mini patch for "race condition" bug with parallel make: https://r-forge.r-project.org/tracker/?func=detail&atid=294&aid=1723&group_id=61]

qr(<sparse>) --> <sparseQR> : finally address the structurally rank-deficient case where cs_sqr() produces "spurious" extra rows in Q i.e. V, and R; add checking code for that case

 as(<csr>, "dgCMatrix") works again; Matrix now *runs* in older R even if installed in R >= 2.15.2

an extra check needed for R-devel (2.16.0) on 32-bit Windows

use R_SVN_REVISION (instead of less precise R_VERSION)

using .External(Mmatrix, ...) instead of .Internal(matrix(...))

+ ByteCompile: yes {as it is 'recommended'}

.sparseDiagonal() & solve() tweaks (which were not really necessary, but should slightly speed up

toeplitz(<sparseVector>);  more use of  paste0 ()

allow writeMM("~/...")

allow solve(<matrix>) to keep list(NULL,NULL) dimnames

need name in 'Maintainer'

CRAN policy compatible maintainer

make arithmetic (and "Ops") considerably more efficient for large diagonal matrices {thanks to a report by Harold Doran}

Version 1.0-9 - for release

more Ops are checked and work; new is.finite() and is.infinite() methods (+ checks for them)

new pack(); more unpack() methods

update(<Cholesky>, non-symmetric) restored to previous behavior; more docu and tests

date (prepare "build")

More docs/examples on chol()/Cholesky()
 new chk.s() "check dots" utility

new  updown() [interface to CHOLMOD_updown] for up-/down-dating a Cholesky decomposition

PROTECT()++ thanks to Andrew Runnalls

version requiring  R >= 2.15.0

made sparse2int.R()  recursion-free which helps only little for sparse.model.matrix() high-interaction factors

fix solve(<dtCMatrix>) seg.fault by using Calloc() instead of Alloca() .. hmm

sparse.model.matrix(*, verbose) {and cleanup of old stuff}

<sparseMatrix> o <numeric> fixed Florent`s bug

plug memory leak (only in error case!) reported by Kasper Kristensen

mini cleanup before release

fix  -Diagonal(3)

we use one MASS data set in one vignette

Matrix_check_class_etc() has been in R-devel for a while now;
NAMESPACE imports for border line situation

skip other test in other-pkts.R; cosmetic

workaround some "CRAN check errors" (Mac Encoding / Solaris seg.fault with SparseM)

mention options("Matrix.quiet" = TRUE) for qr.R()

ready for CRAN release

cbind2() and rbind2() methods for the mix of sparse & dense

ready for 1.0-0

make Matrix_check_class_etc callable from other packages C code

bandSparse(*, giveCsparse=TRUE) - option to return Tsparse, ditto for sparseMatrix()

people should see the "expm" package when using expm()

export diagU2N() and diagN2U()

update before release

Version -> *-3

do not use GNU-make-isms in Makefile

update+comments

 "[<-" : now also replTmat calls into Csparse_subassign for large replacement region

<Csparse>[..] <- val   now goes via .Call()

finish: functionality of sparseVector.c now as "R_INLINE" in Mutils.h

finally a working C level Csparse_subassign() - to solve the memory-explode issue .. *BUT* it`s still too slow, even if I never remove but use drop0(.) in the end ... commit now before South America ... in case I get hit by a truck :-)

update the "UFsparse_download.sh" script; need new SparseM for example

use standard UFlong (instead of UFlong := int hack); updated CHOLMOD/SparseSuite Jan. -- version 0.9996875-0

expand(lu(<non-square dense>)) now works

fix bug in  M[i,j] <- v for cases with *repeated* entries in j

Remove Encoding: directive.

nearPD(): better error message for all negative E.V.

get rid of model.Matrix() {stub to pkg "MatrixModels"}

cosmetics from checkUsagePackage()

fix <ngCMatrix> %*% <dense> -- providing fast C-based coercions

Matrix(): adapted to new .Internal(matrix)

update before release

+ graph:: 

misc. (cosmetic) - preparing release

sparse.model.matrix(): drop.unused.levels = *FALSE* / printSpMatrix() using FormatSpMatrix() and that is modularized

Schur(<dgCMatrix>) no works

need to drop "@factors" when matrix is changed

efficient x[ -i ] for sparse vectors; "Enhances" MatrixModels

add a model.Matrix() "proxy" - for back-compatibility

cleanup Imports; no full, but just importsFrom()

new package "MatrixModels" split off Matrix - for modelMatrix(), prediction & response Modules, etc

move Matrix/ directory "down"

