OrthographicCamera

• left, right, top, and bottom define the visible rectangle in scene units, not pixels.
• zoom scales the view uniformly and is safer than editing frustum bounds.
• setViewOffset() splits the view for tiled or multi-monitor rendering.

const VIEW_HEIGHT = 24; // 纵向可见 24 个场景单位

function fitOrtho(camera, width, height) {
  const aspect = width / height;
  const halfH = VIEW_HEIGHT / 2;
  const halfW = halfH * aspect;

  camera.left = -halfW;
  camera.right = halfW;
  camera.top = halfH;
  camera.bottom = -halfH;
  camera.updateProjectionMatrix();
}

const camera = new THREE.OrthographicCamera(-1, 1, 1, -1, 0.1, 500);
fitOrtho(camera, stage.clientWidth, stage.clientHeight);

camera.position.set(0, 120, 0);
camera.up.set(0, 0, -1);   // 让图纸上方对应世界 -Z
camera.lookAt(0, 0, 0);
camera.zoom = 1.8;
camera.updateProjectionMatrix();

• Changing bounds without updateProjectionMatrix() leaves the old projection in place.
• Aspect ratio mismatch stretches content; derive bounds from the canvas dimensions.