SCM Repository

[matrix] View of /pkg/Matrix/ChangeLog
ViewVC logotype

View of /pkg/Matrix/ChangeLog

Parent Directory Parent Directory | Revision Log Revision Log

Revision 2586 - (download) (annotate)
Sun Jul 25 02:32:06 2010 UTC (9 years, 3 months ago) by mmaechler
File size: 152471 byte(s)
move Matrix/ directory "down"
2010-07-23  Martin Maechler  <>

	* R/Auxiliaries.R (prMatrix): add " (unitriangular)" as we already
	have for sparse matrices.

2010-07-22  Martin Maechler  <>

	* R/Auxiliaries.R (.diagU2N): implement for "dtpMatrix" = old 'FIXME';
	(.dense.diagU2N): new utility, called from .diagU2N()

2010-07-19  Martin Maechler  <>

	* src/dtrMatrix.c (dtrMatrix_dtrMatrix_mm): new for tri %*% tri
	* R/products.R (%*%): ditto
	* tests/matprod.R: test it

2010-07-16  Martin Maechler  <>

	* R/spModels.R (do.defaults): add 'nonMatched.action' with default
	ensuring that typos are caught.

2010-07-16  Douglas Bates <>

	* R/spModels.R (do.defaults): utility function; TODO: move to R

2010-07-16  Martin Maechler  <>

	* DESCRIPTION (Version): 0.999375-43

	* R/AllClass.R (Model): as mother class (of "glpModel")

	* R/spModels.R (IRLS): more options()
	(updateModel): update(<glm4>)

2010-07-13  Martin Maechler  <>

	* DESCRIPTION (Version): 0.999375-42, CRAN-released: 2010-07-15, r 2566

	* R/spModels.R (glm4, IRLS): glm4 [was 'glm1']; tweaks.

2010-07-12  Martin Maechler  <>

	* NAMESPACE: rename, export and
	* man/glpModel-class.Rd: document Doug's new "glpModel" class.

2010-07-08  Douglas Bates <>

	* R/AllClass.R: new "lpMod" class (-> later =: "glpModel"), and
	working function:
	* R/spModels.R (glm1): using linear pred.Model class, and
	Bates-Watts convergence criterion.

2010-07-06  Martin Maechler  <>

	* R/lMatrix.R (whichDense): use arrayInd()
	* R/zzz.R (arrayInd): provide for older R versions

2010-07-05  Martin Maechler  <>

	* src/chm_common.c (chm_triplet_to_SEXP): deal more carefully with
	NAs, needed e.g., on Solaris; thanks to Ruth and Brian Ripley.

	* R/Ops.R (Compare <Csparse>): fix bug uncovered by "alongside" the

2010-07-02  Martin Maechler  <>

	* R/sparseMatrix.R (x[] <- 0): fix shortcut code.
	* tests/indexing.R: and test

	* R/nearPD.R (nearPD): 'doDykstra = TRUE' and *do* use Dykstra's
	correction which was *not* used in Jens' code; thanks to Bernhard
	Spangl for a report

2010-06-26  Martin Maechler  <>

	* R/Matrix.R: fix mean() method; add sparse one, remaining sparse
	at least for trim = 0.

2010-06-08  Martin Maechler  <>

	* DESCRIPTION (Version): 0.999375-41, CRAN-released: 2010-07-03, r 2555

	* R/sparseVector.R (spV2M): enable sparseVector -> (sparse)Matrix
	as(*,.) coercion.

2010-06-07  Martin Maechler  <>

	* R/Tsparse.R (.TM.repl.i.mat): renamed from .TM.repl.i.2col().
	* R/Matrix.R (.repl.i.lSMat): implement logical sparse
	sub-assignment: M[<lsparseMatrix>] <- v; ditto for dense & nsparse.

	* R/Csparse.R (.CM.repl.i.lSMat, ..): direct logical sparse
	sub-assignment for "Csparse".

2010-06-04  Martin Maechler  <>

	* R/sparseMatrix.R (sparseMatrix): re-enable 'dimnames' argument.

2010-06-03  Martin Maechler  <>

	* R/spModels.R (model.Matrix): tweak for NULL contrasts in dense case.

2010-06-02  Martin Maechler  <>

	* tests/spModel.matrix.R (Sparse.model.matrix): adapt to the fact, that
	sparse.model.matrix() returns not just a dgCMatrix.

2010-05-29  Martin Maechler  <>

	* DESCRIPTION (Version): 0.999375-40, CRAN-released: 2010-06-04, r 2546

	* R/AllClass.R: new classes "ModelMatrix", "sparseModelMatrix", etc.

	* R/spModels.R (sparse.model.matrix): now return
	"dsparseModelMatrix" object, notably with 'assign' slot.
	* R/spModels.R (model.spmatrix): faster, using lower level
	cbind2/rbind2 directly.

	* R/spModels.R (model.Matrix): new function,
	returning "ddenseModelMatrix".

	* NAMESPACE: export new classes.

2010-05-18  Martin Maechler  <>

	* src/Csparse.c (Csparse_horzcat, Csparse_vertcat): ensure that
	rBind()/cBind() i.e., rbind2()/cbind2() return logical sparse
	matrices when the components are.
	* tests/bind.R: test

2010-05-15  Martin Maechler  <>

	* R/sparseMatrix.R: A[] <- v ; differentiate dense & sparse
	* R/pMatrix.R: disallow <pMatrix>[] <- v  more consequently

	* tests/indexing.R: test above

2010-05-08  Martin Maechler  <>

	* R/spModels.R (model.spmatrix): deal with "AsIs" components
	* tests/spModel.matrix.R: test that

2010-05-01  Martin Maechler  <>

	* R/condest.R (onenormest, condest): allow to estimate condition
	number for large sparse matrices.
	* condest.Rd: docu

	* R/pMatrix.R (.inv.perm): utility; add [t]crossprod() methods

	* man/sparseLU-class.Rd: A = P'LUQ; add examples, with "identities"

	* R/Auxiliaries.R (mmultCheck): new arg. 'kind'

2010-04-28  Martin Maechler  <>

	* DESCRIPTION (Version): 0.999375-39, CRAN-released: 2010-05-19, r 2540

	* R/spModels.R (fac2sparse): using names(formals(new))[[1]] to
	adapt to a future change in new()'s first argument *name*.

2010-03-31  Martin Maechler  <>

	* R/spModels.R ( update, allowing weights; also
	return residuals, notably for "cholesky" case.

	* man/ examples; comparing with dense case.

	* src/dgCMatrix.c (dgCMatrix_cholsol): comments; also compute

2010-03-30  Martin Maechler  <>

	* R/spModels.R (sparse.model.matrix, model.spmatrix): border case
	'~ 1' should also work.  Add  'transpose = FALSE' argument.

	* tests/spModel.matrix.R: test that.

2010-03-27  Martin Maechler  <>

	* R/sparseMatrix.R (printSpMatrix): ensure returning original argument

2010-03-26  Martin Maechler  <>

	* R/sparseVector.R (coercion from TsparseMatrix): diagU2N() when

	* inst/test-tools.R (checkMatrix): explicit which() test for
	"l" and "nMatrix".  New sparseVector (coercion "and back") check.

2010-03-25  Martin Maechler  <>

	* R/lMatrix.R (which): define methods for which(<[ln]Matrix>).

	* inst/test-tools.R (Q.eq.symmpart): new utility, now called in

	* R/nearPD.R: use symmpart() for non-symmetric x
	* man/nearPD.Rd: improve title

2010-03-24  Martin Maechler  <>

	* R/colSums.R (.diag.Mean): define methods for "diagonalMatrix"

	* src/Mutils.c (m_encodeInd, do_ii_FILL):
	coerce ij to integer if necessary; check that ij are within "dim[]"
	values.  Parametrize do_ii_FILL() to be used in m_encodeInd2() as well:
	* src/Mutils.c (m_encodeInd2): also check bounds (if desired).

	* tests/indexing.R: test the above.

2010-03-19  Martin Maechler  <>

	* src/dgeMatrix.c (dgeMatrix_solve): compute the recip.cond.number
	and also bail out for *computational* singularity {as "base R"
	does}, from (code) suggestion by Daniel Sabanés Bové.

	* tests/dg_Matrix.R: "test" the above.

2010-03-01  Martin Maechler  <>

	* man/rep2abI.Rd: rep2abI() utility is exported now.

	* R/Csparse.R (subCsp_cols, subCsp_rows, subCsp_ij): dimnames() <-
	fix for character subsetting.

	* tests/indexing.R: testing it.

2010-02-26  Martin Maechler  <>

	* R/spModels.R (model.spmatrix): warn and coerce to sparseMatrix if
	result would end up dense (e.g., in case of *no* factors in formula).

	* tests/spModel.matrix.R: test the above.

2010-02-12  Martin Maechler  <>

	* R/dtrMatrix.R: add solve(<dtrMatrix>, <dMatrix>): e.g., for
	 solve(<Cholesky>, <dgCMatrix>) in lme4.

2010-02-09  Martin Maechler  <>

	* DESCRIPTION (Version): 0.999375-38, CRAN-released: 2010-03-31, r 2529

	* NAMESPACE, R/AllGeneric.R, R/zzz.R: change det() into a regularly exported
	function (masking base::det) instead of load-time hack.
	* man/Matrix-class.Rd: \alias, docu

2010-02-05  Martin Maechler  <>

	* DESCRIPTION (Version): 0.999375-37, CRAN-released: 2010-02-05

	* inst/test-tools.R (Qidentical.DN): new
	(Qidentical): all Qidentical.DN(<dimnames>)

	* R/Csparse.R (subCsp_ij, subCsp_cols, subCsp_rows): use
	CHOLMOD's submatrix C code, instead of matrix multiplication;
	now *do* keep dimnames of result, wherever classical matrix
	subsetting does.

2010-02-04  Martin Maechler  <>

	* DESCRIPTION (Version): 0.999375-36, CRAN-released: 2010-02-04

	* R/Csparse.R (subCsp_ij): Fix <sparse>[0,0] bug

2010-02-03  Martin Maechler  <>

	* R/Tsparse.R (.TM.repl.i.2col): <nsparse>[ <ij> ] <- FALSE fix
	* tests/indexing.R, * test that

2010-01-28  Martin Maechler  <>

	* src/Csparse.c (Csparse_crossprod): PROTECT() Tsparse_diagU2N()
	result, from a suggestion by Andrew Runnalls.

2010-01-22  Martin Maechler  <>

	* R/SparseM-conv.R (setAs(., "")): fix typo in method.

2010-01-20  Martin Maechler  <>

	* R/AllGeneric.R: nnzero() is now generic,
	* R/nnzero.R: newly containing all nnzero() methods.

	* R/zzz.R (det): assign base::det such that it uses S4 generic kronecker.

2010-01-18  Martin Maechler  <>

	* R/spModels.R (contr.poly): [the back-compatible-only version]:
	do not use a default for 'scores'; rather rely on stats::contr.poly.

	* tests/spModel.matrix.R: test that case

2009-12-28  Douglas Bates <>

	* DESCRIPTION (Version): 0.999375-35, CRAN-released: 2010-02-03

	* src/init.c, inst/include/Matrix_stubs.c: cholmod_band_inplace()

2009-12-23  Martin Maechler  <>

	* tests/indexing.R: slightly adapt to the very slight [] changes.
	* inst/test-tools.R (Q.C.identical): + checkClass

	* R/Tsparse.R ([-methods): for now go via "Csparse" and drop
	all the sophisticated code dealing with the many cases.

	* R/Csparse.R (subCsp_cols, etc): faster [i,j] via matrix
	multiplication, thanks to suggestions by Greg Jorstad.

	* R/Auxiliaries.R (paste0): more use of paste0()

2009-12-22  Martin Maechler  <>

	* R/diagMatrix.R (.sparseDiagonal): made more general, allowing
	to directly build the equivalent of   Diagonal(n)[, j]

	* man/Diagonal.Rd: document .sparseDiagonal() as well.

2009-12-21  Martin Maechler  <>

	* R/AllClass.R: abIndex@x and rleDiff@first are now "numLike",
	such that also logical can be converted to "abIndex".

	* R/abIndex.R (.diff): new utility, used in num2abI()

2009-12-19  Martin Maechler  <>

	* src/abIndex.c: include new
	* src/t_Matrix_rle.c (Matrix_RLE_): is template for
	Matrix_rle_i() and Matrix_rle_d(); now obeys a 'force' argument.

	* R/abIndex.R: implement <Arith> methods, at least with scalars.
	(all.equal.abI): add also all.equal() methods.

	* tests/abIndex-tsts.R: testing <Arith>, using all.equal().

	* R/AllClass.R: classUnion  "numLike" := {"numeric", "logical"}

2009-12-18  Martin Maechler  <>

	* src/abIndex.c (Matrix_int_rle): UNPROTECT() needed in trivial case.

	* R/abIndex.R (abIseq1, abIseq): new functions for building
	"abIndex" vectors.

	* tests/abIndex-tsts.R (tst.c.abI): test new c("<abIndex>") method.

	* DESCRIPTION (Version): 0.999375-34, CRAN-released: -never-

	* R/Ops.R: use prod(d) instead of d[1]*d[2], as the latter may
	integer overflow; fixes  <large sparseMatrix>  o  <sparseVector>

	* tests/Simple.R: test that.

2009-12-11  Martin Maechler  <>

	* R/sparseVector.R (TsparseM* -> sparseV): symmetricMatrix needs to
	be expanded.  This fixes a bug reported by Yu-Sung Su.
	* tests/indexing.R: testing the fix.

	* inst/test-tools.R (all.equalX): new util

2009-12-09  Martin Maechler  <>

	* R/Ops.R (A.n.M, A.n.M): <sparseMatrix> o <numeric> : remain
	sparse also when *majority* (instead of all) of   0 o v  is 0.

	* tests/group-methods.R: test one such case.

2009-12-06  Martin Maechler  <>

	* DESCRIPTION (Version): 0.999375-33, CRAN-released: 2009-12-11

	* R/Ops.R (Compare<Csp><Csp>): fix case with NA x.

	* R/not.R: fix "typo" in !<sparseVector>

	* R/Ops.R (Ops.spV.spV): fix thinko

2009-12-05  Martin Maechler  <>

	* R/sparseVector.R: setAs(nsparseV*, lsparseV*) etc

	* R/Ops.R (Ops.spM.spV, Ops.spV.spM): sparseVec. o sparseMat.

2009-11-20  Martin Maechler  <>

	* R/Ops.R (Ops.spV.spV): enable sparseVector operations in more cases.

	* R/sparseVector.R ( methods defined.
	* R/sparseVector.R (intIv): also accept "sparseVector"s

	* tests/Simple.R: check the above

2009-11-19  Martin Maechler  <>

	* R/sparseVector.R (newSpV, newSpVec): new utility, dropping 0 "on the fly".
	* R/sparseVector.R (atomic -> sparse*): fix for NA case.

	* R/Ops.R (<spV>): using newSpVec()

	* R/not.R: fix thinko in !<sparseVector>

2009-11-17  Martin Maechler  <>

	* tests/other-pkgs.R: detach(*, unload) Rgraphviz too

2009-11-14  Martin Maechler  <>

	* R/AllClass.R: "abIndex" (and "rleDiff") class
	* R/abIndex.R: with some methods; commit these finally, even if
	it's mostly unfinished.
	* src/abIndex.[ch]: new: currently only for .Call(Matrix_int_rle,*)
	* tests/abIndex-tsts.R: basic consistency checks for "abIndex".

	* R/diagMatrix.R (diagOdiag): "exploding" Matrix.msg() only level 2;
	* tests/ update

2009-11-11  Martin Maechler  <>

	* DESCRIPTION (Version): 0.999375-32, CRAN-released: 2009-11-20

	* src/Csparse.c (Csparse_Csparse_prod, Csparse_Csparse_crossprod):
	PROTECT(.) the dimnames; thanks to Kaspar Kristensen

2009-10-24  Martin Maechler  <>

	* R/Ops.R (Logic.lCMat): to be used for lsC* and ltC* as well,
	effectively replacing previous suboptimal methods.

	* src/chm_common.c (chm2Ralloc): Fix unidiagonal ntC segfault:
	assign 'x' only when non-pattern.
	* src/chm_common.c (as_cholmod_triplet): reallocate now in a way
	that works; fix documentation about return value in diagU2N case;
	ditto for
	* src/chm_common.c (as_cholmod_sparse):

	* R/sparseMatrix.R (printSpMatrix): add 'cld' argument, typically
	passed from printSpMatrix2; and indicate "unit-diagonal"

2009-10-22  Martin Maechler  <>

	* R/lsparseMatrix.R (C2l): fix for case with NA.

	* R/Csparse.R (replCmat): drop "stale" cached @factors
	factorizations after sub-assignments.
	* R/Tsparse.R (replTmat, .TM.repl.i.2col): ditto

2009-10-19  Martin Maechler  <>

	* src/dgCMatrix.c (dgCMatrix_LU): new boolean argument
	'error_on_sing' to allow no error in case of singularity;
	needed for determinant(), e.g.

	* R/Auxiliaries.R (detSparseLU): using lu(*, errSing=FALSE)

	* R/dgCMatrix.R, R/dsparseMatrix.R: lu() methods, using 'errSing'

	* R/sparseMatrix.R (printSpMatrix): fix bug introduced on *-09-10
	* tests/Simple.R: test for that.

2009-10-18  Martin Maechler  <>

	* src/dgeMatrix.c (dgeMatrix_crossprod): do not fail in 0-column case.

