get (X, Y) coordinates of a node in sankey plot

ファイル、テスト、エントリーポイントは指定されていません。まず plotly.py のソースコードとドキュメントで Sankey ノードの位置処理を追跡し、次に、サポートされる座標アクセサーまたは正確な変換動作を仕様として定められるかどうかを判断してください。ドラッグされたノードを対象とするカバレッジを含め、要求された座標または文書化された制限が確立されれば完了です。

Problem definition

I wrote an application that loads properties of a sankey plot from a file and displays it. I'd like to be able to manually (drag and drop) arrange node positions and then save their (X, Y) coordinates to a file to be able to continue the work later. For that I would need to get access to (X, Y) coordinates of nodes but I haven't found anything about that in the documentation and the source code.

Applied workaround

My current workaround is to read HTML source of generated plot and estimate (X, Y) coordinates based on parameters of translate functions defined for each node. However, this approach is inaccurate. Here's simplified code describing a single node:

<g class="sankey-node" transform="translate(13.260000228881836,401.4360046386719)">
    <rect class="node-rect" height="185.32853214031275" width="20.000000000000004">
    </rect>
      ...
</g>

The above node, in a python code, has X coordinate set to: 0.01 and Y coordinate set to 0.45. My current formulas to estimate python coordinates (X*, Y*) based on SVG translations (X, Y) are:

X* = (X - (node_width / 2)) / (plot_width - (margin_left + margin_right) - (padding_left + padding_right))
Y* = (Y - (node_height / 2)) / (plot_height - (margin_top + margin_bottom) - (padding_top + padding_bottom)) + node_height * 0.0009

Referring to the pasted SVG code and corresponding python coordinates the formulas would do the following:
X -> X* : 13.260000228881836 -> 0.01
Y -> Y* : 401.4360046386719 -> 0.45

I wrote that the above formulas are inaccurate because they:
[1] Reproduce python (X, Y) with some error.
[2] The error is affected by a plot size but it's not clear to me what is the relation. The error decreases with increasing plot size. When the plot is maximized on a 2560x1440 screen I get error less then 0.01 which is acceptable.
[3] In case of Y coordinate I had to add the node_height * 0.0009 component because the higher the node was the bigger the error was. I don't know where this number comes from but without it errors for high nodes were significantly bigger.

Questions

[1] Is there a better way read python (X, Y) coordinates of nodes?
[2] If there is no way to do it better could you tell if you plan to add a feature that would allow to access (X, Y) coordinates of nodes?
[3] Could you tell what are formulas used to transfrom python (X, Y) coordinates to parameters in the translate function used in the SVG code of a generated plot?