Data and coordinate model#

GraphX keeps three coordinate systems separate:

1. Source-image coordinates are intrinsic pixels in the original image.
2. Graph coordinates are calibrated X/Y values.
3. Viewport coordinates are browser-workspace display pixels.

Viewport conversion is conceptually $screen=offset+image\times scale$, with the inverse $image=(screen-offset)/scale$. Therefore zooming and panning only change the display transform, never the stored image point.

Authoritative positions#

Each data point has an imagePosition and, once calibration is available, a graphPosition. Which value is authoritative depends on how the point was created or changed.

Normal digitising#

Normal digitising starts with the intrinsic image position you select. GraphX derives the graph coordinate through the current calibration.

Recalibration#

On recalibration, the stored image position stays fixed and GraphX recalculates the derived graph coordinate. The source-image pick remains the durable geometric reference.

Direct coordinate editing#

Direct table editing takes the opposite route: GraphX preserves the typed graph coordinate and uses inverse calibration to calculate its corresponding image position. This keeps the direct edit geometrically tied to the source image without deliberately rounding the entered value.

Precision#

Dataset Precision controls formatting in tables and exports only; it does not quantise stored coordinates. This distinction also explains why PDF/JPEG overlays can be projected accurately against the original image and why extrapolated points can be represented while clearly marked as outside the calibrated range.