	* inst/test-tools.R (Q.eq): new utility
	(checkMatrix): minimally check %*%, crossprod() and tcrossprod()

	* R/products.R: more '%*%' methods for "[ln]?Matrix", "missing"

2009-10-06  Martin Maechler  <>

	* DESCRIPTION (Version): 0.999375-31, CRAN-released: 2009-10-06

	* inst/include/Matrix_stubs.c (M_R_cholmod_error): revert (2009-09-18),
	i.e., no longer restore cholmod_common. {{M_cholmod_defaults()
	still seems not usable from lme4's init.c}}

2009-10-05  Martin Maechler  <>

 	* src/dtrMatrix.c (dtrMatrix_chol2inv): use "dpoMatrix" instead of "dsy"

	* R/dtrMatrix.R: make use of implicit generic for chol2inv() in
	newer R versions.

2009-09-30  Martin Maechler  <>

	* R/CHMfactor.R (solve): fix methods for "ddiMatrix" and "missing" RHS.
	* tests/factorizing.R: test these

	* R/Matrix.R (image): fix Matrix method to work for "ddiMatrix"
	* R/diagMatrix.R: coercion to "dsparse*"
	* tests/Simple.R: test image(<ddiMatrix>)

2009-09-29  Martin Maechler  <>

	* R/AllGeneric.R: rcond is implicit generic in latest versions of R

2009-09-22  Martin Maechler  <>

	* R/Ops.R (A.M.n,A.n.M): replace "Ops" methods by explicit "Arith",
	"Logic", etc, getting rid of ambiguity (notes), and of infinite recursions.

	* tests/group-methods.R: test these systematically.

2009-09-18  Martin Maechler  <>

	* inst/include/Matrix_stubs.c (M_R_cholmod_start): print_function = NULL
	as in src/chm_common.c (2009-07-20)
	(M_R_cholmod_error): ditto, using new M_cholmod_default(), declared in
	* inst/include/cholmod.h

	* R/Tsparse.R (intI): do *not* allow logical subscript (row or
	column) to be too long, compatibly with traditional matrix
	* tests/indexing.R: and assert the error.

2009-09-17  Martin Maechler  <>

	* R/pMatrix.R: as(sparseMatrix, pMatrix)

	* R/CHMfactor.R (solve): method for (CHMfactor, missing)

	* inst/test-tools.R (assertError): use tryCatch()

	* R/diagMatrix.R (.sparseDiagonal): fix shape "g" case.

	* R/Auxiliaries.R (isTriC): do not wrongly return TRUE for a *sCMatrix.

	* man/chol2inv-methods.Rd: document & example

2009-09-16   Douglas Bates <>

	* NAMESPACE, R/dtrMatrix.R, src/dtrMatrix.[ch]: add chol2inv()
	method for dtrMatrix.

2009-09-12  Martin Maechler  <>

	* R/sparseVector.R ([): allow *indexing* with "lsparseVector"

2009-09-11  Martin Maechler  <>

	* R/sparseVector.R (prSpVector): using ":" as in printSpMatrix()
	(Summary): add "Summary" group methods
	* man/sparseVector-class.Rd: and test a bit

2009-09-10  Martin Maechler  <>

	* R/sparseMatrix.R (printSpMatrix): visually differentiate
	non-structural zeros in *logical* sparse matrices, using ":"

	* R/Auxiliaries.R (setparts): new utility, for
	* R/Ops.R (Ops.spV.spV): start implementing sparseVector arithmetic etc

2009-09-08  Martin Maechler  <>

	* R/dgCMatrix.R (qr): for sparseMatrix must coerce to "dgCMatrix".
	* tests/factorizing.R: test qr(<dsCMatrix>)

2009-09-01  Martin Maechler  <>

	* R/CHMfactor.R (determinant): divide previous log(det(.)) by 2;
	now returning det(L); and modifiy the "CHMfactor.warn" message.

	* man/CHMfactor-class.Rd: modify documentation accordingly.

2009-08-21  Martin Maechler  <>

	* R/spModels.R (`contrasts<-`): a version that can also work with a
	"sparseMatrix" value.  This is put into R 2.10.0 (devel) as well.

	* src/Mutils.h: rename any_NA to any_NA_in_x
	* src/Mutils.c (check_scalar_string): add doxygen doc

2009-08-15  Martin Maechler  <>

	* R/spModels.R (fac2Sparse): make also work for 'contrasts.arg' =

2009-07-28  Martin Maechler  <>

	* R/spModels.R (contr.sum): need also to define contr.*() as long
	as we document (./man/) them.

2009-07-27  Martin Maechler  <>

	* DESCRIPTION (Version): 0.999375-30, CRAN-released: 2009-07-28

	* R/Matrix.R (all.equal_Mat): add factorsCheck=FALSE argument
	* R/Auxiliaries.R (attr.all_Mat, attrSlots): ditto

2009-07-25  Martin Maechler  <>

	* R/Auxiliaries.R (attr.all_Mat): fix checking of non-content slots.
	* R/Matrix.R (all.equal_Mat): thinko needed  s/&&/&/
	* R/sparseMatrix.R (all.equal(.) methods): ditto
	===> Note:  all.equal() is more stringent for "Matrix" arguments now!

2009-07-23  Martin Maechler  <>

	* R/spModels.R (model.spmatrix, sparse2int): "complete" re-write

	* tests/spModel.matrix.R: many tests added

2009-07-20  Martin Maechler  <>

	* src/chm_common.c (R_cholmod_l_start): set print_function to NULL,
	as we have long suggested ==> get rid of random strings seen in
	some cholmod warnings.
	(R_cholmod_error): call cholmod_l_defaults() before error(), so
	we restore the Cholmod global in case of errors.

	* R/ldenseMatrix.R (.rcond_via_d): fix thinko

2009-07-18  Martin Maechler  <>

	* R/CHMfactor.R (isLDL): need a "!" as 'type' is "is_ll"
	* src/dsCMatrix.c (dsCMatrix_Cholesky): update, notably when caching.

	* tests/indexing.R: test col.names printing of sparseMatrix

2009-07-16  Martin Maechler  <>

	* inst/test-tools.R (allCholesky): new testing function

2009-07-15  Martin Maechler  <>

	* src/dsCMatrix.c (dsCMatrix_Cholesky): add possibility to set each
	of 'perm', 'LDL', and 'super' to NA (in addition to TRUE / FALSE).
	in these case, a CHOLMOD-heuristic choses the option "sensibly".
	* man/Cholesky.Rd: document the new possibility.

2009-07-14  Martin Maechler  <>

	* R/rankMatrix.R (rankMatrix): diff(sval) <= 0

	* R/spModels.R (model.spmatrix): fix case of missing main effect
	* tests/spModel.matrix.R: new file

2009-07-11  Martin Maechler  <>

	* R/sparseMatrix.R (show, printSpMatrix2): both print() and show()
	now use printSpMatrix2(), and that now already prints
	"<n> x <m> sparse Matrix of class ...".

	* R/CHMfactor.R (isLDL): fix and
	* NAMESPACE: export  isLDL()

2009-07-10  Martin Maechler  <>

	* R/spModels.R (model.spmatrix): mf may be simple data.frame

2009-07-09  Martin Maechler  <>

	* NAMESPACE: export sparse.model.matrix() and
	* man/sparse.model.matrix.Rd: document it

2009-07-08  Martin Maechler  <>

	* R/Tsparse.R (intI): also work for integer dimnames (well ..)

2009-07-07  Martin Maechler  <>

	* R/sparseMatrix.R: "factor out" sparse model things into

	* R/spModels.R (sparse.model.matrix): new model matrix functions

2009-06-20  Douglas Bates <>

	* src/CHMfactor.c: Ensure updated LL stays LL

2009-06-10  Martin Maechler  <>

	* DESCRIPTION (Version): 0.999375-29, CRAN-released: 2009-06-11

2009-06-10  Douglas Bates <>

	* [r2404] src/Mutils.c: Change value of set_factor to be the cached factor
	* [r2403] src/dgCMatrix.c, src/dgCMatrix.h, src/init.c: Comment out
	  unused dgCMatrix_lusol function
	* [r2402] R/dgCMatrix.R: R-level implementation of solve("dgCMatrix",
	* [r2401] src/dgCMatrix.c: Re-arrange LU factorization to always use the
	  cached value.

2009-06-09  Douglas Bates <>

	* [r2399] src/dgCMatrix.c: PROTECT the result from dgCMatrix_LU in

2009-06-06  Martin Maechler  <>

	* R/Tsparse.R: add numeric -> Tsparse* coercion

2009-06-05  Martin Maechler  <>

	* src/Mutils.h, src/dgeMatrix.c: using dngettext(.) only
	on future versions of R.

2009-06-04  Martin Maechler  <>

	* DESCRIPTION (Version): 0.999375-28, CRAN-released: 2009-06-08

	* po/de.po, po/R-de.po: German translations from Chris Leick.
	* inst/po/de/: ditto, via po/

2009-05-28  Martin Maechler  <>

	* src/chm_common.c, src/cs_utils.c, etc: internationalize more
	messages; fix some; thanks to feedback from Chris Leick.

2009-05-27  Martin Maechler  <>

	* man/denseMatrix-class.Rd, etc: 'factors' is *not* a slot in this
	class; found by the upcoming R 2.10.0 codocClasses().

2009-05-25  Martin Maechler  <>

	* po/, Matrix.pot, etc: updated *.pot / *.po files

	* DESCRIPTION (Version, Date): 0.999375-27, CRAN-released today.

	* R/sparseVector.R: add as(<atomic>, "dsparseVector")
	(spV2M): now works (again!?) for "isparseVector" (-> "dgTMatrix").

	* tests/matprod.R: tcrossprod(), sparseVector multiplications, i.e.,
	features of next paragraph.

2009-05-23  Martin Maechler  <>

	* R/products.R: move almost all %*%, crossprod(), tcrossprod()
	methods to a new file.  tcrossprod() compatibility with *fixed*
	base-R; enable operations with "sparseVector"s; some extra methods
	to avoid ambiguity messages.

2009-05-20  Martin Maechler  <>

	* R/Auxiliaries.R (.M.v, .v.M): slight speedup, and *use* them in
	* R/Matrix.R (crossprod)

2009-05-18  Martin Maechler  <>

	* R/sparseVector.R (dim<-): prod(as.integer(.)) may overflow!

	* R/Matrix.R (Matrix): Matrix(sv, ..) now works for a sparseVector

	* R/sparseVector.R (spV2M): allow zero nrow or ncol.

2009-05-16  Martin Maechler  <>

	* R/sparseMatrix.R (dim<-): should also work for diagonalMatrix.

	* inst/test-tools.R (vec): add test for dim(x) <- c(n, 1)

2009-05-13  Martin Maechler  <>

	* R/rankMatrix.R (rankMatrix): new function for matrix rank;
	"to be discussed".

2009-05-07  Doug Bates and Martin Maechler <>

	* src/Mutils.c (Matrix_check_class_etc): diverse smallish
	improvements, stuffing a small leak, adding PROTECT()ion,
	but, alas, not enough.

2009-05-06  Martin Maechler  <>

	* R/HBMM.R (readMM): add checkIJ() to produce nicer error messages,
	on malformed input.
	* tests/write-read.R: and test that..

2009-04-18  Martin Maechler  <>

	* DESCRIPTION (Version, Date): 0.999375-26, CRAN-released on 2009-04-29

	* src/Mutils.h (Matrix_check_class_etc): try to ensure it works
	also when Matrix is loaded but not attached.

	* src/init.c (R_init_Matrix): get Matrix namespace in C.

	* R/zzz.R (.onLoad): a *working* fix aka "hack"

2009-04-15  Martin Maechler  <>

	* DESCRIPTION (Version): 0.999375-25 packaged --> CRAN

2009-04-09  Martin Maechler  <>

	* R/Auxiliaries.R (Matrix.msg): new utility
	(.M.vectorSub): ditto, for vector-indexing; in
	* R/Matrix.R ([): M[i, drop=<LOGI>] should do *vector*-indexing.
	* R/Tsparse.R ([): ditto;
	* R/diagMatrix.R ([, subDiag): ditto.

	* R/Tsparse.R ([): more careful indexing of (triangular)

	* tests/indexing.R: testing the above

	* R/Auxiliaries.R (gT2tT, check.gT2tT): consistency and efficiency
	improvement, using 'do.n' argument.

2009-04-08  Martin Maechler  <>

	* R/Matrix.R: add  as(., "vector")  etc
	* man/Matrix-class.Rd: ditto
	* inst/test-tools.R (checkMatrix): and check them

2009-04-07  Martin Maechler  <>

	* DESCRIPTION (Version, Date): 0.999375-24 ... released to CRAN

	* R/sparseVector.R: fix coercion from xsparse*
	* tests/Simple.R: and check it.

	* man/lu.Rd: document 'warnSing' argument
	* tests/dg_Matrix.R: and test it

	* src/dgeMatrix.c (dgeMatrix_LU): missing 'return'

2009-04-06  Martin Maechler  <>

	* DESCRIPTION (Version, Date): 0.999375-24 only for 2.9.0 (and newer)

2009-03-30  Martin Maechler  <>

	* src/Mutils.h (Matrix_check_class_etc): new version that *computes*
	the environment to be used.
	* src/Tsparse.c, src/cs_utils.c, src/chm_common.c, src/dgCMatrix.c:
	use the above in place of Matrix_check_class_and_super().

2009-03-26  Martin Maechler  <>

	* R/Auxiliaries.R (MatrixClass): use cld@package, not packageSlot(.) !

2009-03-25  Martin Maechler  <>

	* DESCRIPTION (Version, Date): 0.999375-23

	* tests/Class+Meth.R (dotestMat): use getClassDef() for speed;
	adaptions to also work with "Matrix-extensions".
	* inst/test-tools.R (checkMatrix): similar adaptions.

2009-03-24  Martin Maechler  <>

	* R/Auxiliaries.R (MatrixClass, l2d_meth, ...): make use of
	MatrixClass(): functionality for setClass("foo", contains="dgCMatrix")

	* src/Mutils.h (Matrix_check_class_and_super): new utility, to be
	used in lieu of  Matrix_check_class() in most cases.
	* src/Tsparse.c, src/cs_utils.c, src/chm_common.c, src/dgCMatrix.c: use it,
	currently only with R_GlobalEnv {otherwise: API change}

2009-03-12  Martin Maechler  <>

	* man/band.Rd: note and example about band(<dense>) |-> dense

	* R/ddenseMatrix.R (.bandDense): fix typo in argument check.
	* R/Csparse.R: ditto
	* src/dense.c (dense_band): limit index range; thanks to Bill Dunlap.
	* tests/Simple.R (band): check it

2009-03-11  Martin Maechler  <>

	* R/dMatrix.R (Summary): (dsparseMatrix): new logic, fixing
	prod(<sparse with NA>).

	* inst/test-tools.R (eqDeterminant): improve after suggestions from Doug
	* inst/test-tools.R (checkMatrix): message(), not warning(),
	for differing prod().

	* src/dgeMatrix.h, src/init.c: dgeMatrix_LU(x, warn_singularity)
	* src/dgeMatrix.c (dgeMatrix_LU): allow to suppress singularity warning.
	(dgeMatrix_determinant, dgeMatrix_rcond): do *not* warn on exact singularity

2009-03-10  Martin Maechler  <>

	* R/Matrix.R (Summary.l, no conversion to "dMatrix" for all()/any()

	* tests/Simple.R: do not use memory-expensive all(I.. == Diagonal(n))
		for R >= 2.9.0, keep option 'warn = 2', i.e. warnings
		producing errors  ==>

	* R/Ops.R: small fixes, getting rid of warnings in tests/Simple.R

	* TODO: think about providing something like  allCompare(.,., "==")

2009-03-02  Martin Maechler  <>

	* DESCRIPTION (Version): 0.999375-22, ready but not submitted

2009-03-01  Martin Maechler  <>

	* inst/test-tools.R (eqDeterminant): fix for NAs

2009-02-28  Martin Maechler  <>

	* R/diagMatrix.R (prod, all): fix methods (cut & paste remnant);
	for NA.

	* R/Auxiliaries.R (detSparseLU): determinant(<..NA..>) now gives NaN

	* R/sparseMatrix.R (printSpMatrix): workaround R bug
	* tests/Simple.R: test that

2009-02-27  Martin Maechler  <>

	* R/Matrix.R (Matrix): Matrix(x,*) returns x unaltered when
	is(x, "diagonalMatrix"); OTOH, Matrix(x,1,1) should typically
	*not* return a diagonalMatrix.

	* R/diagMatrix.R (setAs(matrix, *)): fix for NAs.
	* tests/Simple.R: test things like Matrix(NA, 1,1)

2009-02-25  Martin Maechler  <>

	* NAMESPACE: add bandSparse() "properly"
	* man/bandSparse.Rd: doc including examples

2009-02-20  Martin Maechler  <>

	* R/bandSparse.R (bandSparse): constructor for band(diagonal)
	sparse matrices.

2009-02-13  Martin Maechler  <>

	* DESCRIPTION (Version): 0.999375-21, released to CRAN

	* src/Mutils.h (Matrix_with_SPQR): #define (or #undef), in order to
	allow easy building "SPQR-free version" of Matrix.

2009-02-11  Martin Maechler  <>

	* R/Csparse.R (replCmat): another check for 'has.x'
	* tests/indexing.R: very large (very sparse) sub-indexing.

2009-01-30  Martin Maechler  <>

	* DESCRIPTION (Version): 0.999375-20
	* man/spqr.Rd: disable example on Windows for now

	* inst/test-tools.R (checkMatrix): simpler for "logical" pMatrix.

