Changes in version 1.22                         

    o   Improved the efficiency of the downsampling algorithm in
	downsampleMatrix(), which will change the results slightly. In
	addition, new options are added to control the output type
	(integer/double) and class (dense matrix or SVT_SparseMatrix).

    o   Removed "median" from the default statistics= in
	summarizeAssayByGroup(). This enables default use of the fast
	code path for improved performance.

    o   Deprecate all functions that have more efficient counterparts
	in other packages, mostly scrapper. Specifically:
	
	  • addPerCellQcMetrics() is replaced by
	    scrapper::quickRnaQc.se().
	
	  • aggregateAcrossCells() is replaced by
	    scrapper::aggregateAcrossCells.se().
	
	  • aggregateAcrossFeatures() is replaced by
	    scrapper::aggregateAcrossFeatures.se().
	
	  • cleanSizeFactors() is replaced by
	    scrapper::sanitizeSizeFactors().
	
	  • geometricSizeFactors() is replaced by
	    scrapper::computeClrm1Factors().
	
	  • librarySizeFactors() is replaced by
	    scrapper::centerSizeFactors().
	
	  • logNormCounts() is replaced by
	    scrapper::normalizeRnaCounts.se().
	
	  • normalizeCounts() is replaced by
	    scrapper::normalizeCounts().
	
	  • numDetectedAcrossCells() is replaced by
	    scrapper::computeRnaQcMetrics().
	
	  • perCellQCFilters() is replaced by
	    scrapper::suggestRnaQcThresholds().
	
	  • perCellQCMetrics() is replaced by
	    scrapper::computeRnaQcMetrics().
	
	  • quickPerCellQC() is replaced by scrapper::quickRnaQc.se().
	
	  • sumCountsAcrossCells() is replaced by
	    scrapper::aggregateAcrossCells().
	
	  • sumCountsAcrossFeatures() is replaced by
	    scrapper::aggregateAcrossFeatures().
	
	  • summarizeAssayByGroup() is replaced by
	    scrapper::aggregateAcrossCells() or
	    beachmat::tatami.sums.by.group().
	
	  • uniquifyDataFrameByGroup() is replaced by
	    scrapper::aggregateColData().
	
	  • readSparseCounts() is replaced by
	    SparseArray::readSparseCSV().

                        Changes in version 1.16                         

    o   addPerCellQCMetrics() will (optionally) add the feature subset
	identities to the row data of the SingleCellExperiment.

    o   Added uniquifyDataFrameByGroup() to collapse grouped rows in a
	DataFrame into their unique values.

                        Changes in version 1.10                         

    o   .subset2index now converts factor inputs to character vectors,
	rather than treating them as integers.

                        Changes in version 1.8.0                        

    o   Removed support for use.altexps= in aggregateAcrossCells() and
	logNormCounts().

    o   Added swap.rownames= option to makePerCellDF() to allow easy
	access by rowData aliases. Also moved the extracted features to
	the end of the data frame for consistency.

                        Changes in version 1.2.0                        

    o   Migrated whichNonZero() to beachmat.

    o   Bugfixes for factor-based colData aggregation in
	aggregateAcrossCells(). Added proper support for Vectors.

    o   Bugfix for correct response to use.altexps= in
	perCellQCMetrics(), perFeatureQCMetrics().

    o   Added a normalize.all= option to normalizeCounts(). Removed
	unnecessary warning when down.target= is not specified. Exposed
	the default size.factors= in the SingleCellExperiment method.

    o   Modified the SingleCellExperiment method of logNormCounts() so
	that manually specified size factors do not apply to
	alternative Experiments. Only relevant if size.factors= and
	use.altexps= are specified.

    o   Deprecated use.altexps= in favor of applySCE() in
	logNormCounts() and aggregateAcrossCells().

    o   Renamed addPerCellQC() and addPerFeatureQC() to
	addPerCellQCMetrics() and addPerCellFeatureMetrics(), for
	consistency. Soft-deprecated the old functions.

    o   Moved most of quickPerCellQC() functionality into the new
	perCellQCFilters() function. Repurposed the former to directly
	return a filtered SummarizedExperiment object.

    o   Migrated scran's normalization-related functions into this
	package. Added pooledSizeFactors(), computePooledFactors(),
	cleanSizeFactors() and computeSpikeFactors().

    o   Added transform="asinh" to normalizeCounts() and
	logNormCounts() for inverse hyperbolic transformations of
	CITE-seq data.

    o   Modified isOutlier() to now return outlier.filter objects.
	These are simply logical vectors that preseve the "thresholds"
	attribute upon subsetting.

    o   Migrated correctGroupSummary() from scater, to compute
	corrected versions of group-level summary statistics.

                        Changes in version 1.0.0                        

    o   Split off scuttle from scater by migrating all
	non-visualization code from the latter.

    o   Began transition to dot-separate argument names from original
	snake case format.

    o   Added a geometricSizeFactors() function, deprecated
	geometric=TRUE in librarySizeFactors().

    o   Single-object downsampling in downsampleBatches() now behaves
	more consistently with multi-object downsampling.