IRLS() tweaks; update( <glm4> ); "Model" mother class

IRLS(): more convergence crit. checking.  s/glm1/glm4/

"glpModel": docu etc; "CholeskyFactorization"

bug fixes for "!=", thanks to Ruth & Brian Ripley

nearPD() etc; see ChangeLog

fix mean(<sparse with NA>), thanks to Ruth Ripley

Create -40.1 for SVN only allowing a dependency in lme4a for testing

M [ <lMatrix> ]  <- value  now works

sparseMatrix(*, dimnames=.) should work (again!)

model.Matrix() tweak

new model.Matrix(), sparse.model.matrix() return "modelMatrix" class

rBind(<lsparse>, <lsparse>) now is <lsparse>

fix <sparseMatrix>[] <- value

update condest() help page + cosmetic

fac2sparse(): no longer use "Class" [= 1st arg of new()] explicitly

update CHOLMOD etc from Tim Davis -- using src/scripts/UFsparse_download.sh and CSparse_install.sh

lm.fit.sparse(x,y, w, *) (+ weights!) etc

enable sparse.model.matrix( ~ 1)

which() methods + checks; small things

det(): export regularly (with "masking warning") instead of base namespace hack

ore <sparse>[.,.] fixes, using cholmod_submatrix() now

fix <sparse>[0,0] -> version 0.999375-36

Fix <n.TMatrix>[ <ij-matrix> ] <- FALSE

add a PROTECT() [hint from A.Runnalls]

nnzero() is generic; det[erminant]() cosmetic

Matrix`s own contr.poly() had an infelicitous default

New version

Use Matrix-multiplication for sparse "[" for now

many "abIndex" methods and tests ++ see ChangeLog

fix *sTMatrix -> sparseVector; etc

<sparseMatrix> o <numeric> remains sparse in more cases

Add URL: key

implement missing sparseVector "Ops"

<sparseVector>  method improvements

add missing PROTECT()

checkMatrix() does crossprod() now; also see ChangeLog

revert cholmod_common access in *exported* cholmod error handler

release ready

solve(<CHMfactor>); more chol2inv(); more methods against "disambiguity" notes, notably "[" ones, including a buglet fix

sparseVector Ops; printing "non-structural FALSE" in logical sparseMatrix; more use of paste0()

determinant(<CHMfactor>) now does return  det( "L" )

update

sparse.model.matrix() rewrite and tests

export isLDL, test it; small fixes WRT contr.*(., sparse) 

regr.test for "keep LL"

New release

ready for release

remove extraneous slots in some *-class.Rd files

move %*%, [t]crossprod() methods to own file products.R; + "sparseVector" methods

just *r*ecommended, not *R*...

"fix" for the problem that .M.classEnv is not found when Matrix is only loaded but not attached

as(., "vector"); M[i, drop=<LOGI>] vector-indexing; fix (& check) rare problem in <TsparseTriangular[i,]; some cosmetics

lu(*, warnSing)

rid code of pre-2.9.0 if()s; also use 2.9.0-based tests/*.Rout.save where appropriate

more robust MatrixClass(); in C, use methods:::classEnv() via wrapper

use C-level Matrix_check_class_and_super(); adapt R code (and checks!) that work with class names similarly, using new MatrixClass()

band(x, .,.) fixes, notably for dense x

address remaining warnings in tests/Simple.R, see ChangeLog

misc.cleanup; mostly related to "all NA sparse mat."; see ChangeLog

<ngCMatrix>[, j] <- val  -- fix

ready to release *-20; do not call spqr() on Windows; minor coercion changes, notably for "pMatrix", see ChangeLog

long standing small tweaks; release prep

2nd steps towards an R - SPQR interface

typos in sparseMatrix.Rd; ..

for release

new sparseMatrix(): construct *sorted* Csparse from un-sorted input, using modify-in-place under the hood

New release

Prepare for new release

more Makevars & Makefile changes from Brian

ready for release again

Makefile changes for parallel make

Prepare for new release.  Update license details.