	* R/ngTMatrix.R: fix ngT -> nge / lge coercion, and
	* tests/Simple.R: adapt test.

	* R/pMatrix.R: coercion to "matrix": more sensical to coerce to logical
	instead of 0/1 integer.   This a mild back-compatibility breach.
	* man/pMatrix-class.Rd: adapt, including example
	* R/sparseMatrix.R (printSpMatrix): print as logical, too.

2009-01-29  Martin Maechler  <>

	* R/Auxiliaries.R (geClass): define for "pMatrix"
	* R/pMatrix.R: pMatrix  ->  ngeMatrix coercion
	* man/pMatrix-class.Rd:

	* DESCRIPTION (Version): 0.999375-19  considering release.

	* R/ngTMatrix.R: coercing correctly to ngeMatrix via lge*.

	* man/dgTMatrix-class.Rd: remove no-more-existing coercion method;
	mention a bit less, and note.

	* R/diagMatrix.R: ensure that <diagonal> * <dgeMatrix> is diagonal
	even with new method dispatch

	* R/Matrix.R (.M.sub.i.logical): change comments only, and extend
	error message which is never called currently.

	* R/Ops.R (Ops.x.x): fix for new "inherited method" dispatch

2009-01-28  Martin Maechler  <>

	* R/spqr.R: put spqr() / class "SPQR" related code in one file for now.
	* NAMESPACE: export "SPQR" class
	* man/SPQR-class.Rd: document it

2009-01-27  Martin Maechler  <>

	* R/Auxiliaries.R (is.na_nsp): should produce "nsCMatrix" only when
	dimnames are symmetric too.

	* R/sparseQR.R (solve): method for (<sparseQR>, <ANY>)

2009-01-26  Douglas Bates <>

	* src/cs_utils.c (Matrix_as_cs): in diagU2N case: Force sorted
	columns after expanding unit diagonal.

2009-01-21  Martin Maechler  <>

	* R/Tsparse.R (intI): for now give intelligible error on NA indices.

	* R/Matrix.R (subset.ij): should deal correctly with NA indices in
	the 2-column-matrix index case.

	* src/Mutils.c (m_encodeInd, m_encodeInd2): prepare to deal better
	with NA in indices.

2009-01-20  Martin Maechler  <>

	* inst/doc/Intro2Matrix.Rnw: mention sparseMatrix
	* man/sparseMatrix.Rd: fix typos found by parse_Rd

2009-01-19  Martin Maechler  <>

	* DESCRIPTION (Version,Date): release 0.999375-18 (as of yesterday)

2009-01-18  Douglas Bates <>

	* [r2319] src/Csparse.c, src/Csparse.h: Added an untested create_Csparse

2009-01-17  Martin Maechler  <>

	* R/sparseMatrix.R (sparseMatrix): recycle 'x' if necessary;
	use 'index1' rather than 'index0' {as proposed by Doug}.

	* R/dgeMatrix.R: drop two unnecessary (maybe very very slightly
	faster) methods for %*%.

2009-01-12  Douglas Bates <>

	* [r2313] R/sparseMatrix.R, man/sparseMatrix.rd: Use intermediate
	triplet rep in sparseMatrix.

2009-01-07  Martin Maechler  <>

	* R/sparseMatrix.R (sparseMatrix): new function to be used in
	place of new(), notably for CsparseMatrix objects.
	* man/sparseMatrix.Rd: its doc; plus examples

	* man/CsparseMatrix-class.Rd: fix long-standing thinko; update the
	\note{} to reflect the fact that the validity method no longer
	modifies its argument.

	* R/Csparse.R (.validateCsparse): new utility accessing
	* src/Csparse.c (Csparse_validate_, Csparse_validate2): new
	utilities allowing the "sort-in-place" that used to be part of
	former Csparse_validate().

2008-12-10  Douglas Bates <>

	* [r2309] DESCRIPTION: Release 0.99375-17

2008-12-05  Douglas Bates <>

	* [r2308] inst/include/Matrix_stubs.c, inst/include/cholmod.h: Adding
	  const for picky compilers; Using CHM_FR typedef

2008-11-17  Martin Maechler  <>

	* [r2307] ChangeLog, tests/validObj.R: update (as of Oct.26)

2008-10-31  Douglas Bates <>

	* [r2306] inst/include/Matrix_stubs.c: Consistency with cholmod.h re
	  const qualifier

2008-10-26  Douglas Bates <>

	* src/Csparse.c (Csparse_validate): do not sort, but report error
	on unsorted row-indices within column.

	* tests/validObj.R: adapt the regression test.

2008-10-17  Douglas Bates <>

	* [r2300] inst/include/Matrix_stubs.c: Consistency with SparseSuite names
	* [r2299] src/AMD/Source/Makefile, src/CHMfactor.c,
	  src/CHOLMOD/Lib/Makefile, src/COLAMD/Source/Makefile, src/Csparse.c,
	  src/Tsparse.c, src/UFconfig/UFconfig.h, src/chm_common.c,
	  src/chm_common.h, src/dense.c, src/dgCMatrix.c, src/dsCMatrix.c,
	  src/init.c, src/t_gCMatrix_colSums.c: Compile only the UF_long version
	  of SparseSuite but setting UF_long to be int

2008-10-17  Martin Maechler  <>

	* src/Mutils.h: include <alloca.h> also for AIX

2008-10-16  Martin Maechler  <>

	* src/scripts/ : move *.sh, *.mkf and *.mk files from src/ to src/scripts/

	* src/*/{Lib|Source}/Makefile: change to non-GNU-make style, and
	other improvements from Brian Ripley.

2008-10-15  Martin Maechler  <>

	* src/Makevars, src/ replacing
	* src/Makefile, src/ which are no longer needed

2008-10-14  Martin Maechler  <>

	* src/Makefile, src/*/Makefile: changes from Brian Ripley enabling
	parallel make.

2008-10-13  Douglas Bates <>

	* [r2285] src/SPQR/Include/spqr.hpp: Include the C string declarations.

2008-10-06  Douglas Bates <>

	* [r2284] src/Makefile, src/SPQR/Lib/Makefile_SPQR, src/SPQR/Makefile,
	  src/ Modifications for compilation under Windows
	* [r2283] tests/validObj.R: remove reference to undefined object
	* [r2282] inst/doc/UFsparse/SPQR.txt: SPQR license information

2008-10-04  Douglas Bates <>

	* [r2281] src/Makefile, src/chm_common.c, src/chm_common.h,
	  src/dgCMatrix.c, src/init.c: SparseSuiteQR interface

2008-10-03  Martin Maechler  <>

	* src/Csparse.c (isValid_Csparse): new utility
	* src/chm_common.c (as_cholmod_sparse): add validity check early
	* tests/validObj.R: test the above

2008-10-02  Douglas Bates <>

	* [r2277] TODO: Think of a better way of structuring include files
	* [r2276] src/AMD/Source/Makefile, src/CHOLMOD/Lib/Makefile,
	  src/COLAMD/Source/Makefile, src/Makefile,
	  src/SPQR/Include/SuiteSparseQR_C.h, src/chm_common.h, src/dgCMatrix.c,
	  src/dgCMatrix.h, src/init.c, src/sparseQR.h: Add dgCMatrix_SPQR and
	  modify other code to allow its compilation and linkage

2008-10-02  Martin Maechler  <>

	* R/sparseMatrix.R (printSpMatrix2): another validObject(.) call
	* src/Makefile: getting rid of SUBSTAMP etc, thanks to Simon U.

2008-10-01  Douglas Bates <>

	* src/Makefile, inst/doc/UFsparse/, inst/include/: Adjustments for
	SuiteSparse configuration.
	* src/CHOLMOD, src/UFconfig: update versions of SuiteSparse
	* src/SPQR/*  add "SPQR", from Tim Davis' "SuiteSparse" collection,
	not yet with an R interface.

2008-09-25  Martin Maechler  <>

	* Release 0.999375-15 to CRAN.

2008-09-23  Martin Maechler  <>

	* src/dsyMatrix.c (dsyMatrix_as_dspMatrix): copy 'factors' slot
	* tests/dpo-test.R: test for it

	* R/Tsparse.R (intI): improve one error message.

2008-09-22  Martin Maechler  <>

	* DESCRIPTION (Version): 0.999375-15 to be released

	* R/diagMatrix.R (diag o <Matrix>): explicit setMethods,
	in order to keep result diagonal in many more cases.
	(coerce) to denseMatrix now *does* coerce.

	* man/diagonalMatrix-class.Rd: \alias{} for these.

	* R/Auxiliaries.R (.dense.prefixes): ".diMatrix" has not been dense

	* R/dMatrix.R: as(<dMatrix>, <nMatrix>): enable Cholesky/BunchKaufman

2008-09-20  Martin Maechler  <>

	* R/lsparseMatrix.R (C2l): as(<d.Csparse>, "lMatrix") should
	preserve NAs.

	* R/Ops.R (Arith, Compare): keep diag="U" in more cases when sensible

2008-09-19  Martin Maechler  <>

	* R/Ops.R (Arith): better <ddenseMatrix> o <numeric> which
	preserves symmetry / triangularity in "obvious" cases.

	* R/dpoMatrix.R: setAs(., "lMatrix") and setAs(., "nMatrix")
	* R/dppMatrix.R: ditto
	* man/dpoMatrix-class.Rd:

	* inst/test-tools.R (checkMatrix): add [dln]Matrix <-> [dln]Matrix
	coercion checks.
	* tests/ more ambiguity warnings from the new checks.

	* R/dMatrix.R: dMatrix -> nMatrix: fix dense case.

2008-09-18  Martin Maechler  <>

	* tests/factorizing.R: test expand(<sparseLU>)

	* src/dgCMatrix.c (dgCMatrix_LU): fill @Dim slot correctly.

	* R/Tsparse.R (replTmat): also optionally warn when sub-assignment
	loses symmetry of Matrix.

2008-09-17  Martin Maechler  <>

	* R/Tsparse.R (.TM.repl.i.2col): fix sub-assignment of "dsCMatrix";
	bug reported by Jacob van Etten.

	* tests/indexing.R, tests/ testing it

2008-09-10  Douglas Bates <>

	* [r2260] src/Mutils.h, src/chm_common.c, src/cs_utils.c,
	  src/dsCMatrix.c: Update doxygen comments

2008-09-10  Martin Maechler  <>

	* man/ddiMatrix-class.Rd: docu update : sparse, not dense (see 2008-07-28).
	* man/ldiMatrix-class.Rd , man/diagonalMatrix-class.Rd: ditto.

2008-09-08  Martin Maechler  <>

	* DESCRIPTION (Version): 0.999375-14  to be released to CRAN,
	in order to comply to to pending changes in R-devel (2.8.0).

2008-09-05  Martin Maechler  <>

	* R/Matrix.R (solve,..): use (Matrix,ANY)... instead of (Matrix,Matrix)
	* R/Ops.R: ditto; most of these from John Chambers.
	* man/Matrix-class.Rd

	* R/Auxiliaries.R (.diagU2N): fix for Rsparse*
	* tests/Simple.R: test it

2008-09-02  Martin Maechler  <>

	* man/drop0.Rd:
	* R/Auxiliaries.R (drop0): new argument 'tol = 0' (and
	'is.Csparse'); entails updates in
	* R/Csparse.R, R/lMatrix.R, R/dMatrix.R

2008-08-30  Martin Maechler  <>

	* DESCRIPTION (Version): 0.999375-13 released to CRAN

2008-08-29  Martin Maechler  <>

	* R/Auxiliaries.R (attrSlotNames, attrSlots, attr.all_Mat): new
	utilities, for now to be used in all.equal() methods.

	* R/Matrix.R (all.equal_Mat): be more careful (less tolerant) in
	all.equal() methods.
	* R/sparseMatrix.R: ditto

2008-08-28  Martin Maechler  <>

	* DESCRIPTION (Version): 0.999375-12 released to CRAN

	* R/Ops.R (Compare(<Csparse>,<Csparse>): fix for nsparseMatrix

2008-08-27  Douglas Bates <>

	* R/sparseMatrix.R: fac2sparse() for NA's

2008-08-26  Martin Maechler  <>

	* R/sparseVector.R (all.equal.sparseV): make work for nsparseVector's

2008-08-25  Martin Maechler  <>

	* src/dgCMatrix.c (dgCMatrix_LU): partially revert change
	r2175 (2008-04-23) and do give an error for a singular matrix.
	* man/lu.Rd, R/Auxiliaries.R (detSparseLU): adapt to that.

	* R/LU.R: expand( <sparseLU> )

	* NAMESPACE, man/all.equal-methods.Rd:
	* R/Matrix.R (all.equal): methods for Matrices,
	* R/sparseMatrix.R, R/sparseVector.R: sparseMatrices and -Vectors

2008-08-23  Douglas Bates <>

	* [r2243] R/CHMfactor.R, man/CHMfactor-class.Rd: Added expand method and
	  documentation for CHMfactor class
	* [r2241] R/CHMfactor.R: Added trivial coercion method for CHMfactor ->

2008-08-19  Martin Maechler  <>

	* R/nsCMatrix.R (setAs(., dgTMatrix))): remove unused method

2008-08-18  Martin Maechler  <>

	* R/Ops.R (.Ops2dge.via.x, and many others): eliminate never-used
	variable assignments {from "next version" codetools}.
	* R/Csparse.R, R/Tsparse.R: ditto

2008-08-17  Martin Maechler  <>

	* R/sparseVector.R (sp2vec, coerce to sparseVector):
	make sure no integer overflows happen, and use double precision
	(n, index) where appropriate.

	* tests/Simple.R: test "large" sparse vectors.

2008-08-14  Martin Maechler  <>

	* R/AllClass.R (sparseVector): 'length' and 'i' are "numeric", not
	just integer (allowing much longer length).

2008-07-28  Martin Maechler  <>

	* R/AllClass.R (diagonalMatrix): extend  "sparseMatrix" instead of
	"denseMatrix".  This renders "scarceMatrix" dispensable and
	invalidates part of MM's presentations on "space of Matrix
	classes", but seems cleaner overall.

	* R/diagMatrix.R, etc: eliminate "scarceMatrix", replacing it by
	"sparseMatrix" in method signatures;
	further, instead of coercing to "sparseMatrix", coerce to
	"TsparseMatrix" now.

2008-07-26  Martin Maechler  <>

	* src/dgCMatrix.c (dgCMatrix_qrsol): allow third argument 'order' in
	* R/sparseMatrix.R (, src/dgCMatrix.h, src/init.c: ditto

2008-07-24  Martin Maechler  <>

	* R/dgeMatrix.R: need solve(<dge>, <sparse>) against infinite
	recursion in new test in
	* tests/matprod.R: testing the above and another solve() case, below

	* R/sparseMatrix.R ( slightly more efficient for
	e.g. triangular 'x'.

	* src/dgCMatrix.c (dgCMatrix_qrsol): use AS_CSP() hence allowing dtC*
	* src/dgCMatrix.c (dgCMatrix_cholsol): use AS_CHM_SP() to work
	correctly with unit-triangular x

	* src/dsCMatrix.c (dsCMatrix_Csparse_solve): use AS_CHM_SP()
	instead of not checking diagU.

	* R/diagMatrix.R, R/Auxiliaries.R: tweaks to allow later
	experiments where diagonalMatrix would extend sparse*.

2008-07-23  Martin Maechler  <>

	* src/dgCMatrix.c (compressed_non_0_ij): for nnzero(), use
	"x@p[ncol(x)]" instead of length(x@i).
	* R/Auxiliaries.R (nnzero): ditto

	* src/chm_common.c (as_cholmod_sparse): Do not use LENGTH(islot)
	since that fails for over-allocated i.

	* tests/validObj.R: more testing of over-allocated (i,x) slots

2008-07-23  Martin Maechler  <>

	* src/chm_common.c (as_cholmod_sparse): Add 4th argument
	'sort_in_place' and set it to TRUE in call from
	* src/Csparse.c (Csparse_validate): sort in place, making sure that
	validObject() continues sort the columns if needed.

	* tests/validObj.R: now tests that more thoroughly, and
	* man/CsparseMatrix-class.Rd: documents it.

2008-07-22  Douglas Bates <>

	* [r2233] src/chm_common.c: sort columns more carefully in
	* [r2231] src/chm_common.c: check for sorted columns in
	  as_cholmod_sparse instead of assuming it
	* [r2228] src/cs_utils.c: Use R_alloc for semi-permanent storage when
	  expanding diagonal

2008-07-21  Martin Maechler  <>

	* src/cs_utils.c (Matrix_as_cs): add check_Udiag switch
	* src/cs_utils.h (AS_CSP & AS_CSP__): two versions
	* src/dtCMatrix.c (dtCMatrix_sparse_solve): no longer needs

	* R/diagMatrix.R (.sparseDiagonal): new utility, used in "old"
	.spDiagonal() and new .trDiagonal().

	* R/dtCMatrix.R (solve): make use of .trDiagonal()

2008-07-19  Martin Maechler  <>

	* R/Auxiliaries.R (dimNamesCheck): fix very long standing buglet,
	stumbled upon by Michael H.

	* tests/validObj.R: testing the bug-fix

	* src/chm_common.h (AS_CHM_SP__, AS_CHM_TR__): the non-diagU2N

	* src/chm_common.c (as_cholmod_sparse, as_cholmod_triplet): new
	boolean argument 'check_Udiag' (formerly implicitly was FALSE).

	* src/Csparse.c (Csparse_Csparse_prod, ...): etc: use the new
	  AS_CHM_SP() which includes diagU2N.

