Version 2.5.6

v2.5.6

• Merge CompatHelper: bump compat for CSV to 0.9
• Update Project/Manifest
• remove PI function and gfk function

Version 2.5.5

• CompatHelper: bump compat for "BenchmarkTools" to "1.0"
• Update Project/Manifest
• Fix not passing bounds for lambda1 and lambda2 to regularization solver
• Fix restrictive assumptions for diffusional broadening treatement
• Fix restrictive assumptions for tandem DMA transfer

Version 2.5.4

Merge CompatHelper: bump compat for Distributions to 0.25

Version 2.5.3

Merge CompatHelper: bump compat for DataFrames to 1.0

Version 2.5.2

Merge CompatHelper: bump compat for "BenchmarkTools" to "0.7"
Test for potential Julia 1.6 regressions (none found)

Version 2.5.1

Merge CompatHelper: bump compat for "BenchmarkTools" to "0.6"

Version 2.5

• Add gfₖ function for multicharge treatment of mobility distribution function in HTDMA
• Add TDMA model functions for matrix based TDMA data inversion
• Add memoization of slow functions (setupSMPS, setupDMA, setupSMPSdata, and initializeDefaultMatrices). This speeds up code when called with the same input parameters.
• Remove most global variables in dmafunctions; the globals resulted in sometimes inconsistent results when multiple DMAs were initialized in the same script.
• One global variable remains in the code for compatibility reasons, otherwise the code should be pure
• Memoization means that rinv(R, δ) has no longer a performance penalty when called multiple times with the same DMA δ. This makes the method rinv2(R) unnecessary and it is removed.
• Removal of globals required moving the function Ωₐᵥ into setupSMPS, due to a previous implicit dependence on a global. Reference to this function in the docs is removed.

Version 2.4

• Update documentation for rinv2 and add logo
• Merged CompatHelper: bump compat for "Interpolations" to "0.13
• Add rinv2 function based on RegularizationTools.jl. Intended to supersede rinv and supports higher order inversions and performance gains.
• Revisions related to the addition of rinv2
• Thinned out Project.toml and added Compat section.

Version 2.3

• Various small bugfixes discovered during documentation
• Add benchmark scripts to track performance across CPU architectures and over time.
• Add self-contained examples folder with code that matches the docs.
• Add docstrings and write Documenter.jl documentation for the package.
• Add links to Docker container with precompiles sys image.
• Add links to narrated tutorial notes to manual.
• Add interpolateSizeDistributionOntoδ function. This is match grids for chained DMA systems.
• Add recommendation for MKL library for additional performance gain
• Signficant speedup in inversion by switching to Choleskey decomposition for computing the inverse
• Revise search functions for performance gain.
• Limit number of BLAS threads to 1 for factor 4 speedup in inversion.

Version 2.2

• Add interpolateDataFrameOntoδ function. This is used to convert a measured size distribution stored as a DataFrame and to a SizeDistribution that matches a DMA grid. The function simplifies using the package with real data.
• Preallocate identity matrix
• Fix performance regression caused by changes in 2.1
• Add subtraction operator: 𝕟1 - 𝕟2
• Remove Plots.jl/PlotlyJS.jl/ORCA.jl dependencies for core project. The packages are still needed for executing the notebooks but they don't need to be installed when the package is embedded for regular use and carry a lot of overhead.