fix one more (minor) TODO

less TODOs; notably coercion -> [dln]Matrix; diagonal arithmetic

doc updates for diagonalMatrix

read for releasing (upcoming R-devel change)

drop0(., tol); <Matrix,ANY> changes from JMC; diagU2N()

all.equal()  TRUE only if e.g. classes match

all.equal() and Ops for nsparse(Matrix|Vector)

all.equal() methods; expand(<sparseLU>)

make diagonalMatrix extend sparseM* instead dense*; dgCMatrix_qrsol(.,., order) {in hidden lm.fit.sparse()}

two more solve() cases; tweaks related to diagMatrix

re-activate the Csparse_validate(x) feature of sorting x

fix longstanding dimNamesCheck() buglet

bdiag() improvements of last week

M[...] <- "sparseVector" etc, see ChangeLog

bdiag(): use Doug`s better algorithm, fix "bugs"

changes related to 2197; also one small step at reducing the ambiguity warnings

Ops methods for <diag> o <dMatrix>

also import "Ops" (as of May 7 !)

less TODO; more "Done"

checkMatrix() now checks the [dl]sparseMatrix -> nMatrix coercion

slight change in class-definition of [d]diMatrix; better pos.definite checks for dpo* and dpp*

more determinant(); MANY consistency cleanups, found by checkMatrix(); .. (see ChangeLog)

require R 2.6.0

move declaration such that old gcc does not complain

prepare release 0.999375-9

BunchKaufman(); fix <ddi> o <numeric> bug; Version++

no longer use "non-standard" chol() and qr() generic

several cosmetics

Prepare for new release with chm_factor_ldetL2, etc.

fix diagonal <-> sparse  buglets

ready for releasing 0.999375-5

replace many "callGeneric(U, V)" by direct <FUN>(U, V)

solve(<CHMfactor>, <numeric>) and related

fix rcond(..) infinite recursion

update 'License' to conform to Kurt's standards

some more dense solve() / lu()

make alloca working for Sun

add explicit ... to function args in certain setMethod("[" ..) calls  {for R-devel}

new nearPD() {from Jens` nearcor()}; some TODO cleanup

add more x

release date

norm() for sparse; systematic checking of norm() and all "Summary" methods

require 2.5.1

Prepare for new, bug-fix, release

merge 'branches/Matrix-APIchange' back into the trunk

printSpMatrix() {exported, extended, documented}

not dropping dimnames (notably of sparse matrices)

add session- and hardware-info to Comparisons

Math/Math2: setGeneric adaption for todays` R-devel

more strict validity checking for Csparse

few misc.

use correct "na.rm" default in all() and any(); tests for that

export drop0() {name is open for discussion}; document colSums

sparseVector. <- v; rowSums/colSums; summary(<sparse>); spMatrix(.) utility

part of sparseVector additions; in order to avoid merge conflicts

new version for CRAN - windows problem

fixed symmetric Tsparse indexing

Remove superfluous Lapack files

+ diff(.), + "dim<-";  cBind/rBind + cbind/rbind; indexing fixes - current state

Prepare for 0.9975-11 release

Add image method for CHMfactor class

update

workaround platform-dependent segfault in R_to_CMatrix

drop(); tweaks to Ops & indexing; Rsparse coercions for less exceptions in tests/Class+Meth.R

many more coercions for diag(), "!", etc; see ChangeLog

accomplished quite a few small FIXMEs; see ChangeLog

indexing etc till yesterday

move Ops to new Ops.R; start using "Logic"

more "[<-" fixes; validity of triangular sparse

fix last (? ha!) indexing bug; cbind2/rbind2

any(), all() and "Compare"(<sparse>) - for release of 0.9975-7

fix symmetric-sparse-indexing bug

export cholmod_analyze_p; new version

ready for release

0.9975-6 not -7

crossprod(<CsparseMatrix>,<CsparseMatrix>) added

another set of fixes/improvements; the last prompted by Harri Kiiveri

miscellaneous "fixes" mostly in ChangeLog and/or related to diagonalMatrix

update for release

small changes related to "corMatrix" and other newly seen classes in tests/Class+Meth.R

fix Matrix() %*% Diagonal() bug

Remove registration of write_HarwellBoeing and update version number

Matrix(0, d,d) should also give sparse (not diagonal); general_to_symmetric; notably for lsC*; more coercions; first cut at Csparse_drop()

New release to export more CCallable functions

more coercion methods (for rare matrices)  => less exceptions

"Compare" for "dMatrix"; plus a few small ones

Date ++

small update

Matrix <-> graph coercion update (for R 2.4.0)

Date

new nMatrix and subclasses; lsparse change to have x slot; many consequences

Update description after removal of Metis et al.