	* inst/include/Matrix_stubs.c (M_as_cholmod_sparse): similarly
	adapt to 3 arguments.

2008-07-17  Douglas Bates <>

	* [r2220] src/dtCMatrix.c: Correct number of columns for rhs in
	* [r2219] src/cs_utils.c: eye->nz should be -1 for compressed column
	* [r2217] R/dtCMatrix.R, src/dtCMatrix.c, src/dtCMatrix.h, src/init.c:
	  Replace dtCMatrix_solve by more general dtCMatrix_sparse_solve and add
	  new solve method
	* [r2216] src/cs_utils.c: Utility csp_eye in case we want to fold
	  Csparse_diagU2N functionality into Matrix_as_cs

2008-07-17  Martin Maechler  <>

	* inst/doc/factor2sparse.Rnw: revive a year-old unfinished vignette

2008-07-16  Douglas Bates <>

	* [r2212] R/sparseMatrix.R: fac2sparse gets a drop.unused.levels
	  argument with default TRUE
	* [r2211] inst/include/Matrix.h, inst/include/Matrix_stubs.c: Export
	* [r2210] R/dtCMatrix.R, man/dtCMatrix-class.Rd, src/dtCMatrix.c,
	  src/dtCMatrix.h, src/init.c: Remove vestigial methods based on the
	  parent array; new solve method for signature(a = "dtCMatrix", b =

2008-07-14  Martin Maechler  <>

	* R/diagMatrix.R (.bdiag): also keep result triangular

2008-07-13  Douglas Bates <>

	* [r2208] src/dtCMatrix.c, src/dtTMatrix.c: Revise wording in error

2008-07-12  Martin Maechler  <>

	* R/diagMatrix.R (.bdiag): make more sophisticated, keeping
	symmetric- or nsparse-Matrix, also fixing the bug introduced with
	the new version (.999375-10).

	* tests/Simple.R: regression test for the bugs.

2008-07-07  Martin Maechler  <>

	* R/sparseVector.R (spV2M): fix for "isparseVector".
	Further, fix <diag> -> <sparseVector> coercion
	(repSpV): add rep(<sparseVector>, .) method

	* R/sparseMatrix.R ([<-): when the RHS is a "scarce"Matrix, do *not*
	use as.vector(.).

	* R/Tsparse.R ([<-): & [CR]sparse.R: methods for value = "sparseVector"

	* R/AllClass.R: new (hidden) class union "scarceMatrix"
	* R/AllClass.R: sparseVector: add prototype

	* src/dsCMatrix.c (chk_nm): fix "Cholesky" check, thanks to
	Kasper Kristensen

2008-06-28  Martin Maechler  <>

	* tests/other-pkgs.R: add more interesting Matrix -> graph

2008-06-27  Martin Maechler  <>

	* R/sparseMatrix.R (Tsp2grNEL): add 'need.uniq' argument for speed
	improvement in "[CR]sparseMatrix" -> "graphNEL" coercion.

2008-06-25  Martin Maechler  <>

	* DESCRIPTION: release 0.999375-10 to CRAN

	* R/diagMatrix.R (.bdiag): use more efficient Bates-algorithm for
	* man/bdiag.Rd: update, including test.

2008-06-24  Martin Maechler  <>

	* tests/Simple.R: add minimal bdiag() checks

	* R/diagMatrix.R (bdiag): fix bdiag(diag(4:5)) case;
	  factor out the Tsparse - internal computation into new .bdiag()

2008-06-14  Martin Maechler  <>

	* R/nsparseMatrix.R: specific fast all() and any() methods.

	* src/dgTMatrix.c (MAKE_gTMatrix_to_geMatrix): better error message
	when trying to produce too large result.

	* inst/test-tools.R (checkMatrix): add 'do.matrix' with a check for
	very large matrices.

	* R/sparseMatrix.R (printSpMatrix2): new function, used by

	* R/diagMatrix.R (show): print only diagonal entries when nrow(.) >= 50.

2008-06-13  Martin Maechler  <>

	* src/Mutils.c (m_encodeInd, m_encodeInd2):
	* R/*.R: .Call(m_encodeInd2?, ..) instead of encodeInd2?(..).

	* R/Auxiliaries.R (encodeInd2?): care against integer overflow (for
	big matrices): -> 2nd argument 'di' instead of 'nr' !

2008-06-09  Martin Maechler  <>

	* R/dtCMatrix.R: faster dgC* -> dtC* method
	* tests/Simple.R: (hence removed "FIXME")

	* R/Auxiliaries.R (copyClass): add 'check = .copyClass.check'
	which can be be turned off for possible speed gains.

2008-06-02  Martin Maechler  <>

	* R/dMatrix.R, R/Tsparse.R: get rid of some disambiguation warnings

	* R/not.R: fix !<ldiMatrix>
	implement two old "FIXME"s:  !<ldenseMatrix> & !<ndenseMatrix>

	* R/diagMatrix.R (Ops): fix one-off thinko in
	<diagonal> o <numeric> methods

	* inst/test-tools.R: checkMatrix(): drop0(m12) unconditionally

2008-05-30  Martin Maechler  <>

	* R/diagMatrix.R (Ops): more <diagonal> o <Matrix> methods
	* tests/Simple.R: testing the above

2008-05-07  Martin Maechler  <>

	* NAMESPACE: also import "Ops" {found via new tools:::get_S4_generics_with_methods()}

2008-05-03  Martin Maechler  <>

	* R/Matrix.R (dimnames<-): dimnames(.) <- NULL works with a message;
	* NAMESPACE (unname): finally exported

2008-04-28  Martin Maechler  <>

	* R/Auxiliaries.R (cholMat): possibly keep packed

	* inst/test-tools.R (checkMatrix): add fixed determinant.matrix()
	for R < 2.7.0.

	* R/Tsparse.R ([): for triangularMatrix, check if result may be
	* tests/indexing.R: regression test

2008-04-26  Martin Maechler  <>

	* inst/test-tools.R (checkMatrix): test <Ops> not only for dMatrix
	* R/Ops.R: more methods, for lMatrix/nMatrix

2008-04-24  Martin Maechler  <>

	* R/expm.R: new file for all expm() methods; notably new ones for
	diagonal*, triangular*, symmetric* and also simple "matrix".

2008-04-23  Martin Maechler  <>

	* R/dsyMatrix.R: setAs(.) rather than setIs(.)  "to dpoMatrix"

	* inst/test-tools.R (assert.EQ.mat): better message when not equal

	* src/dgeMatrix.c (dgeMatrix_matrix_crossprod): allow integer RHS.

	* R/AllClass.R (ddiMatrix,ldiMatrix): extend [dl]Matrix instead of
	[dl]denseMatrix {identical slots; but more logical method dispatch}

2008-04-23  Martin Maechler  <>

	* R/sparseMatrix.R (summary): use length(<i-slot>) instead of nnzero()

	* R/diagMatrix.R (determinant): method

2008-04-22  Martin Maechler  <>

	* src/dsCMatrix.c (dsCMatrix_Cholesky): suppress CHOLMOD printing too

	* R/Rsparse.R ( uplo fix for symmetric & triangular

	* R/dsCMatrix.R (determinant): switch to use Cholesky( 'LDL' ) and
	its diagonal

2008-04-21  Martin Maechler  <>

	* R/dMatrix.R (Summary): short-cut for prod(<sparse>)

	* R/not.R: fix !<"U"-diag-ltrMatrix> ; drop {R < 2.6.0}-branch

	* R/Auxiliaries.R (tT2gT): fix for "U"-diag-ltTMatrix

	* R/AllClass.R: ltTMatrix validity: call (more stringent) tTMatrix_validate

2008-04-19  Martin Maechler  <>

	* R/Ops.R (Ops.x.x): aux.function, not just for (dMatrix,dMatrix)

	* R/Ops.R (.do.Logic.lsparse): fix NA case

	* R/Tsparse.R (replTmat): fix r.sym case, using indTri()

	* R/Auxiliaries.R (nnzero): fix special cases, using
	* R/Auxiliaries.R (indDiag): new utility
	* R/Auxiliaries.R (indTri): new argument 'diag'

	* R/dMatrix.R: Summmary(<ddenseMatrix>): different branch for dsy*

2008-04-18  Martin Maechler  <>

	* R/diagMatrix.R: "Summary" methods, and more "Arith" / "Ops"

	* src/Csparse.c (Csparse_drop): preserve (uplo,diag) for
	".tCMatrix" triangular matrix.

	* R/Auxiliaries.R (drop0): use diagU2N(); according to helpfile,
	'clx' can be class *or* class representation.
	(nnzSparse): new.
	(non0ind): new 'xtendSymm' argument; used in
	* R/Ops.R:

	* R/pMatrix.R: more efficient determinant() method

2008-04-17  Martin Maechler  <>

	* [deactivated] R/Matrix.R (det): det() as base::det(), but with Matrix environment
	* [deactivated] NAMESPACE: export(det)

	* R/diagMatrix.R (mkDiag): new substitute for diag()

	* R/Auxiliaries.R (nnzero): fix for "diagU2N"-case
	(as0, as1): new utility.

	* R/Csparse.R (diag, band): need diagU2N()

	* src/dgeMatrix.c (get_norm): return NA when 'x' slot has NA/NaN.

	* R/lMatrix.R: coerce(nMatrix |-> lMatrix) fix

	* R/Ops.R (Compare): fix <nsp> o <nsp> case

	* R/nsparseMatrix.R, R/pMatrix.R: method

2008-04-16  Martin Maechler  <>

	* R/Auxiliaries.R (signPerm): new utility for
	* R/Auxiliaries.R (detSparseLU): determinant() via sparseLU

	* src/dsCMatrix.c (dsCMatrix_LDL_D): suppress 'CHOLMOD warning';
	since we have our own error handler (and can use tryCatch() from R).

2008-04-15  Martin Maechler  <>

	* R/dgTMatrix.R (image): ha! set col=NA if lwd is small -- very nice!

	* R/dsCMatrix.R (determinant): use tryCatch() and switch to lu()
	when not positive definite.

	* R/Auxiliaries.R (asTri): new auxiliary
	(non0.i): make *much* faster for Tsparse with many non-zero
	entries; add 'uniqT = TRUE' argument to be used when sorting is undesired.

	* tests/Class+Meth.R (dotestMat): now build on checkMatrix() from
	test-tools.R , see below

	* R/dMatrix.R: fix "Summary" method: + diagU2N()

	* NAMESPACE, R/Matrix.R: add mean() method

2008-04-12  Martin Maechler  <>

	* R/dgTMatrix.R (image): allow to *not* take abs(), and use
	cold-hot colors; try *changed* default useAbs = FALSE
	* man/image-methods.Rd:  nice examples

2008-04-11  Martin Maechler  <>

	* inst/test-tools.R (checkMatrix): new function for modularizing part
	of tstMatrixClass() in tests/Class+Meth.R

	* R/dsCMatrix.R: coercion from "matrix"
	* R/ltTMatrix.R, R/ntTMatrix.R: ditto (change it).

	* tests/Class+Meth.R (tstMatrixClass): some cleanup;
	add test for as(matrix(,0,0), <class>)

2008-04-09  Martin Maechler  <>

	* src/dgeMatrix.c (dgeMatrix_determinant): for n == 0,
	work as determinant(matrix(,0,0)).

	* src/dgCMatrix.c (dgCMatrix_LU): return {L,U} as dtCMatrix
	* man/sparseLU-class.Rd: ditto

	* R/dgCMatrix.R (determinant): method building on lu()
	* R/sparseMatrix.R, Matrix.R, ...: ditto

	* R/Auxiliaries.R (mkDet): auxiliary for determinant()

2008-04-07  Martin Maechler  <>

	* R/sparseMatrix.R (summary<sparseMatrix>): no 'x' column for
	pattern matrices.

2008-04-02  Martin Maechler  <>

	* src/dense.c (dense_to_Csparse): all declarations *before*
	R_CheckStack(); needed e.g. for ancient gcc 2.96.

2008-03-29  Martin Maechler  <>

	* DESCRIPTION (Version): 0.999375-9 --- need to release
	for R-devel (and R 2.7.0 alpha soon).

	* R/AllClass.R: drop "double" from "atomicVector" class union

	* R/AllGeneric.R (rcond): check for base::rcond instead of R version
	* R/dgeMatrix.R: ditto

	* R/sparseMatrix.R (summary): count NAs

	* inst/doc/Intro2Matrix.Rnw: changes, aim for *one* introduction.

2008-03-28  Martin Maechler  <>

	* R/AllGeneric.R: rcond(., norm) instead of rcond(., type), in
	order to become compatible to new R 2.7.0 base::rcond().

2008-03-25  Martin Maechler  <>

	* DESCRIPTION (Version,Date): 0.999375-8 --- released to CRAN

	* R/diagMatrix.R (Ops): fix newly introduce bug in <.di> o <numeric>

	* inst/test-tools.R (isValid): new utility, used much in
	* tests/simple.R:

	* man/BunchKaufman-methods.Rd: added too

2008-03-24  Martin Maechler  <>

	* R/dsyMatrix.R: add BunchKaufman() methods.
	* R/dspMatrix.R:

	* src/init.c: add dsyMatrix_trf

2008-03-23  Douglas Bates <>

	* DESCRIPTION (Version): release 0.999375-7
	* src/CHMfactor.c (CHMfactor_update): fix

2008-03-22  Martin Maechler  <>

	* src/dsCMatrix.c (dsCMatrix_LDL_D): cleanup, also using

	* R/AllGeneric.R: do *not* define a "chol" generic
		(but rather use the implicit one *without* making pivot
		part of the signature)
	* R/*.R: drop the 'pivot' from chol's signature and make
		'pivot=FALSE' a default argument of method definitions.

	* .Rbuildignore: add 'wrld_1deg': I.e. do not put it into released
	version of Matrix

2008-03-18  Martin Maechler  <>

	* R/Tsparse.R (.T.2.n): drop 0's before coercion to "nMatrix"

	* R/sparseMatrix.R ( new simple method
	* R/denseMatrix.R ( ditto.

	* R/diagMatrix.R (.symDiagonal): newly exported utility.

	* R/diagMatrix.R (Ops): <diag> * <numeric> should not become dgeMatrix!

	* src/ -->
	* src/CHOLMOD/: update to CHOLMOD version 1.6 (Nov.2007)

2008-03-17  Martin Maechler  <>

	* src/dsCMatrix.c (dsCMatrix_LDL_D): even faster utility with same
	functionality; barely tested in
	* tests/factorizing.R

	* src/Csparse.c (diag_tC): new functionality to enable faster
	determinant(<dsCMatrix>, .) in
	* R/dsCMatrix.R

2008-03-17 18:53  Douglas Bates <>

	* R/CHMfactor.R, inst/include/Matrix.h, inst/include/Matrix_stubs.c,
	  src/CHMfactor.c, src/CHMfactor.h,
	  man/CHMfactor-class.Rd, src/init.c, tests/factorizing.R:
	  Log-determinant of the parent matrix from a CHMfactor object as
	  chm_factor_ldetL2; documentation and support

2008-03-15  Martin Maechler  <>

	* R/dsCMatrix.R: enable determinant(<dsCMatrix>, .) via chol(.)

2008-03-14  Martin Maechler  <>

	* R/dsCMatrix.R: setAs(., "dsCMatrix") but with a deprecation

2008-03-13  Martin Maechler  <>

	* DESCRIPTION (Version, Date): bug-fix release 0.999375-6

	* R/diagMatrix.R (diag2tT.u): utility to be smarter in
	  <diagonal> o <sparse_triangular>
	  setAs(., "nMatrix") added.

	* R/diagMatrix.R (Diagonal): Diagonal(4, x=3) now works too

	* R/Auxiliaries.R (.diagU2N): more careful coercion in 2 steps
		new argument 'checkDense = FALSE'.

2008-03-07  Martin Maechler  <>

	* src/dgeMatrix.c (dgeMatrix_exp): fix the octave-origin bug in the
	back-permutation of the matrix exponential.

	* tests/matr-exp.R: test the fix (for an example where expm() was wrong).

	* DESCRIPTION (Date,Version): ready to release 0.999375-5

	* tests/simple.R: testing diagN2U

2008-03-06  Martin Maechler  <>

	* R/Auxiliaries.R (diagN2U): interface to

	* src/Csparse.c (Csparse_diagN2U): .Call()able SEXP version of chm_diagN2U()

	* tests/matprod.R: test for triangularity preserving <dtC> %*% <dtC>

2008-03-05  Martin Maechler  <>

	* src/chm_common.c (chm_diagN2U): new utility.

	* src/Csparse.c (Csparse_Csparse_crossprod, Csparse_Csparse_prod):
	make use of chm_diagN2U() and hence now preserve triangularity and

	* DESCRIPTION (LicenseDetails): new; such that 'License:' becomes "canonical"

2008-03-04  Martin Maechler  <>

	* R/diagMatrix.R (subDiag): fix when x[..] became a vector

	* src/Tsparse.c (Tsparse_diagU2N): new utility, necessary for e.g.
	* src/Csparse.c (Csparse_crossprod): use [CT]sparse_diagU2N() !!

	* R/Auxiliaries.R (.diagU2N): make use of new Tsparse_diagU2N

	* R/Ops.R ("-" (<sparseMatrix>)): use diagU2N()

	* src/chm_common.c (AS_CHM_FINISH): add note about problem
	for triangular (diag = "U").

