Changes in version 1.17                         

    o   Fix handling of missing values in `rowSums2` if `cols` is a boolean vector

    o   Implement optimized code path for `rowSums2` and `rowMeans2` if `cols` is provided

                        Changes in version 1.15                         

    o   Throw error if length of 'center' in colVars, colSds, colMads (and
	corrresponding row-functions) does not match size of 'x'.
	This behavior was never supported by `sparseMatrixStats` and previously
	could return incorrect results. This change matches the upcoming behavior of
	matrixStats (https://github.com/HenrikBengtsson/matrixStats/issues/254).

                        Changes in version 1.13                         

    o   Make sparseMatrixStats compatible with matrixStats release v1.0.0.
	In particular change 'useNames'to 'TRUE' by default.

    o   Add fast path for 'rowSums2(x, cols = logical_vector)'

    o   Add useNames parameter to all functions

    o   fix incomplete method signature of rowQuantiles

                         Changes in version 1.3                         

    o   Avoid C++14 features. This should fix the reoccuring installation problems.

    o   Support center argument in colVars (and friends). This can speed up
	the calculation if both are needed. Note that center must be a proper
	estimate of the column means.

    o   Fix bug that broke test on ARM architectures

    o   Add extensive section on how to fix installation problems to README

    o   Fix compatibility with matrixStats v.0.58.0

    o   Fix bug in colCounts, colAnys, colAlls, colRanks if the sparse matrix
	contained explicit zeros.

                         Changes in version 1.2                         

    o   Use numerically stable algorithms for sum, mean, and variance

    o   Implement dedicated function for rowSums2, rowMeans2, rowVars
	which avoid transposing the input matrix to call the colXXX version

    o   Support logical input (lgCMatrix) for all methods that matrixStats
	supports. Introduce new class xgCMatrix as union class of dgCMatrix
	and lgCMatrix.

    o   Fix bug in colWeightedVars() that returned slightly off results on
	Linux, because there abs() internally returns integers instead of
	doubles

    o   Change argument name in colRanks from `preserve.shape` to
	`preserveShape` to match matrixStats (thanks @LTLA for reporting)

    o   Fix bug in low-level `quantile_sparse()` related to +- Inf
	 (thanks @LTLA for reporting)

    o   Move MatrixGenerics from Imports to Depends so that loading
	matrixStats does not break calls to sparseMatrixStats
	(thanks @hpages for reporting)

    o   Fix bug in [col|row]Quantile which ignored drop argument if
	 length(probs) == 1

                 Changes in version 1.0.5 (2020-05-17)                  

    o   Fix links in documentation to get rid of WARNINGS

                 Changes in version 1.0.4 (2020-05-17)                  

    o   Fix bugs in colTabulates

    o   Update documentation to avoid warnings in build on Windows

                 Changes in version 1.0.3 (2020-05-17)                  

    o   Fix bug in colAnys and colAlls if value = TRUE

                 Changes in version 1.0.2 (2020-05-10)                  

    o   Fix bug in colMaxs, colMins related to missing values

                 Changes in version 1.0.1 (2020-05-08)                  

    o   Fix bugs in colMaxs, colMins, colAnys

    o   Fix bug in colLogSumExps

                 Changes in version 0.0.99 (2020-04-03)                 

    o   Submitted to Bioconductor