JuliaData/CategoricalArrays.jl: v1.0.0

CategoricalArrays v1.0.0

Diff since v0.10.8

Breaking changes

• unique(::CategoricalArray) and levels(::CategoricalArray) return a CategoricalArray instead of unwrapping values, consistent with unique(::AbstractArray) in Base and levels(::AbstractArray) in DataAPI (#358, #425).

• cut always closes the last interval on the right (#409).

• cut(x, breaks) rounds breaks to generate shorter labels (#422).

• cut(x, ngroups) takes breaks from actual values instead of using quantile estimates which are generally longer (#416). This only changes group labels, not their contents.

• T(::CategoricalArray{U}) and convert(T, ::CategoricalArray{U}) now consistently return an Array{U} for T in Array, Vector, Matrix. This avoids creating Array{<:CategoricalValue} objects unless explicitly requested (#420).

• All deprecations have been removed (#419).

New features

• Support reading from and writing to Arrow files (#415).

• Improve performance of recode (#407).

• Support weighted quantiles in cut (#423).

Bug fixes

• Fix performance regression on Julia 1.11 and above (#418).

• Fix cut corner cases with duplicated breaks (#410).

Merged pull requests:

• fix unique() behaviour, add unique!() (#358) (@alyst)
• make recode! type stable (#407) (@tiemvanderdeure)
• remove JET runtime dispatch error (#408) (@matthijscox-asml)
• Make cut close last interval on the right (#409) (@nalimilan)
• Fix corner cases of cut with duplicated breaks (#410) (@nalimilan)
• CompatHelper: add new compat entry for Statistics at version 1, (keep existing compat) (#411) (@github-actions[bot])
• Bump julia-actions/setup-julia from 1 to 2 (#413) (@dependabot[bot])
• Bump actions/cache from 2 to 4 (#414) (@dependabot[bot])
• Support reading from and writing to Arrow files (#415) (@nalimilan)
• Choose different quantile cutpoints in cut(x, n) (#416) (@nalimilan)
• Fix allocations by dropping CategoricalPool type parameter (#418) (@nalimilan)
• Remove deprecations (#419) (@nalimilan)
• Add Array constructors and convert methods (#420) (@nalimilan)
• Simplify default cut labels (#422) (@nalimilan)
• Support weighted quantiles in cut (#423) (@nalimilan)
• Fix CI (#424) (@andreasnoack)
• Make levels return a CategoricalArray (#425) (@nalimilan)
• Release version 1.0 (#426) (@nalimilan)

Closed issues:

• Fix showcompact() (#40)
• Move CategoricalValue and CategoricalPool into separate package (#64)
• Implement optimized == and isequal() for arrays with different pools (#82)
• == for CategoricalPool (#108)
• Implement deleteat! method (#115)
• Adding levels to ordered categorical array (#141)
• cut combined with labels and extend throws ArgumentError (#160)
• missing append!(::CategoricalArray, ::SubArray{<:CategoricalArray}) (#170)
• maximum/minimum over dimension fails (#178)
• Extreme slowdown in IndexedTable creation with CategoricalArrays (#272)
• Missing CategoricalArray{String}(::SentinelArrays.ChainedVector, ::CategoricalPool) (#285)
• Add Vector conversion (#294)
• Can't precompile on Julia v1.5.2 (#304)
• another take at cut (#314)
• How to check if an array contains categorical values? (#332)
• CategoricalArray creation (#333)
• Plotting categorical values as colors (#351)
• unique() returns levels, not the CategoricalArray (#357)
• How to compare to value? (#363)
• Make last interval closed in cut (#373)
• Compact printing in cut (#381)
• Strange cut for imbalanced distributions (#382)
• Add promotion rules for CategoricalArrays and views of CategoricalArrays (#386)
• Consider adding RecipesBase as a direct dependency (#388)
• Make levels return a CategoricalArray (#390)
• ordered vs ordinal (#398)
• Define value(x) alternative to DataAPI.unwrap(x) (#399)
• docstring/repl issue in package development (#403)
• CategoricalValue allocates on julia 1.11 (#412)
• Request to become an active maintainer (#417)