2008-02-21  Martin Maechler  <>

	* R/Auxiliaries.R (as_Csparse2): drop this, replacing by
	* R/bind.R:  .Call(dense_to_Csparse, *)

2008-02-20  Martin Maechler  <>

	* R/Matrix.R (Matrix): no longer use coercions to specific classes
		(since we are discouraging them in user code).

	* tests/*.R: also replaces coercions to specific classes by
	coercions to super classes.

	* R/denseMatrix.R (.dense2C): simplified by using forceSymmetric()

2008-02-19  Martin Maechler  <>

	* man/CAex.Rd: example: coerce to symmetric*, not dsC*

	* src/dense.c (dense_band): generalized from ddense_band

	* R/ddenseMatrix.R (.trilDense,.triuDense, .bandDense):
	  now for "denseMatrix" (instead of just "ddense*"); much
	  simplified setMethod()s for these.

	* src/dense.c (dense_to_symmetric): generalized from ddense_*

2008-02-18  Martin Maechler  <>

	* R/AllGeneric.R: forceSymmetric() generic: fast no-checking
	version of as(*, "symmetricMatric").

	* src/dense.c (ddense_to_symmetric): add 'symm_test' argument, and
	* R/symmetricMatrix.R: set it TRUE in coercion to "symmetricMatrix"

2008-02-16  Martin Maechler  <>

	* R/Matrix.R (subset.ij): utility, as first step to faster
	   M [ <ij-matrix> ] indexing.

	* R/Matrix.R (.M.sub.i.logical): M[ <logical> ] : try better.

	* src/dense.c (ddense_symmpart, ddense_skewpart): new functions
	for more efficient symmpart() and skewpart() methods.

	* src/Mutils.c (equal_string_vectors): utility

	* src/dense.c (ddense_to_symmetric): new function used in

	* R/symmetricMatrix.R: setAs(., "symmetricMatrix")

	* R/sparseMatrix.R, et_al (isSymmetric): add '...' to formals, in
	order to match the generic (and evade .local(..)).

	* R/dsCMatrix.R: dgC -> dsC: use C code!

	* NAMESPACE, R/AllGeneric.R (symmpart, skewpart): new functions
	as per TODO

	* R/Auxiliaries.R (setZero, all0Matrix): new utilities

	* R/symmetricMatrix.R: obvious symmpart(), skewpart() methods

2008-02-15  Martin Maechler  <>

	* R/Ops.R (.Arith.Csparse): use diagU2N when needed in triangular

	* R/Auxiliaries.R (non0.i): take out of non0ind

	* R/ddenseMatrix.R (.trilDense, .triuDense, .bandDense):
	make tril(), triu(), band() work for "matrix" and all "dense*"

	* R/triangularMatrix.R (.tri[lu].tr): need .diagU2N(*)
	* tests/simple.R: test the fix above

	* R/sparseMatrix.R ([): simplification: no coerce to <general>
		before we use as(., <Tsparse>)[...] anyway

	* R/Rsparse.R ( mostly instead of

	* R/triangularMatrix.R (isTriangular): methods for all subclasses
	instead of "triangularMatrix" -- just to disambiguate

2008-02-14  Martin Maechler  <>

	* tests/Class+Meth.R (dotestMat): add m[FALSE, FALSE] { <-> TODO}
	* tests/indexing.R: example that fails above

2008-01-26  Martin Maechler  <>

	* R/Matrix.R (.M.sub.i.2col): fix for logical matrix indexing
	* R/Tsparse.R (.TM.sub.i.2col, .TM.repl.i.2col): ditto;
		now, M[lower.tri(M)] and  M[lower.tri(M)] <- v  work better

	* src/Tsparse.c (Tsparse_to_tCsparse): new fast utility.
	* src/Tsparse.h, init.c: ditto
	* R/Auxiliaries.R (isTriT): new; faster than going via Csparse
	     both isTriC() and isTriT() now return TRUE with "kind" or FALSE.
	* R/sparseMatrix.R (isTriangular): hence simplified

2008-01-24  Martin Maechler  <>

	* R/Ops.R (.Arith.Csparse): new utility factored out of former
	  <dgC> o <dgC>, extended for triangular and also used in  <dtC> o <dtC>.

2008-01-23  Martin Maechler  <>

	* tests/factorizing.R (checkSchur): and more tests for checking Schur()

	* inst/test-tools.R (isOrthogonal): new function; also file restructured

2008-01-22  Martin Maechler  <>

	* R/ngTMatrix.R: allow as(<numeric matrix>, "ngTMatrix") and hence
	coercion to "nMatrix" and "pMatrix"

	* R/AllClass.R: "Schur" class;  "number" class union
	* man/number-class.Rd:
	* man/Schur-class.Rd:

	* R/eigen.R (.dgeSchur): utility, and return "Schur" class
	(.simpleSchur): Schur() method for diagonal matrices
	(.triSchur):    Schur() method for triangular matrices
	(Schur.sym):    Schur() for symmetric matrices {building on eigen()}.

2008-01-21  Martin Maechler  <>

	* src/dgCMatrix.c (dgCMatrix_QR): set @Dim slot (as per doc)

2008-01-15  Martin Maechler  <>

	* R/CHMfactor.R (solve): method for b="numeric", but also b="ANY"
		in order to ensure 'system = *' is not lost;
		formals()$system instead of cut&paste.
	* tests/factorizing.R: test solve(<CHM..>, <numeric>)

2008-01-11  Martin Maechler  <>

	* DESCRIPTION (Date): make ready for release --> 0.999375-4

	* R/dgeMatrix.R: fix rcond() method for "dgeMatrix".

2007-12-08  Martin Maechler  <>

	* R/pMatrix.R: as(*, "matrix") now returns 0/1 *integer* matrix,
		and hence does as.vector(.).
	* man/pMatrix-class.Rd: docs
	* R/sparseMatrix.R: fix for printing "integer sparse"

	* tests/Class+Meth.R (tstMatrixClass): test M[FALSE], M[2] etc

	* R/Matrix.R and others: use "exact" function argument list for
		both "["   : (x, i,j, ..., drop)
		and  "[<-" : (x, i,j, ..., value)

	* R/denseMatrix.R: M[i] and M[i] <- v  (i vector) now work
	* R/Tsparse.R (replTmat):    ditto
	* R/diagMatrix.R (replDiag): ditto
	* R/Csparse.R (replCmat):    ditto {was it worth the pain?}
	* tests/indexing.R: testing the above

2007-12-07  Martin Maechler  <>

	* R/sparseMatrix.R (cov2cor): method for sparse matrices

	* R/diagMatrix.R ([<-): fix   D[ cbind(i,j) ] <- v

	* R/bind2.R: fix for Rsparse* and rbind2(dense,dense)

	* tests/Class+Meth.R: test cbind2, rbind2  and  diag<-

2007-12-06  Martin Maechler  <>

	* R/Matrix.R: "generic" cov2cor() method

	* R/nearPD.R: new 'only.values', 'keepDiag' arguments;
		speed up Q %*% D %*% t(Q)
	* tests/dpoMatrix.R: test nearPD()

2007-12-05  Doug Bates and Martin Maechler

	* R/sparseMatrix.R: xtabs(*, sparse=.) function; an extention of
	stats::xtabs() allowing to create sparse matrices.

2007-10-08  Martin Maechler  <>

	* DESCRIPTION (Version): *-4 (released *-3 two days ago)

2007-10-06  Martin Maechler  <>

	* R/pMatrix.R: solve(<Matrix>, <pMatrix>)

2007-10-05  Martin Maechler  <>

	* R/LU.R: solve() method for "denseLU"

2007-10-01  Martin Maechler  <>

	* DESCRIPTION (Version): 0.999375-3 preparing for release

	* R/AllGeneric.R: simplify  if(.) .. else .. for R <= 2.5.1

	* R/Matrix.R (Matrix): .Internal(matrix(..)) different for R >= 2.7.0

2007-09-26  Martin Maechler  <>

	* R/pMatrix.R (.m.mult.pMat): fix <matrix> %*% <pMatrix>,
	  thanks to Kasper Kristensen.
	* tests/matprod.R: regression test for that.

2007-09-23 17:32  Douglas Bates <>

	* [r4778] R/AllGeneric.R: Check R version before defining generics for

2007-09-13  Martin Maechler  <>

	* R/denseMatrix.R (rcond): method
	* R/sparseQR.R (rcond): method, use x or t(x)

2007-09-12  Martin Maechler  <>

	* R/dgeMatrix.R (rcond): method: work via qr.R() for non-square
	* R/sparseMatrix.R: Ditto for all other rcond() method definitions.

	* man/rcond.Rd: mention the more general definition, and add
	example for non-square matrices.

	* man/chol.Rd: new file, for the S4 chol() generic and all methods.

2007-09-11  Martin Maechler  <>

	* R/sparseQR.R: add qr.R() method [to be used for rcond()]

2007-09-01  Martin Maechler  <>

	* R/Matrix.R ([<-): add (Matrix,missing,ANY,Matrix) etc
	* tests/indexing.R: add new regression for the above cases.

2007-08-30  Martin Maechler  <>

	* src/Mutils.h (__sun): clause for alloca.h on Solaris

2007-08-16  Martin Maechler  <>

	* DESCRIPTION (Date, Version): 0.999375-2

2007-08-15  Martin Maechler  <>

	* R/HBMM.R (readMM): make work for pattern matrices as well

2007-08-14 13:07  Douglas Bates <>

	* [r4730] src/Mutils.h: declare alloca
	* [r4734] NAMESPACE, R/AllGeneric.R, R/Csparse.R, R/HBMM.R, R/Tsparse.R,
	  R/dgCMatrix.R, R/dgTMatrix.R, R/dsCMatrix.R, R/dsTMatrix.R,
	  man/externalFormats.Rd, src/Csparse.c, src/Csparse.h, src/DEPS.mkf,
	  src/HBMM.c, src/HBMM.h, src/Mutils.h, src/SOURCES_C.mkf, src/init.c,
	  src/mmio.c, src/mmio.h: Remove deprecated function writeHB and its
	  methods; switch writeMM to use CHOLMOD code; repair readMM

2007-08-14  Martin Maechler  <>

	* R/nearPD.R, man/nearPD.Rd (nearPD): new function built on Jens
	Oehlschlaegel's ... result type still to be discussed!

2007-08-10  Martin Maechler  <>

	* man/image-methods.Rd: new, combining all image() methods

2007-08-09  Martin Maechler  <>

	* R/dgCMatrix.R: define qr() and lu() methods for "sparseMatrix"
	to work via dgC...

	* R/Matrix.R (Matrix): special treatment for "table" (S3)

2007-08-07  Martin Maechler  <>

	* R/dgTMatrix.R (image): change defaults to
		aspect = "iso", colorkey = FALSE

2007-08-06  Martin Maechler  <>

	* src/dsyMatrix.c (dsyMatrix_matrix_mm): 'b' might be matrix;
	must copy its *expanded* x slot before LAPACK call.

	* tests/matprod.R: test the last days' changes.

2007-08-06 16:43  Douglas Bates <>

	* [r4712] inst/include/Matrix_stubs.c: Change R_cholmod_printf in stubs
	  as well as in sources
	* [r4713] src/dsyMatrix.c: Duplicate the contents of the RHS before
	  Lapack call

2007-08-03  Martin Maechler  <>

	* R/Matrix.R (%*%, crossprod, tcrossprod): add method for
		("Matrix", "matrix") which is needed in some cases.
	Ditto for solve().

	* R/colSums.R (.as.dge.Fun): need ddenseMatrix methods to avoid
	infinite recursion in dispatch for some cases.

2007-08-02 08:48  Martin Maechler <>

	* [r4693] src/chm_common.c: R_cholmod_printf() instead of Rprintf() just
	  so pointers match

2007-08-02  Martin Maechler  <>

	* DESCRIPTION (Date): set ready for release -- 0.999375-1

2007-08-01 15:44  Douglas Bates <>

	* [r4686] inst/include/Matrix.h, inst/include/Matrix_stubs.c,
	  src/chm_common.c, src/chm_common.h: Change API for
	  numeric_as_chm_dense and N_AS_CHM_DN

2007-08-01  Martin Maechler  <>

	* src/dtrMatrix.c (dtrMatrix_matrix_mm): fix dimensionality check (!)
	* tests/matprod.R: regr.test for it

2007-07-20  Martin Maechler  <>

	* R/dMatrix.R: fix from Brian for   Math2(., digits = "missing")
	* tests/group-methods.R: and regression-test it

2007-07-19 19:45  Douglas Bates <>

	* [r4642] inst/include/Matrix.h, inst/include/Matrix_stubs.c,
	  inst/include/cholmod.h, src/chm_common.c, src/init.c: Export
	  triplet_to_sparse, documentation, use typedefs

2007-07-18  Martin Maechler  <>

	* man/dpoMatrix-class.Rd: added 'corMatrix' example

	* src/dsyMatrix.[ch] (dsyMatrix_as_matrix): new 'keep_dimnames' arg
	* src/dtrMatrix.[ch] (dtrMatrix_as_matrix): ditto
	* src/init.c, R/dsyMatrix.R, R/dtrMatrix.R, R/lgTMatrix.R: ditto

	* R/lsparseMatrix.R: bug fix in "all" method
	* R/Ops.R (.do.Logic.lsparse): "|" bug in borderline case

	* R/dsyMatrix.R (coerce->dsTMatrix): (i,j) slots should *not* have names
	* R/ngTMatrix.R (coerce->ngTMatrix): ditto;
		+ matrix |-> nMatrix coercion
	* R/pMatrix.R: + setAs()  to dMatrix and from nMatrix
	* man/pMatrix-class.Rd: ditto

	* R/Matrix.R (Summary): method for non-dMatrix

	* tests/Class+Meth.R (extraValid): new check about "dirty" slots
	* tests/Class+Meth.R (tstMatrixClass): test norm(.);
		test all Summary methods.

2007-07-16  Martin Maechler  <>

	* R/dgeMatrix.R (norm, rcond): methods for 'matrix'

2007-07-14  Martin Maechler  <>

	* R/sparseMatrix.R (norm): simple methods for sparseMatrix

	* R/pMatrix.R (t?crossprod): methods for pMatrix

2007-07-10  Douglas Bates  <>

	* src/dgeMatrix.c (dgeMatrix_colsums): Get the logic straight.

2007-07-09 20:45  Douglas Bates <>

	* [r4579] src/dgeMatrix.c: Untangle horrible code in dgeMatrix_colsums
	  trying to fix a subtle bug - which has been somewhere else.

2007-07-09 19:43  Martin Maechler <>

	* [r4578] src/dgeMatrix.c: "cleaned" dgeMatrix_colsums() - but did not
	  solve the bug

2007-07-08  Martin Maechler  <>

	* src/dgCMatrix.c (compressed_to_TMatrix):

2007-07-07  Martin Maechler  <>

	* src/Csparse.c (Rsparse_validate): new, to have some validity
	checking for RsparseMatrix
	* src/dgCMatrix.c (xRMatrix_validate): ditto
	* src/dtCMatrix.c (tRMatrix_validate): ditto

2007-07-07  Douglas Bates <>

        * [r4567] R/AllClass.R: Slots already in RsparseMatrix were redefined in
        * [r4568] DESCRIPTION: Prepare for bug-fix release
        * [r4570] src/CHOLMOD/Check/cholmod_write.c: Include cholmod_matrixops.h
          for declaration of cholmod_symmetry

2007-07-06  Martin Maechler  <>

	* DESCRIPTION (Version): 0.999375 merged into the trunk; ready for

2007-07-06 14:11  Douglas Bates <>

	* [r4559] src/iohb.c, src/iohb.h: Remove Harwell-Boeing input/output
	  functions - no longer used
	* [r4560] src/HBMM.c, src/Mutils.c, src/dgTMatrix.c, src/dgeMatrix.c,
	  src/dspMatrix.c, src/dsyMatrix.c, src/dtCMatrix.c,
	  src/factorizations.c, src/sparseQR.c: Replace most calls to Calloc by

2007-07-06 13:14  Martin Maechler <>

	* [r4558] inst/doc/Comparisons.Rnw, src/CHMfactor.c, src/Csparse.c,
	  src/Tsparse.c, src/chm_common.c, src/chm_common.h, src/dense.c,
	  src/dgCMatrix.c, src/dsCMatrix.c, src/dtTMatrix.c, src/sparseQR.c,
	  src/t_gCMatrix_colSums.c: more R_CheckStack()s

2007-07-05 18:12  Douglas Bates <>

	* [r4550] inst/include/Matrix.h: Add the macro N_AS_CHM_DN to "alloca"
	  the required amount of memory then call M_numeric_as_chm_dense.
	* [r4556] src/Mutils.h, src/dgCMatrix.c: Define and use the Alloca macro
	  (like Calloc but calling alloca)

2007-07-05  Martin Maechler  <>

	* R/sparseMatrix.R (printSpMatrix): renamed from prSpMatrix()
	and extended with more sophisticated 'col.names' option.
	* NAMESPACE: export  printSparseMatrix()
	* man/printSpMatrix.Rd: document, incl. examples

2007-07-04 16:21  Douglas Bates <>

	* [r4543] src/cs_utils.c, src/cs_utils.h, src/dgCMatrix.c,
	  src/dtCMatrix.c, src/sparseQR.c: CSP typedef for *cs and macro
	  AS_CSP. API change - pass the empty structure to Matrix_as_cs.

2007-07-04  Martin Maechler  <>

	* DESCRIPTION (Version): 0.99875-4

	* tests/Class+Meth.R (tstMatrixClass): add dimnames, and hence test
	some dimnames perservation.