Merging changes from Matrix-nolmer branch

Clarify the sparse matrix libraries used

with the branch in place, up the Version and require R >= 2.4.0

Increment version number after last upload.

miscellaneous tweaks; moved some method definitions higher up

Date for release

several smallish fixes

add comment on why Rdefines.h

made Date: current

use speedier crossprod() for "dgC" - still slower than t(.) %*% .

library(Matrix) *outside* if; more graph tests

recreate the S4 Matrix examples at data load time

miscellaneous mostly related to sparse indexing

Wrap lines in the License section at a shorter line length.  Cite correct directory doc/UFsparse.

update date

non0ind() extended for symmetric sparse matrices -> correct printing

Preparing for new release in early July

mcmcsamp()[,"deviance"] was wrong column

coercion to and multiplication of diagonal and sparse matrices

Change the Depends:

useDynlib(*, .registration=TRUE)

single argument [t]crossprod now returns a symmetric matrix

tT -> tC; and diagU2N() for solve(<dtC>)

coercions from dgC to triangular (dtC) and symmetric (dsC), etc

modifying return value of ranef method and adding residuals method

Fix dpotrf call for BLAS under Mac OS X

put "mm" and "y" datasets into list "KNex" and change the docu, examples, tests, and one vignette accordingly;
 ready for release of 0.995-6

sparse: "[<-" and better arithmetic; other tweaks

Fix case of cbind(successes, failures) for lmer fitting a GLMM (see Changelog)

Date++

comment in simple.R; ready for 0.995-4 release

Class cleanup; particularly ldense* ones

Version 0.995-3

Version ++

Date++

Munging the data frame

Really only need R-2.2.0 (for nlminb)

Preparation for new version

0.995 ; util; R >= 2.3.0

add ", drop = FALSE" in lmer.coef constructions; plus a few comments

Release 0.99-6

Fix due to Peter Dalgaard for segfault with multiple non-nested grouping factors; Version 0.99-5

"diagonalMatrix" class and subclasses with a few methods; very basic "ltTMatrix" methods

depend on "utils"; new version for ready for CRAN

update

new Version (needed for build)

Release 0.99-2

SystemRequirements: GNU make

Release 0.99-2

New release

Remove references to mer2 and mltloc.rda which are now in a branch

mer2_initial

Added release information

too many changes: rbind2 for dM*, cbind2(), "Math", "Arith" for sparse ones; more  "s/gTMatrix/TsparseMatrix/" fixes; quite a few more tests, incl some for *all* classes

date

cbind() / cbind2() methods for R >= 2.2.0

up "Version" (for those using svn version)

Version 0.98-6: "[" methods for dense matrices; first "[<-"; examples

Release 0.98-5

Release 0.98-4

Release 0.98-3

Harwell-Boeing and MatrixMarket formats

Version 0.98-1

Release 0.97-4 with fix suggested by DTL

Release 0.97-3

Preliminary version of glmmMCMC

Updates for release 0.97-2

New release

Release 0.96-3

release 0.96-2

Prepare for release 0.96-1

Moving lmer to the Matrix package

New release

Release 0.95-11

Release 0.95-10

Preparing for release 0.95-10

Release 0.95-9

Release 0.95-8 with bug fix for sparse symmetric solve

Getting ready for a release

Moving Doxyfile

Release 0.95-6

"[" and show() methods for sparse matrices => RFC

Repaired a thinko in the lmer_fitted function

Prepare for release 0.95-4

Preparing for 0.95-4 release

dgCMatrix: needs explicit "dim" method (should not be needed!)

dimnames(), "dimnames() <-", better t(), etc

more group methods and testing them

Release 0.95-2

Release 0.95-1

update Version (just in case)

some matrix exponential tests

Getting package to pass CMD check

Preparation for new release

New release

Release 0.8-23

Change to the lmer representation

Working on the lmer classes

New release

New release

Updates for factorization name and default value

New release

Add registration of .Call functions

Release 0.8-15

check for lattice package in image methods

Prepare for new release

New version

Preparing for a new release

Remove taucs, add lmeRep, new release

Add the lmeRep class

Remove warnings, new release, change gradient calculation method

Remove pdMat classes

Ready for next release

Prepare for next release

Doxygen configuration file

Change release number

New release

Release 0.8-4

New release

Release 0.8-3 to fix Makefile.win

Release 0.8-2

Initial import