But interestingly drop and fishy fails in --release mode.

This may hint to a bug in one of the unsafe code blocks.

GDB output

Since drop fails in 3 of the 4 combinations of feature and mode, this is the relevant gdb output from running them:

With use-simd enabled

cargo build --example drop && gdb ./target/debug/examples/drop

(gdb) run
Starting program: .../footile/target/debug/examples/drop 
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".

Program received signal SIGILL, Illegal instruction.
core::core_arch::x86::sse2::_mm_slli_si128_impl (a=..., imm8=8) at /rustc/fc50f328b0353b285421b8ff5d4100966387a997/src/libcore/../stdsimd/crates/core_arch/src/x86/sse2.rs:429
429	/rustc/fc50f328b0353b285421b8ff5d4100966387a997/src/libcore/../stdsimd/crates/core_arch/src/x86/sse2.rs: No such file or directory.

cargo build --release --example drop && gdb ./target/release/examples/drop

(gdb) run
Starting program: .../footile/target/release/examples/drop 
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".

Program received signal SIGILL, Illegal instruction.
0x00005555555613a2 in <footile::gray8::Gray8 as footile::pixel::PixFmt>::over ()
(gdb)

With use-simd disabled

cargo build --release --example drop && gdb ./target/release/examples/drop

(gdb) run
Starting program: .../footile/target/release/examples/drop 
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".

Program received signal SIGILL, Illegal instruction.
0x0000555555560a52 in <footile::gray8::Gray8 as footile::pixel::PixFmt>::over ()
(gdb) 

System Info

Ubuntu 18.04 kernel 4.15.0-48-generic

CPU:

processor	: 5
vendor_id	: AuthenticAMD
cpu family	: 16
model		: 10
model name	: AMD Phenom(tm) II X6 1055T Processor
stepping	: 0
microcode	: 0x10000dc
cpu MHz		: 803.788
cache size	: 512 KB
physical id	: 0
siblings	: 6
core id		: 3
cpu cores	: 6
apicid		: 5
initial apicid	: 3
fpu		: yes
fpu_exception	: yes
cpuid level	: 6
wp		: yes
flags		: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm 3dnowext 3dnow constant_tsc rep_good nopl nonstop_tsc cpuid extd_apicid aperfmperf pni monitor cx16 popcnt lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt cpb hw_pstate vmmcall npt lbrv svm_lock nrip_save pausefilter
bugs		: tlb_mmatch apic_c1e fxsave_leak sysret_ss_attrs null_seg amd_e400 spectre_v1 spectre_v2
bogomips	: 5625.75
TLB size	: 1024 4K pages
clflush size	: 64
cache_alignment	: 64
address sizes	: 48 bits physical, 48 bits virtual
power management: ts ttp tm stc 100mhzsteps hwpstate cpb

This is an important piece of functionality for a 2d library. I have been playing around with the code needed to support this kind of thing here: https://github.com/jrmuizel/sw-composite/blob/master/src/lib.rs

Feel free to crib anything from it that you'd like.

Transforming within a path allows rotation, etc.