Compression for FrontEndObjects

[JerryI]

Should improve #340

Use it inside FrontEndObjects extension

CompressedExpressionJSON[“...”, "Base64"] (implement on both sides)
CompressedWXF["...", "Base64"] (implement on WL)

CreateFrontEndObject (if is too big) -> CompressedWXF[...] -> transport -> uncompress using MMA extension to JS array -> interpretate

When syncing happens the original CompressedWXF will be returned to the Kernel, which will also reduce the size of the notebook storage.

[JerryI]

Here is some graphs for the comparison

exportCompressed[expr_] := With[{arr = Normal[ExportByteArray[expr, "ExpressionJSON"] ]},
  With[{data = BaseEncode[ByteArray[Developer`RawCompress[arr](*`*) ] ]},
    uncompress[data] // Hold
  ]
];

exportNormal[expr_] := ExportByteArray[expr, "ExpressionJSON"]
measure[e_] := Quantity[Round[ByteCount[e] / 1024.0 / 1024.0, 0.001], "Megabytes"]

measureComplete[test_] := {
  test // measure,
  test // exportNormal // measure,
  test // exportCompressed // measure
}

{Head[#], measureComplete[#]} &/@ {
  Plot[x^2, {x,0,1}],
  Plot3D[x y, {x,-1,1}, {y,-1,1}],
  ExampleData[{"TestImage","Aerial"}],
  ExampleData[{"TestImage3D","CTengine"}],
  ExampleData[{"Audio","Apollo11ReturnSafely"}]
} //Transpose;

TableForm[Drop[%, 1][[1]], TableHeadings -> {%[[1]], {"WXF", "JSON", "Compressed JSON"}}]

We are currently using ExpressionJSON for frontend objects, since we do not have a fully functioning WXF parser written in JS. WXF is the closes format, which represents how WL stores the data in Kernel. This is a goal for the future.

However, if we still just keep using ExpressionJSON, but compress it using gzip, the results are amazing. Deflation on Javascript side becomes trivial.

[JerryI]

Real life example