Flamegraph prototype + text improvements

[janpaul123]

Adds a basic "flamegraph prototype" which is eventually to be integrated with Pyroscope; but a bunch more work has to be done for that (#177).

This also refactors text rendering a bunch so that we get ellipsis truncation even when using custom layouting. And adds some unit tests to that.

Best viewed by individual commits!

[janpaul123]

cc @thomasballinger if you're interested in having a look :)

[stevekrouse]

Why don't we use a JSON representation and then parse it, given that we already have to be able to deal with their standard JSON representation? (Seems like work to have to maintain this file)

[janpaul123]

Good point. I'm not sure what their internal format is — whether it is that JSON or if they first transform it to something else internally. Once I've figured that out I'll update this fixture to what makes most sense.

[stevekrouse]

I don't see any assertions here, so this test just makes sure that no error is thrown? If so, is there to be explicit about the fact that that's what's being tested here?

[janpaul123]

Ah this is not a test; this is a helper for use within tests. I'll add a comment! (Tests would be marked with #[test])

[stevekrouse]

When I demoed this to someone recently (I forget whom), they asked "wait does this handle function run on every event Zaplib detects? Every mousemove? How is that not super slow and inefficient?"

(I share this question)

[stevekrouse]

Could be nice to have indentation but this isn't great:

        self.window.begin_window(cx);
        {
            self.pass.begin_pass(cx, Vec4::color("300"));
            {
                self.main_view.begin_view(cx, LayoutSize::FILL);
                {
                    cx.begin_padding_box(Padding::top(30.));
                    {
                        self.flame_rects.resize_with(self.spans.len(), Default::default);
                        for (span_index, span) in self.spans.iter().enumerate() {
                            self.flame_rects[span_index].draw(cx, span_index, span, self.zoom_pan);
                        }
                    };
                    cx.end_padding_box();
                };
                self.main_view.end_view(cx);
            };
            self.pass.end_pass(cx);
        };
        self.window.end_window(cx);

I think this is valid syntax but cargo fmt doesn't like this:

        self.window.begin_window(cx); {
          self.pass.begin_pass(cx, Vec4::color("300")); {
            self.main_view.begin_view(cx, LayoutSize::FILL); {
              cx.begin_padding_box(Padding::top(30.)); {
                self.flame_rects.resize_with(self.spans.len(), Default::default);
                for (span_index, span) in self.spans.iter().enumerate() {
                    self.flame_rects[span_index].draw(cx, span_index, span, self.zoom_pan);
                }
              }; cx.end_padding_box();
            }; self.main_view.end_view(cx);
          }; self.pass.end_pass(cx);
        }; self.window.end_window(cx);

[stevekrouse]

Regardless we'll need to explain what's happening here and how you can imagine it like JSX but with imperative syntax (to get around the borrow checker)

[stevekrouse]

We may want to explain here why we can pass a pointer to the span down into the FlameRect component but not back up (we'd need to copy it or use a ref counter, which is similar to a weak ref in JS, because of lifecycle issues which always happens for recursive things like this)