	* R/dsTMatrix.R (t-method): keep dimnames
	* R/dtTMatrix.R:  ditto

	* R/sparseMatrix.R (prSpMatrix): print colnames when non-trivial
	and ncol(.) < 10

	* src/cs_utils.c: drop check_class() and use Matrix_check_class()
	from Mutils.h

	* src/lgCMatrix.c ([ln]csc_to_matrix): no longer lose dimnames,
	e.g. in as(<ngCMatrix>, "matrix")

2007-07-01 13:27  Douglas Bates <>

	* [r4529] .: Create a branch for the API changes from the 0.99875
	  series to the 0.999375 series
	* [r4530] DESCRIPTION, inst/include/Matrix.h,
	  inst/include/Matrix_stubs.c, src/CHMfactor.c, src/Csparse.c,
	  src/Mutils.h, src/Tsparse.c, src/chm_common.c, src/chm_common.h,
	  src/dense.c, src/dgCMatrix.c, src/dsCMatrix.c, src/dtTMatrix.c,
	  src/t_gCMatrix_colSums.c: API change - pass the empty structure to the
	  as_cholmod_x functions

2007-06-30 09:05  Martin Maechler <>

	* [r4527] trunk/Matrix/DESCRIPTION, trunk/Matrix/NAMESPACE,
	  trunk/Matrix/inst/doc/Comparisons.Rnw: add session- and hardware-info
	  to Comparisons

>>>>>>> .merge-right.r4561
2007-06-29  Martin Maechler  <>

	* DESCRIPTION (Version): 0.99875-3 ready to be released.

	* R/sparseMatrix.R (spMatrix): make spMatrix(3,4) working

	* R/AllGeneric.R: set "Math" (and "Math2") group generics in a way
	that should also work in a future version of R.

2007-06-21  Martin Maechler  <>

	* NAMESPACE, R/AllClass.R: "xsparseVector" class union.

	* R/sparseVector.R: more *sparseVector coercions, notably for
	non - double ones.

2007-06-19  Martin Maechler  <>

	* R/colSums.R: new file for all (col|row)(Sums|Means) methods,
	  notably the new ones building on the new .Call(.)s:

	* src/dgCMatrix.c (DEF_gCMatrix_COLSUMS): use to define
	all 4 of [dlin]gCMatrix_colSums().

2007-06-18 16:12  Douglas Bates <>

	* [r4472] src/Syms.h, src/init.c: Added Matrix_lengthSym
	* [r4473] src/dgCMatrix.c: Modified dgCMatrix_colSums for sparseVector

2007-06-16  Martin Maechler  <>

	* R/kronecker.R: fix typo (could lead to inf.recursion)
	* test/simple.R: testing that

	* R/sparseMatrix.R (prSpMatrix): change to be used as print()
	method as well (which can have arguments, show() can't).

2007-06-16 15:52  Douglas Bates <>

	* [r4466] R/dgCMatrix.R, src/dgCMatrix.c, src/dgCMatrix.h, src/init.c:
	  added dgCMatrix_colSums for [col,row][Sums,Means]

2007-06-15 23:15  Douglas Bates <>

	* [r4460] R/sparseMatrix.R, man/dgCMatrix-class.Rd, src/dgCMatrix.c,
	  src/dgCMatrix.h, src/init.c: added (unexported),
	  coercion of "factor" to "dgCMatrix" and dgCMatrix_cholsol
	* [r4461] R/AllClass.R, man/sparseMatrix-class.Rd: draft "indicators"
	* [r4463] R/sparseMatrix.R, man/dgCMatrix-class.Rd,
	  man/sparseMatrix-class.Rd: Don't need an "indicators" class - use the
	  row names to store the levels - duh!	Added an example.

2007-06-14  Martin Maechler  <>

	* src/Csparse.c (Csparse_validate): check for *repeated* entries
	thanks to example from Christian Buchta; with a test here:
	* tests/simple.R:

2007-06-07  Martin Maechler  <>

	* R/Auxiliaries.R (callGeneric): another fix, needed for some cases
	of colSums(*, sparseResult = TRUE)

2007-06-06  Martin Maechler  <>

	* R/lsparseMatrix.R, R/ldenseMatrix.R (all, any): change default to
	'na.rm = FALSE' as "everywhere" else in R.

2007-06-05 Douglas Bates <>

	* [r4421] src/ Modify for new organization of
	  CSparse package
	* [r4425] src/ Update to version 3.0.0 of
	* [r4426] src/Makefile: add ./UFconfig to the include path for
	* [r4427] src/cs.[ch]: update to CSparse version 2.2.0
	* [r4428] inst/doc/UFsparse/* src/{AMD,CHOLMOD,COLAMD}/*
	  src/UFconfig/UFconfig.h: Update to version 3.0.0 of SuiteSparse

2007-06-05  Martin Maechler  <>

	* R/Auxiliaries.R (emptyColnames): + argument msg.if.not.empty, used in
	* R/sparseMatrix.R (prSpMatrix): now gives a message about
	suppressed column names.

2007-06-04 17:13  Douglas Bates <>

	* [r4418] src/Csparse.c, src/HBMM.c, src/Mutils.c, src/Mutils.h,
	  src/Tsparse.c, src/chm_common.c, src/chm_common.h, src/dgCMatrix.c,
	  src/dgeMatrix.c, src/dpoMatrix.c, src/dpoMatrix.h, src/dppMatrix.c,
	  src/dppMatrix.h, src/dsCMatrix.c, src/dspMatrix.c, src/dspMatrix.h,
	  src/dsyMatrix.c, src/dsyMatrix.h, src/dtpMatrix.c, src/dtrMatrix.c:
	  Remove warnings after change to const char* CHAR

2007-06-04 17:11  Douglas Bates <>

	* [r4417] inst/include/Matrix_stubs.c, inst/include/cholmod.h: Corrected
	  type of M_cholmod_ssmult

2007-06-03 14:42  Douglas Bates <>

	* [r4412] inst/include/Matrix_stubs.c, inst/include/cholmod.h,
	  src/init.c: Yet another cholmod export - cholmod_ssmult

2007-05-23  Martin Maechler  <>

	* NAMESPACE: exported drop0(), since
	* man/drop0.Rd: I have seen several cases, I really wanted to use
	it, so our users may want too.

2007-05-22  Martin Maechler  <>

	* man/colSums.Rd: separately document colSums() etc, since these
	have the extra argument 'sparseResult'.

2007-05-21  Martin Maechler  <>

	* R/sparseMatrix.R (spMatrix): utility (T)sparse Matrix constructor;
	* man/spMatrix.Rd: docu., including examples

	* R/Auxiliaries.R (sp.colMeans): etc, using a patched callGeneric(),
	in order to make colMeans() etc fast *and* correct.
	* R/sparseVector.R (replSPvec): "[<-" functionality for
	sparseVectors; tested in
	* tests/simple.R:

2007-05-19  Martin Maechler  <>

	* R/sparseMatrix.R (print.sparseSummary): and summary() method for
	(very) sparse Matrices; output similar to Matlab's print().

2007-05-17  Douglas Bates  <>

	* src/HBMM.c (Matrix_writeMatrixMarket): Write 1-based, not
	0-based, indices (Jose Quesada <>).

2007-05-16  Douglas Bates  <>

	* R/CHMfactor.R: Added solve methods for a CHMfactor object.

2007-05-16  Martin Maechler  <>

	* R/Auxiliaries.R (sparsapply): new utility, much faster than
	tapply1() for large sparse matrices.

2007-05-15  Martin Maechler  <>

	* R/Matrix.R (dim<-): reshape now via sparseVector.

	* R/sparseVector.R: methods and function for
	* R/AllClass.R: new "sparseVector" class and daughters.
	* NAMESPACE: export new classes

2007-05-14  Martin Maechler  <>

	* DESCRIPTION (Version): 0.99875-1
	* src/ also remove Lapack code from here (cf. 04-25).

2007-05-11  Martin Maechler  <>

	* R/Tsparse.R ([, Tsparse): fix last case: *duplicated*, symmetric
	* tests/indexing.R: test set for that.

2007-05-08  Martin Maechler  <>

	* R/Tsparse.R (replTmat): fix the case of *duplicated* index
	* tests/indexing.R(out): add regression test for it

2007-04-30  Martin Maechler  <>

	* R/(l(dense|sparse))?Matrix.R (!): use 'x', not 'e1' as argument
	name for "!" method definitions.

2007-04-26  Martin Maechler  <>

	* R/Tsparse.R (intI): new utility, used for "[" :
		Cleanup up there, and fixes for duplicated indices - more TODO!

	* tests/indexing.R(out): more tests

2007-04-25  Douglas Bates  <>

	* DESCRIPTION,src/Makefile: require R>= 2.5.0 and remove Lapack
	code that is now part of the R Lapack library.

	* src/init.c,inst/include/{Matrix_stubs.c,cholmod.h}:export
	cholmod_factorize_p (used in lme4 for GLMMs and NLMMs).

2007-04-21  Martin Maechler  <>

	* R/Matrix.R (image): method for all Matrices, not just sparse ones.

2007-04-17  Martin Maechler  <>

	* R/Auxiliaries.R (tapply1): unname(.) -> colSums() etc don't end
	up with extraneous names '0'...'<n-1>'

2007-04-12  Martin Maechler  <>

	* R/dgTMatrix.R (mat2dgT): care about NAs

2007-04-11  Martin Maechler  <>

	* R/kronecker.R: triangularity preserving methods

2007-03-27  Martin Maechler  <>

	* R/kronecker.R: new file collecting kronecker() methods in one
		place.  Goal: become much faster!

2007-03-23  Martin Maechler  <>

	* src/dtCMatrix.c (dtCMatrix_solve): use the new code from Kasper
	Kristensen based cs_spsolve() instead of _lsolve & _usolve which
	can be much faster.

	* tests/matprod.R: add regression tests for these (upper & lower).

2007-03-19  Martin Maechler  <>

	* R/Matrix.R (diff): method for our Matrices.

	* R/sparseMatrix.R (isDiagonal): check dim()!

2007-03-17  Martin Maechler  <>

	* R/Matrix.R (dim<-): new method for "reshape()" built on a
	proposal from Tamas Papp.

2007-03-16  Martin Maechler  <>

	* R/AllGeneric.R: remove all  if(!isGeneric(.))  clauses

	* R/zzz.R (.onLoad, .onUnload): do *not* leave bind_activation(TRUE);
	rather define and export  cBind() and rBind() only.
	--> useRs *must* change code that used to have cbind()/rbind() !!

	* R/bind.R: change tests from cbind() to cBind() and similarly to rBind()
	* R/ ditto

2007-02-16  Douglas Bates  <>

	* DESCRIPTION (Date, Version): 0.9975-11 with new date
	* src/dgCMatrix.c (R_to_CMatrix, compressed_to_TMatrix): remove
	const modifier on declaration of the array 'valid'

2007-02-12  Douglas Bates  <>

	* R/CHMfactor.R: Add image method (coercion to sparseMatrix).

2007-02-05  Martin Maechler  <>

	* DESCRIPTION (Date, Version): 0.9975-10 with new date.

	* R/Ops.R (Arith): make sure Csparse o Csparse also works for e.g. ntCMatrix
	* tests/simple.R: test the above and some of these coercions:

	* R/nsparseMatrix.R: coercing "Csparse" to "lsparseMatrix";
		be careful to avoid infinite recursion, using new coercions in
	* R/ngCMatrix.R and nsC... and ntC...
	* R/lsparseMatrix.R: ditto

	* R/SparseM-conv.R: more conversion, notably for triplet matrices.

	* src/dgCMatrix.c (R_to_C_Matrix): port Doug's fix and
	* R/Rsparse.R: reactivate .Call()s

	* tests/Class+Meth.R: a bit more on actual classes

2007-02-04  Douglas Bates  <>

	* src/dgCMatrix.c (compressed_to_TMatrix): fix memory bug using strdup()

2007-02-03  Martin Maechler  <>

	* DESCRIPTION (Version): 0.9975-10 to upload

	* tests/Class+Meth.R (tstMatrixClass): require coercions to
	specific classes less unconditionally.

	* R/Auxiliaries.R: get rid of as_Tsparse() and as_Rsparse()

	* R/Tsparse.R (triu): etc, use as(*, "TsparseMatrix") instead of as_Tsparse()

	* R/Rsparse.R (.R.2.T): R-level workaround using compressed_to_TMatrix.
	* R/Rsparse.R (.R.2.C): R-level workaround since C-level
	R_to_CMatrix segfaults on one platform.
	Eliminate most coercion method to *specific* classes, and replace
	with virtual classes coercions.

2007-02-01  Martin Maechler <>

	* src/init.c: export the CHM...._validate() placeholders, since
	they *are* called.

	* tests/Class+Meth.R (classCanCoerce): and starting to test
	all as(<from>, <to>)

2007-01-30  Martin Maechler  <>

	* R/Tsparse.R ([): more care when subsetting triangular Tsparse
	* tests/indexing.R: tested now
	* tests/ updated

	* src/Csparse.c (Csparse_to_dense): use Rkind = -1 for PATTERN to
	* src/chm_common.c (chm_dense_to_SEXP): return "ngeMatrix" when

	* NAMESPACE: export a trivial
	* R/Matrix.R: drop() Matrix-method

	* R/AllClass.R: moved all prototypes to virtual super classes.

	* R/Rsparse.R: many more coercions to have less exceptions in
	* tests/Class+Meth.R:

	* R/Ops.R (Compare): tweak for case with NA
	* tests/simpl.R: hence another 'FIXME' eliminated

2007-01-29  Martin Maechler  <>

	* R/diagMatrix.R (solve): the obvious methods for diagonalMatrix

	* tests/Class+Meth.R (tstMatrixClass): now testing diag(), nnzero(),
	  and more of "!", "&", "|", all, any; coercions

	* R/Rsparse.R: many coercions (which enable quite a few other
	methods), thanks to enhancements in
	* src/dgCMatrix.c (R_to_CMatrix): new, and
	* src/dgCMatrix.c (compressed_to_TMatrix): now for (d,l,n) ,
	symmetric & triangular and ..RMatrix objects.

	* src/TMatrix_as.c (Matrix_T_as_DENSE,Matrix_T_as_GENERAL):
	renamed file from src/dsTMatrix.c;
	now dealing with symmetric and triangular Tsparse coercions, both
	to dense and general.

2007-01-27  Martin Maechler  <>

	* src/dsTMatrix.c: has now "l" and "n" methods besides the "d" ones.

	* R/Ops.R (Arith): <dgCMatrix> o <numeric> now remains sparse
	where sensible when the <numeric> is of length > 1.

2007-01-26  Martin Maechler  <>

	* R/Matrix.R ([<-): for M[] <- value: fix length

2007-01-25  Martin Maechler  <>

	* R/Auxiliaries.R (n2l_Matrix): new, to be used in
	* R/ndenseMatrix.R: new coercions n* -> l*

2007-01-22  Martin Maechler  <>

	* R/triangularMatrix.R: new file; simple triu() and tril() methods.

	* R/Ops.R ("Logic"): and other "Ops", many updates

2007-01-18  Martin Maechler  <>

	* src/Mutils.h (SET_DimNames): new utility

	* R/Auxiliaries.R (nnzero): improved and now exported via

	* NAMESPACE: + nnzero(); length() == prod(dim(.)) method for all "Matrix" objects

2007-01-17  Martin Maechler  <>

	* R/diagMatrix.R (!): fix typo.

2007-01-16  Martin Maechler  <>

	* R/Auxiliaries.R (as_Csparse): and quite a few others:
		allow to pass class definition --> speedup
	* R/sparseMatrix.R: apply the above

	* R/Csparse.R: coercion Csparse* to dense* now preserves shape properties.

	* src/Mutils.h (mMatrix_as_geMatrix): new, based on
	* src/Mutils.c (dup_mMatrix_as_geMatrix): new; generalization of
	old dup_mMatrix_as_dgeMatrix), eliminating a long-standing "FIXME".

	* src/dense.c (dense_to_Csparse): use new mMatrix_as_geMatrix()

	* R/denseMatrix.R (.dense2C): based on dense_to_Csparse: name it,
	and use it for "sparse*" as well, since it's faster than the
	as_Csparse(.) way.

2007-01-15  Martin Maechler  <>

	* R/Ops.R ("Logic"): more methods, notably an  <lgC> o <lgC> one.

2007-01-12  Martin Maechler  <>

	* R/Tsparse.R (.TM.repl.i.2col): new internal function to be used
	as method for   M[ ij ] <- v
	* R/Csparse.R:: go via Tsparse for "M[ij] <- v"

	* R/Ops.R: "Compare" for (C|R)sparse: need pointer slot for all
	FALSE answer

	* R/Csparse.R (replCmat): fix the "all non-zero" case with reordering
	* tests/indexing.R: test it, and some of the above

2007-01-05  Martin Maechler  <>

	* R/Auxiliaries.R (is_duplicatedT): new utility

2007-01-05  Douglas Bates  <>

	* src/init.c (R_init_Matrix): export cholmod_scale

2006-12-30  Martin Maechler  <>

	* R/zzz.R (tmp): for R >= 2.5.0, extend formals of our
	base::as.matrix to (x, ...)

2006-12-28  Martin Maechler  <>

	* R/Ops.R ("Arith" etc): move almost all "Ops" methods to new R
	file; start using "Logic", hence

	* DESCRIPTION (Depends): R >= 2.4.1 (since we want "Logic")
	* NAMESPACE: import and export "Logic"

2006-12-27  Martin Maechler  <>

	* src/zpotfr.f and dependencies: use LAPACK 3.1 version
	only needed previously to R version 2.5.0.

