Description
https://github.com/plotly/plotly.js/releases/tag/v2.28.0 introduces an option to set objects for encoded typedarrays for basically all fields that provide data to traces (e.g. x
and y
for a cartesian plot).
an object with keys dtype, bdata, and optionally shape. In this 3rd form, dtype is one of f8, f4. i4, u4, i2, u2, i1, u1 or u1c for Uint8ClampedArray. In addition to shorthand
dtype
above one could also use the following forms: float64, float32, int32, uint32, int16, uint16, int8, uint8 or uint8c for Uint8ClampedArray.bdata
is either a base64-encoded string or the ArrayBuffer of an integer or float typed array. For either multi-dimensional arrays you must also provide its dimensions separated by comma viashape
. For example usingdtype
: f4 andshape
: 5,100 you can declare a 2-D array that has 5 rows and 100 columns containing float32 values i.e. 4 bits per value.shape
is optional for one dimensional arrays.
Plotly.NET's API is statically typed and currently uses sequences of #IConvertible
for these fields. This was a clever trick because this allows setting any of the CLR runtime types Boolean
, SByte
, Byte
, Int16
, UInt16
, Int32
, UInt32
, Int64
, UInt64
, Single
, Double
, Decimal
, DateTime
, Char
, and String
while also making it very easy to serialize.
The new way of setting data_array
s seems to have speed advantages, and is therefore worth to be supported well from our side. A straightforward way of implementing this in the API providing a new type, e.g.
type EncodedTypedArray {
bdata: <string or maybe byte []?>
dtype: <some type abstraction>
shape: <some shape abstraction>
}
this type would also need functionality to encode the base64 string in a way that is understood by plotly.js. Maybe the python implementation's codebase can help here, as they seem to implement these functions as well.
Due to these large changes that also must be reflected on the CSharp lib, support for plotly.js >= v2.28.0 will be moved to the major release of Plotly.NET 6.0
Activity
kMutagene commentedon Feb 5, 2024
Alternative: implement a really basic version of this first that offloads base64 encoding, shape, and datatype selection purely to the user and just provide the
EncodedTypedArray
type and overloads for the API taking these as alternative arguments for data.This way, it would be possible to add this to v5 or a minor release of v5.