Developer log

This document will house any development logs that were not part of a structured issue or pull request

Array sparsity (tested in v0.12.0)

• array sparsity was tested with scipy.sparse csc_matrix for column-level sparsity storage, with the hypothesis that column storage would generally be more efficient than row storage (most ROI dimensions have more columns than rows)
• sparsity was tested in order to potentially save disk space when creating cached Serverside objects from callback invocation: sparsity was used only for the grey scale raw matrices and not for the RGB blend layers
• Result: sparsity did not improve disk storage for the pickle hash tables of raw arrays, and in fact sometimes increased the amount of disk space required. Additionally, using a function to constantly convert the sparse arrays back to dense made the blend operations slightly slower
• decision: maintain arrays in standard dense format without any conversions between sparse and dense for any of the blend steps. With the addition of the key input for Serverside objects for overwriting previous pickle objects, sessions can avoid continuously using large amount of tmp storage per session