2006-12-26  Martin Maechler  <>

	* DESCRIPTION (Date, Version): 0.9975-8, ready for release

	* R/Tsparse.R (replTmat): fix subassignment of triangular
	* R/Csparse.R (replCmat): ditto
	* tests/indexing.R: more tests, incl the above fix

2006-12-23  Martin Maechler  <>

	* R/Auxiliaries.R (drop0): extend for non CSparse
	* R/Auxiliaries.R (diagU2N): should work for all sparseMatrix

	* src/Csparse.c (Csparse_to_Tsparse, Csparse_general_to_symmetric):
		use uplo correctly (!); other places: use uplo_P() macro

	* R/Csparse.R (replCmat): call diagU2N() when needed
	* R/Tsparse.R (replTmat): ditto

	* src/dtCMatrix.c (tCMatrix_validate): new
	* src/dtTMatrix.c (tTMatrix_validate): new, used in
	* R/AllClass.R: for validity of dtC, dsC, and dtT, dsT.

	* R/diagMatrix.R (replDiag): to use in [<-

2006-12-22  Martin Maechler  <>

	* R/Auxiliaries.R (as_Csparse2, as_geSimpl): new functions;
		also more general diagU2N().

2006-12-21  Martin Maechler  <>

	* R/bind2.R: new file for all cbind2(), rbind() methods moved here
	from R/<foo>Matrix.R files. Better diagonal & improved sparse methods.

2006-12-20  Martin Maechler  <>

	* tests/bind.R: a few more cases
	* R/Auxiliaries.R (.M.kind): also work for atomic vectors

	* R/denseMatrix.R (cbind2/rbind2): moved here (and generalized) from
	* R/ddenseMatrix.R (cbind2/rbind2)
	* R/Tsparse.R (replTmat): final(?!) fix for "[<-" ..
	* tests/indexing.R
	* tests/ updated

2006-12-18  Martin Maechler  <>

	* R/Tsparse.R (replTmat): fixed a remaining "[<-" bug in
	* tests/indexing.R

2006-12-15  Martin Maechler  <>

	* R/sparseMatrix.R (prSpMatrix): "." alignment much improved:
	align with proper position of "0", i.e., right for integers.
	argument 'align' by default is "fancy".

2006-12-14  Martin Maechler  <>

	* R/sparseMatrix.R: delegate "Compare" to "Csparse.."
	* R/Csparse.R: and fix "Compare" for more cases.
	* tests/Class+Meth.R: test some of these (m == m, m != m)

2006-12-13  Martin Maechler  <>

	* R/lsparseMatrix.R: all() and any() methods
	* R/ldenseMatrix.R:	 ditto
	* NAMESPACE, R/Matrix.R: ditto

	* man/all-methods.Rd: document them minimally

	* tests/simple.R: add a few examples for these

2006-12-11  Martin Maechler  <>

	* R/Tsparse.R ([): fix long standing typo in symmetric case
	* man/dsCMatrix-class.Rd: add example exhibiting the above case

2006-12-10  Douglas Bates  <>

	* src/CHMfactor.c (CHMfactor_to_sparse): change LDL factorization
	to LL before converting to a sparse matrix. (The LDL form can be
	converted to a sparse matrix but it is implicitly a unit
	triangular matrix and a diagonal matrix overwritten on the diagonal.)

2006-12-09  Douglas Bates  <>

	* src/chm_common.c (chm_factor_to_SEXP): allocate and fill the Dim slot.

2006-12-08  Douglas Bates  <>

	* DESCRIPTION (Version): updated -> release 0.9975-7
	* src/{init.c,chm_common.c}, inst/include/*: export cholmod_analyze_p

2006-11-30  Martin Maechler  <>

	* R/diagMatrix.R (%*%): write a direct [diag o Csparse] method

2006-11-29  Douglas Bates  <>

	* src/dgeMatrix.c (dgeMatrix_solve): Check error code from dgetri.
	* tests/dg_Matrix.R: Add Barry Rowlingson's test of a matrix that
	is exactly singular.

2006-11-07  Martin Maechler  <>

	* DESCRIPTION (Date): updated -> release 0.9975-6

2006-11-06  Martin Maechler  <>

	* R/Csparse.R (replCmat): symmetric indexing of symmetric matrix
	now returns symmetric.

	* R/zzz.R ("diag<-"): replace "diag<-" in base for R <= 2.4.x

	* R/Matrix.R (.M.sub.i.2col):  new, for  M[ cbind(i,j) ] indexing.
	* R/Matrix.R (.M.repl.i.2col): new, for  M[ cbind(i,j) ] <- value

	* R/Auxiliaries.R (.type.kind): added

2006-11-04  Martin Maechler  <>

	* src/cs.[ch]: updated to CSparse Version 2.0.3 by simply
	running src/

	* R/denseMatrix.R: "[": keep symmetric on symmetric indexing.

2006-11-03  Martin Maechler  <>

	* src/dsCMatrix.c (dsCMatrix_Csparse_solve): new
	* R/dsCMatrix.R (solve): "fully-sparse" using the above.

	* R/AllClass.R: "pMatrix" now also inherits from "generalMatrix"

	* tests/Class+Meth.R (tstMatrixClass): now assure
	the (important in method programming) property :

		###>>  Every "Matrix" is either
		###>>  "general*", "symmetric*", "triangular*" or "diagonal*"

		(where "*" stands for "Matrix")

	* R/Auxiliaries.R (diagU2N): now .Call()s Csparse_diagU2N for <Csparse>

	* R/dMatrix.R (Compare(<dMatrix>,<dMatrix>): update and
	* tests/validObj.R: checks for "comparison"

	* R/sparseMatrix.R ([): improved indexing for sparse;
		trying to keep <symmetric>[ n, n] symmmetric

	* tests/indexing.R: indexing for logical sparse now ok

2006-11-02  Martin Maechler  <>

        * src/Tsparse.c: use xTsparse_validate() , and hence remove

        * src/{ltC,lsC,lgT}Matrix.[ch]: removed

2006-11-02  Martin Maechler  <>

	* R/AllClass.R (Matrix-class): check length of dimnames in validity.

	* tests/simple.R: validObject() checking the above.

	* src/dgCMatrix.c (xCMatrix_validate): new, small and simple,
	replacing both dgCMatrix_validate and lgCM*.

	* src/Csparse.c (Csparse_dense_prod, etc): do not lose dimnames;
	fix dimnames setting in other places.
	* src/chm_common.c (chm_dense_to_SEXP): now can pass dimnames

2006-11-01  Martin Maechler  <>

	* R/Csparse.R,src/Csparse.c, etc: tcrossprod(<CsparseM>,<CsparseM>)

	* R/sparseMatrix.R (isSymmetric): drop 'factors' slot for
	symmetry test, via
	* R/Auxiliaries.R (.as.dgC.0.factors):

2006-11-01  Douglas Bates  <>

	* R/Csparse.R,src/Csparse.c,tests/matprod.R,
	man/CsparseMatrix-class.Rd: crossprod(<CsparseMatrix>,
	<CsparseMatrix>) added

2006-10-30  Martin Maechler  <>

	* tests/matprod.R: add a variation of Harri's example

	* R/dsparseMatrix.R: fix crossprod(<dsparse>, <dge>) to *not*
	recursive infinitely.

	* R/dgCMatrix.R: + solve(<sparse>, <sparse>)

	* tests/indexing.R: add test for the "<" bug fixed 10-27 in R/dMatrix.R

2006-10-28  Martin Maechler  <>

	* tests/Class+Meth.R (tstMatrixClass): more: use non-trivial
	matrix if possible; test m+m == 2*m; now test dgRMatrix.
	* R/dgRMatrix.R (.to.dgR): a few more coercions, in order to
	satisfy the above test.

2006-10-27  Martin Maechler  <>

	* R/Matrix.R (Ops):  <Matrix> o <matrix>  method added

	* R/dgCMatrix.R: solve(a, b="missing") based on
	* src/dgCMatrix.c (dgCMatrix_matrix_solve): extend to work with

	* R/diagMatrix.R (diagdiagprod): extend %*% etc to ldiMatrix;
	add more (needed) [t]crossprod() methods.

	* man/ddiMatrix-class.Rd: more info, notably on 'diag'

	* R/Auxiliaries.R (as_CspClass): cleanup
	(drop0): internal utility for "Csparse_drop(*, 0)"
	(.bail.out.2): encourage active feedback

2006-10-26  Martin Maechler  <>

	* R/dMatrix.R(Compare): new(), then slots [no validity check]

	* src/Csparse.c (Csparse_validate): fixed (and more efficient in
	non-valid or 'sorted' case).

	* R/dsparseMatrix.R: add "chol" method.
	* R/ddenseMatrix.R: ditto

	* R/diagMatrix.R (Ops): group methods for  <diagonal> o <sparse>
	* R/diagMatrix.R (diag2T): simple utility used "higher level"
	coercion; deprecating direct lower level coercions.

	* R/*.R (seq): use seq_len() and seq_along() where possible.

2006-10-23  Martin Maechler  <>

	* DESCRIPTION (Version): 0.9975-5 ready for release

2006-10-20  Douglas Bates  <>

	* src/init.c (R_init_Matrix): export more cholmod CCallable functions.

2006-10-20  Martin Maechler  <>

	* R/AllClass.R (corMatrix): add 'validity' check;
		comment out unused "LDL" class definition
	* NAMESPACE: mention, but do not export "LDL" class
	* R/corMatrix.R: new (simple), needed for R-devel with
	* tests/Class+Meth.R (tstMatrixClass): 1 exception for corMatrix
	  coerce and t() exceptions for all 5 'Mat.MatFact' classes.

2006-10-19  Douglas Bates  <>

	* src/chm_common.h: Add R_cholmod_start to initialize cholmod to
	use Rprintf and R's error handling.

2006-10-17  Martin Maechler  <>

	* R/diagMatrix.R (%*%): rep(*, each = .) in Matrix %*% diagonal.
	* tests/matprod.R: add tests for the bug fixed.

2006-10-11  Douglas Bates  <>

	* src/HBMM.[ch]: remove HarwellBoeing format for writing.
	* src/SOURCES_C.mkf (SOURCES_C): no longer compile iohb.c

2006-10-06  Douglas Bates  <>

	* R/d[gs]CMatrix.R: deprecate the writeHB function.  Use writeMM instead.

2006-10-06  Martin Maechler  <>

	* DESCRIPTION (Version): 0.9975-3

	* R/diagMatrix.R (bdiag): new function constructing block diagonal
	(sparse) matrices.
	* man/bdiag.Rd: docu + examples

	* R/Csparse.R (replCmat): calling new Csparse_drop() now.

	* src/Csparse.c (Csparse_general_to_symmetric, Csparse_drop): new functions
	* R/lsCMatrix.R: three more coercions to lsC (thanks to the above)

	* R/diagMatrix.R (Diagonal): '[<-' method for diag.matrices such
	that result is sparse or diagonal (and not dense).

	* man/Subassign-methods.Rd: fix examples

	* R/Matrix.R (Matrix):  Matrix(0, *) or Matrix(*, sparse=TRUE)
	should always return a sparse (and not sometimes a diagonal) matrix.

2006-10-05  Martin Maechler  <>

	* R/Matrix.R ([<-): also for value "Matrix" or "matrix"

2006-10-04  Douglas Bates  <>

	* DESCRIPTION (Version): 0.9975-2

	* inst/include/Matrix_stubs.c (M_cholmod_sparse_to_triplet): export more symbols

2006-10-02  Douglas Bates  <>

	* tests/dg_Matrix.R: Simplify test taking into account new code.

2006-09-29  Martin Maechler  <>

	* R/Csparse.R (replCmat): improve for missing i / j in non-simple cases

	* R/lsTMatrix.R: new files w/ missing methods
	* R/nsTMatrix.R:  "  'for completeness'

	* tests/Class+Meth.R: a bit less 'not.ok.classes'

	* R/Tsparse.R (t): generalized "t" method from "dgT*" to "Tsparse*"

2006-09-28  Douglas Bates  <>

	* src/dppMatrix.h: Ensure definition of dspMatrix_validate is

	* src/init.c, inst/include/{Matrix.h,Matrix_stubs.h,cholmod.h}:
	Export C-callable functions used in Zt_create in lme4.

2006-09-28  Martin Maechler  <>

	* DESCRIPTION (Version): 0.9975-1

	* tests/simple.R: less checks fail; using NA, found that our
		kronecker() is not base-compatible with NA's.

	* R/dMatrix.R: "Compare" method now implemented for all cases

	* R/Auxiliaries.R (indTri): == which([lower/upper].tri( * )) new utility

	* man/dtpMatrix-class.Rd: mention length of 'x' slot

	* src/dtpMatrix.c (dtpMatrix_validate): fix check
	* src/dspMatrix.c (dspMatrix_validate): ditto

	* R/dtTMatrix.R (gt2tT): fix ("l" -> "n")
	  including coercion to [nl]tTMatrix.

	* R/diagMatrix.R (show): print a header line as for other classes.

2006-09-27  Martin Maechler  <>

	* src/ (SUBDIRS): fix typo

2006-09-19  Martin Maechler  <>

	* DESCRIPTION (Date): ready to release 0.9975-0 to CRAN

2006-09-18  Douglas Bates  <>

	* R/[CT]sparse.R (crossprod and tcrossprod): Handle the cases for x
	symmetric and y missing in R code using %*% (cholmod_aat doesn't
	accept a symmetric matrix).
	* tests/group-methods.R: Uncomment test of crossprod applied to
	lsCMatrix objects.

2006-09-18  Martin Maechler  <>

	* R/AllClass.R (symmetricMatrix): add validity method (available
	in C for a long time). Many "n..Matrix": drop (wrong) validity arg.

	* src/lgCMatrix.c (lgCMatrix_validate): check 'x' slot (!)

	* tests/
	* tests/indexing.R: additions, mainly for 'lsparse'

	* R/diagMatrix.R (Diagonal) & coercion to lgTMatrix: fixes for NA

	* R/Auxiliaries.R (nz.NA): new utility now used in nnzero()

2006-09-16  Martin Maechler  <>

	* R/sparseMatrix.R (prSpMatrix): print logical NAs "visibly"
		as 'N' (1-letter - Ok?)

	* tests/group-methods.R: add test for logical + NAs

	* R/dMatrix.R ("Compare"): fix to work with NA's

	* R/AllClass.R: "Cholesky" etc now inherit from MatrixFactorization.

	* src/lgCMatrix.c (ncsc_to_matrix): renamed from lcsc_to_matrix()
		which is implemented.

2006-09-15  Martin Maechler  <>

	* src/chm_common.c: coerce logical <-> double instead of
	typecasting;  needed for "l" matrix handling in cholmod.

	* tests/other-pkgs.R (graph): small extension in "graph" checks.

	* R/sparseMatrix.R (graphNEL -> Tsparse): method for weight case.
	(Tsp2grNEL): other fixes needed

2006-09-11  Martin Maechler  <>

	* R/AllClass.R ("nMatrix"): and subclasses for "nonzero pattern"
		Matrices, since "lMatrix", also "lsparseM" can have NA
	* R/ndenseMatrix.R, etc: new source files
	* man/nsparseMatrix-classes.Rd, etc: new help files
	* tests/: adaptions
	* src/chm_common.c (chm_dense_to_SEXP): and others:
		new 'Rkind' argument: "l*" and "d*" both use CHOLMOD_REAL
	* src/Csparse.c, etc: ditto

2006-09-11  Douglas Bates  <>

	* src/Mutils.[ch],init.c inst/include/*.h: Move the
	alloc_d**Matrix functions to the lme4 package.

2006-09-09  Douglas Bates  <>

	* src/dsCMatrix.c (dsCMatrix_Cholesky): igoring LDL = FALSE now

2006-09-09  Martin Maechler  <>

	* R/lMatrix.R: new

	* R/sparseMatrix.R (Tsp2grNEL): do not yet use graph::foo()

	* R/dgeMatrix.R: do not define tcrossprod() methods for "matrix"
	* man/tcrossprod.Rd: ditto

2006-09-08  Douglas Bates  <>

	* inst/include/Matrix_stubs.c,Matrix.h: Add declarations and stubs
	for exported functions
	* src/Makefile, src/CHOLMOD/Lib/Makefile, src/Metis,CAMD,CCOLAMD:
	Remove partitioning algorithms for sparse matrix reordering.  The
	copyright on the Metis code was problematic and the methods were
	rarely used.
	* src/triplet_to_col.[ch],MMHB.[ch]: Remove triplet_to_col.  Such
	operations are now done entirely in CHOLMOD code.

2006-09-06  Douglas Bates  <>

	* src/Mutils.h: Remove functions that are no longer used.

