Changes in version 1.41.1                        

New Features

  - SummarizedExperiment Support: Added native, seamless support for
    SummarizedExperiment inputs (and lists containing them) in the main
    doppelgangR() function. The objects are automatically coerced to
    ExpressionSet internally using robust S4 coercion (as(x,
    "ExpressionSet")), enabling direct compatibility without manual type
    conversions.

Migration to future Parallel Framework

  - Replaced BiocParallel package usage with the CRAN future and
    future.apply framework for parallel computing.
  - Removed BPPARAM parameter from doppelgangR() to align with future
    package best practices (enabling global backend configuration via
    future::plan()).
  - Added a dummy parameter catcher in doppelgangR(...) that throws a
    helpful deprecation warning if legacy BPPARAM is used.

Performance and Memory Optimizations

  - Cartesian Product Vector Optimization: Added a vector fast path in
    .outer2df to construct coordinates directly using vector replication
    (rep), yielding a 46x speedup on 2,000 samples and eliminating
    massive memory spikes from legacy string splits.
  - Phenotype Distance Lookups: Precalculated value frequencies once per
    column in phenoDist(), allowing vectorWeightedDist() to perform
    $O(1)$ lookups instead of scanning columns dynamically. This reduced
    complexity from cubic to quadratic and achieved a 2.3x speedup
    on 500 samples.
  - Hamming Distance Helper Optimization: Replaced length(na.omit(z))
    with sum(!is.na(z)) inside vectorHammingDist() for an 8x speedup per
    call.

Bug Fixes & Compliance

  - Recycling Warning Fixed: Resolved a vector length mismatch recycling
    warning in doppelgangR() intermediate pruning when smoking guns are
    enabled. Replaced the element-wise logical OR on vectors of unequal
    sizes with a robust, key-based merging and subsetting strategy.
  - BiocCheck & Code Style Compliance:
      - Migrated unsafe sapply() calls to type-safe vapply().
      - Replaced manual integer ranges (1:...) with seq_len().
      - Removed paste() and paste0() from inside error and warning
        condition signals.
      - Replaced raw cat and print calls in trace logging blocks with
        message().

                        Changes in version 0.5.0                        

  - Add knn imputation by default.

                        Changes in version 0.1.1                        

  - remove "smoking gun" columns from calculation of phenotype
    similarity.

                        Changes in version 0.1.0                        

  - initial commit