You're assuming the rotated line maintains the same pixelwise thickness. A straight 1 pixel line edge-to-edge may not translate to a diagonal 1 pixel line corne

You're assuming the rotated line maintains the same pixelwise thickness. A straight 1 pixel line edge-to-edge may not translate to a diagonal 1 pixel line corner-to-corner. I think that's what the parent is saying. No data is being 'compressed' here; the shear operation simply preserves all pixels uniquely.