2006-09-04  Douglas Bates  <>

	* src/dtCMatrix.c (dtCMatrix_validate): rename functions.
	* src/DEPS.mkf: update

2006-09-02  Martin Maechler  <>

	* created branches/Matrix-for-R-2.3.x;  on trunk: do
	* DESCRIPTION (Version): 0.9975-0
	(Depends): R (>= 2.4.0)

2006-09-01  Douglas Bates  <>

	* R/sparseMatrix.R: Added direct method to CsparseMatrix from graphNEL

2006-09-01  Martin Maechler  <>

	* R/sparseMatrix.R: add coercion from "ANY" to "sparseMatrix"
	* R/denseMatrix.R:  add coercion from "ANY" to "denseMatrix"

	* R/Matrix.R ([): use nargs() to disambiguate M[i] and M[i,]

2006-08-31  Martin Maechler  <>

	* R/sparseMatrix.R (Arith): moved Arith group method one-level up
	from "dsparse" to "sparse" and now go via "Csparse" instead of "dgC"
	* R/dsparseMatrix.R: ditto
	* R/Csparse.R: ditto

2006-08-31  Martin Maechler  <>

	* R/dMatrix.R (Compare): improve availability of "<", etc
	* R/Auxiliaries.R (asTuniq): new; also make use of R 2.4.x print(*,max)

2006-08-30  Martin Maechler  <>

	* R/dgCMatrix.R: aargh: "Arith(<dgC>, numeric)" was wrong because
	of a 0-index which was used as 1-index (..hmm)

	* R/sparseMatrix.R (prSpMatrix): fix printing an all-0 sparse Matrix

	* R/Auxiliaries.R (all0, is0): for 0-testing in presence of NA's
	* R/Auxiliaries.R (isTriMat): use all0() for 0-testing.
	(.is.diagonal): ditto

	* R/lgTMatrix.R: as("matrix", "lgTMatrix"): warn about NA's

	* R/Matrix.R (Matrix): also work for NA data

2006-08-28  Martin Maechler  <>

	* R/Matrix.R (Matrix): + 'forceCheck' argument; dimnames setting
	in all cases.

2006-08-27  Douglas Bates  <>

	* src/dense.[ch],init.c, R/ddenseMatrix.R, man/band.Rd
	(ddense_band): Added triu, tril and band for ddenseMatrix objects.

2006-08-25  Martin Maechler  <>

	* src/Mutils.c (dup_mMatrix_as_dgeMatrix): added all subclasses of
	subclasses of "ddenseMatrix"
	* src/init.c et al: outcomment dtrM*_as_dge* and dsyM*_as_dge*_
	* R/ddenseMatrix.R et al: setAs("ddenseMatrix", "dgeMatrix", ..)
	instead of half dozen specialized ones.

2006-08-25  Douglas Bates  <>

	* R/lmer.R (qqmath method): Bug fix provided by Emmanuel Tillard
	<> - ordering standard errors to match effects.

2006-08-24  Douglas Bates  <>

	* src/lsCMatrix.c (lsCMatrix_trans): Remove lsCMatrix_chol based
	on R_ldl.
	* R/lCholCMatrix.R,
	src/[dl]CholCMatrix.[ch],R_ldl.[ch],Metis_utils.[ch]: removed
	* src/dsCMatrix.c (dsCMatrix_to_dgTMatrix): use CHOLMOD
	* many files in ./R and ./src: Use more general version of
	dup_mMatrix_as_dgeMatrix to simplify method definitions.
	* src/Mutils.c (dup_mMatrix_as_dgeMatrix): Add ddiMatrix,
	dtpMatrix, dspMatrix and dppMatrix conversions.

2006-08-23  Douglas Bates  <>

	* R/AllClass.R,lCholCMatrix.R,src/SOURCES_C.mkf,init.c,NAMESPACE:
	Remove classes lCholCMatrix and dCholCMatrix based on R_ldl code.
	* src/dgeMatrix.c: Ensure 'factors' slot exists in result of
	dgeMatrix_crossprod (may need to do this in other places).
	* R/AllGeneric.R,dsCMatrix.R, src/dscMatrix.[ch]: Add Cholesky
	generic and method for dsCMatrix.  Use CHOLMOD for chol().

2006-08-22  Douglas Bates  <>

	* src/Mutils.c (dup_mMatrix_as_dgeMatrix): updated for general
	types of classed Matrices, matrices or numeric or logical vectors.
	* src/init.c: register dup_mMatrix_as_dgeMatrix for coercions.
	* src/chm_common.c,Mutils.h (as_cholmod_factor): Move check_class
	to Mutils.h (as an inline) and change name to Matrix_check_class;
	fixes in as_cholmod_factor and chm_factor_to_SEXP.
	* src/dsCMatrix.[ch]: Use CHOLMOD for dsCMatrix_chol and
	dsCMatrix_matrix_solve.  Comment out vestigial functions.
	* src/Csparse.c: use diag_P and uplo_P macros.

2006-08-21  Douglas Bates  <>

	* src/lmer.c (internal_mer_RZXinv): Fix memory leak caught by
	* tests/matprod.R: Add tests to verify that 'solve' and '%*%' are
	* src/sparseQR.c (sparseQR_validate): add new validation test, fix
	-Wall warnings.
	* src/dppMatrix.c,dtrMatrix.c,dgCMatrix.c,dgeMatrix.c,dspMatrix.c:
	Use dup_mMatrix_as_dgeMatrix.

2006-08-20  Douglas Bates  <>

	* src/sparseQR.c: Fix thinko in sparseQR_resid_fitted.
	* tests/sparseQR.R: Added
	* man/sparseQR-class.Rd: Document methods for qr.* generics
	* R/sparseQR.R: Return correct # of coefs; fix cut-and-paste errors

2006-08-19  Douglas Bates  <>

	* NAMESPACE, R/sparseQR.R, src/init.c,sparseQR.[ch],SOURCES_C.mkf:
	Added methods for sparseQR for qr.qy, qr.qty, qr.coef, qr.resid
	and qr.fitted.
	* src/Mutils.[ch]: Added dup_mMatrix_as_dgeMatrix utility
	* src/dgCMatrix.c: Check for ordering before storing q in dgCMatrix_QR

2006-08-18  Martin Maechler  <>

	* R/AllGeneric.R: add "qr" and "chol" generics; via 'trick' since
	the base version has no "..." argument.

	* R/sparseMatrix.R (prSpMatrix): fix the triangular unit diagonal case.

	* R/Matrix.R: define and
	* NAMESPACE: export as.numeric() and as.logical() methods.
		Let's hope these do not badly slow down something...
	Finally export the zapsmall() method.

2006-08-17  Douglas Bates  <>

	* src/dgCMatrix.[ch] (dgCMatrix_matrix_solve), src/init.c,
	R/dgCMatrix.R: solve methods for dgCMatrix and dense RHS.
	* src/dtCMatrix.c :Remove code that is no longer used
	* R/dtCMatrix.R: Use C code for diagU2N in CsparseMatrix classes

2006-08-14  Douglas Bates  <>

	* src/Csparse.[ch],init.c (Csparse_to_logical): Added utilities
	Csparse_to_logical and Csparse_symmetric_to_general.
	* R/dgCMatrix.R,dsCMatrix.R,Csparse.R : Consolidate general
	coercion methods between CsparseMatrix and TsparseMatrix.

2006-08-14  Douglas Bates  <>

	* R/dtCMatrix.R,dgCMatrix.R,Csparse.R src/init.c,dgCMatrix.[ch],
	Csparse.[ch],chm_common.c,dense.c : Use CHOLMOD code and methods
	for CsparseMatrix, TsparseMatrix, denseMatrix or matrix when

2006-08-12  Douglas Bates  <>

	* src/chm_common.[ch],Csparse.[ch],Tsparse.[ch],CHMfactor.c,dtTMatrix.c,
	  dgTMatrix.c,dgCMatrix.c,dsCMatrix.c,dtTMatrix.c, R/dsTMatrix.R,
	  generalize conversions between TsparseMatrix, CsparseMatrix,
	  denseMatrix and matrix.  Preserve triangularity property and
	  propagate Dimnames in the easy cases.

2006-08-10  Douglas Bates  <>

	* src/lmer.c: adjust checks in glmer_init for S4SXP
	* tests/validObj.R: check of all(eq) should be all(eq@x) - worked
	before S4SXP but for the wrong reasons.

2006-08-08  Douglas Bates  <>

	* src/Csparse.c (Csparse_crossprod): tcrossprod result has stype = -1.
	Later modified to return the upper triangle only.
	* R/dgTMatrix.R: Remove vestigial crossprod and tcrossprod methods.

2006-08-07  Douglas Bates  <>

	* src/Csparse.c (Csparse_crossprod): Set stype on result to avoid
	R-level conversion to a sparse symmetric class.
	* R/Tsparse.R,R/Csparse.R (crossprod and tcrossprod methods):
	Remove conversion to sparse symmetric classes (now done in C code).

2006-08-07  Martin Maechler  <>

	* R/dgCMatrix.R: disable old crossprod and tcrossprod methods
	* man/band.Rd:  adapt the \dontshow{} unit test.

2006-08-07  Martin Maechler  <>

	* DESCRIPTION (Version): 0.995-14  for CRAN

	* tests/other-pkgs.R: oops: library(Matrix) outside *if* !

	* R/sparseMatrix.R (Tsp2grNEL): fixed Tsparse -> graph coercion
	* tests/other-pkgs.R:  more checks, sparseMatrix -> graph

	* R/Auxiliaries.R (as_Tsparse, as_Rsparse): new
	* R/Tsparse.R (tril, triu, band): methods
	* R/dgRMatrix.R (tril, ...):	ditto
	* man/band.Rd:			ditto

2006-08-04  Martin Maechler  <>

	* R/Matrix.R (head, tail): directly use utils:::head.matrix
		to be up-to-date automatically.

2006-08-03  Martin Maechler  <>

	* DESCRIPTION (Version): 0.995-13 to be released to CRAN

	* DESCRIPTION (Lazydata): no, instead of 'yes' because it fails for:
	* data/CAex.R, inst/external/CAex_slots.rda: replacing data/CAex.rda
	* data/KNex.R, inst/external/KNex_slots.rda: replacing data/KNex.rda
	such that the S4 objects are always created by the current version
	of R and 'Matrix' class definitions.

2006-08-01  Douglas Bates  <>

	* R/lmer.R (LMEoptimize method), tests/lmer.R,
	inst/external/test3comp.rda: Added warnings for convergence on
	boundary and test cases.

	* src/lmer.c (mer_postVar): Modified to return the variances that
	are marginal to the fixed effects, not conditional on them (which
	is what the bVar slot contents represent).

2006-07-31  Douglas Bates  <>

	* NAMESPACE, src/lmer.c, R/lmer.R (hatTrace):  Add the hatTrace
	function which calls the C function mer_hat_trace2.

	* man/ranef.Rd: Include description of "postVar" argument and
	producing a caterpillar plot.

2006-07-31  Martin Maechler  <>

	* NAMESPACE:    change  "correlation" to "corMatrix"
	* R/AllClass.R: to avoid clash with S3 class in 'nlme'.
	* R/dpoMatrix.R: ditto
	* R/lmer.R:	 ditto

2006-07-28  Douglas Bates  <>

	* src/lmer.c (internal_mer_RZXinv): Split the calculation of the
	RZXinv slot's contents into a separate internal function that can
	be used in mer_hat_trace.

2006-07-22  Martin Maechler  <>

	* R/Matrix.R: Coercions "Matrix" -> (sparse|dense)Matrix
	via new smart
	* R/Auxiliaries.R (as_Csparse, as_dense, .M.shapse): new
	utility functions.

2006-07-21  Martin Maechler  <>

	* R/Csparse.R (tril, triu, band): do return *triangular* classed
	matrices when appropriate; band() even symmetric ones.
	(replCmat): extend to potentially all "CsparseMatrix"
	* R/Tsparse.R (replTmat): extend to all "TsparseMatrix"; hence
	allow subassignment for special sparse matrices.

	* R/Auxiliaries.R (as_geClass): factor out the  .M.kind() functionality

	* src/lmer.c (mer_MCMCsamp, glmer_MCMCsamp): new 'verbose'
	argument; in glmer_*(): print only if(verbose).

	speed-optimize a few places by moving REAL(.) out of loops.

	* src/lmer.h, src/init.h, R/lmer.R:  related to above.

2006-07-20  Martin Maechler  <>

	* R/Matrix.R("["): disable ("Matrix", i = "logical", j = "missing"),
		since that wrongly triggers also for M[ logi , ]

	* R/denseMatrix.R: "[" methods now also work e.g. when indexing
	a symmetric matrix that results in a non-symmetric one.
	* R/Auxiliaries.R (as_geClass): new function used in "[" above.

	* R/dMatrix.R: make round(M) work as round(M, 0)

	* R/dgTMatrix.R (image): coordinate system and axis now use
	1-based indices, not 0-based ones.

	* R/Tsparse.R (.ind.prep for "["): get rid of max(<empty>) warning.
	* tests/indexing.R: test it.

	* NAMESPACE: export isSymmetric(); has been a generic in "base"	for a while;
	* man/isSymmetric-methods.Rd: and document it.

	* R/SparseM-conv.R: added coercion methods for some 'SparseM' matrices.
	* man/SparseM-conv.Rd: docu them

	* tests/other-pkgs.R: renamed from tests/graph.R and add example
		for 'SparseM' conversions

2006-07-17  Douglas Bates  <>

	* R/Matrix.R (head): added head() and tail() methods.

2006-07-17  Martin Maechler  <>

	* DESCRIPTION (Version): 0.995-12 released to CRAN

2006-07-15  Martin Maechler  <>

	* tests/simple.R: add check for correct dsT -> dgT coercion;
		add check for correct printing of symmetric sparse matrices.

	* R/Auxiliaries.R (non0ind): return *all* non-0 entry indices also
	for sparse symmetric matrices.

	* src/dsTMatrix.c (dsTMatrix_as_dgTMatrix): do not copy the
	diagonal twice.

2006-07-11  Douglas Bates  <>

	* src/dsTMatrix.c (dsTMatrix_as_dgTMatrix): Fix a case of INTEGER
	being applied to the x slot (detected by Brian Ripley).

2006-07-10  Martin Maechler  <>

	* src/dgCMatrix.c (dgCMatrix_validate): 'p' slot must have correct length.

	* R/Auxiliaries.R (isTriC): fix buglet (we were 1-based!)

2006-07-08  Martin Maechler  <>

	* src/lgCMatrix.c (lgCMatrix_diag): new function
	* R/lgCMatrix.R (diag): for new method

	* R/AllClass.R (TsparseMatrix): do use Tsparse_validate
	     ==> construction of illegal "*gTMatrix" via new() should
	     now be much less easy:
	* tests/Class+Meth.R: assertError() for some illegal "dgT*"

	* R/Matrix.R (Matrix): Matrix(0, nrow,ncol) now "goes sparse"
	* man/Matrix.Rd: documents it.

2006-07-06  Douglas Bates  <>

	* src/pedigree.c (pedigree_inbreeding): Correction in
	initialization.  This function is not currently being used and is
	not fully tested.

	* NAMESPACE, R/{AllClass.R,lmer.R}, src/{init.c,lmer.c}:
	Introduced the glmer class.  Added code for mcmcmsamp on glmer
	objects.  Modified validity check on pedigree objects to account
	for nonparallel patterns of missingness of parents.

	* man/{lmer-class.Rd,mcmcsamp.Rd}: Update documentation for glmer

2006-07-01  Martin Maechler  <>

	* R/pMatrix.R: coercion pMatrix -> Tsparse

2006-06-12  Douglas Bates  <>

	* DESCRIPTION (Version): 0.995-11 released to CRAN

	* R/lmer.R (mcmcsamp method): Corrected arrangments of names on
	the output from mcmcsamp.

2006-06-10  Douglas Bates  <>

	* R/lmer.R (simulestimate): added C code for calculating the trace
	of the hat matrix.

2006-06-09  Martin Maechler  <>

	* R/diagMatrix.R (setAs): define coercion methods to sparse matrix
	* R/sparseMatrix.R etc: multiplication of diagonal and sparse

2006-06-08  Martin Maechler  <>

	* R/dgTMatrix.R (colSums): etc. All four of
	colSums(), rowSums(), colMeans(), rowMeans() now should work for
	all "Matrices".

2006-06-01  Douglas Bates  <>

	* R/lmer.R ( Add a reference line at zero and a
	background grid to the qqmath plot of ranef.lmer.

2006-05-30  Douglas Bates  <>

	* R/lmer.R (expandSlash): Functions (non-exported) to allow nested
	grouping factors to be specified in the formula as (1|foo/bar).

2006-05-27  Douglas Bates  <>

	* R/lmer.R (findbars and others): Change check of is.numeric to
	!is.language to resolve the bug reported by Jacob Wegelin.

	* src/pedigree.c (pedigree_inbreeding): Initial implementation of
	code to evaluate inbreeding coefficients without calculating T,
	based on code in Sargolzaei and Iwaisaki's paper.

2006-05-27  Douglas Bates  <>

	* R/{lmer.R,AllGeneric.R}, src/{init.c,lmer.[ch]}: Added local
	generic and mer methods for isNested and denomDF. This denomDF was
	an attempt to emulate that in lme but I don't think that makes
	sense.  Use the trace of the hat matrix instead.

2006-05-17  Martin Maechler  <>

	* R/sparseMatrix.R: Matrix <-> graph methods: can no longer use
	the C code depending on a slot structure that's no longer valid.
	* src/dgTMatrix.c: ditto (also: src/init.c src/dgTMatrix.h)

2006-05-17  Douglas Bates  <>

	*  R/{AllGeneric.R,Csparse.R},man/band.Rd,NAMESPACE: changed name
	of lowerTriMatrix generic and methods to tril (also upper to triu)
	and added a general band extractor.

2006-05-16  Douglas Bates  <>

	* R/pedigree.R (pedigree): Replace sire and dam values outside the
	allowable range with NAs. Added a corresponding check in the
	validity check for the pedigree class.

	* R/[CT]sparse.R ([t]crossprod): The result of single-argument
	crossprod methods now inherits from symmetricMatrix.

2006-05-15  Douglas Bates  <>

	* R/AllGeneric.R (lowerTriMatrix): Added (but did not export)
	generics lowerTriMatrix and upperTriMatrix along with methods for
	the Csparse virtual class.  Also added a C function Csparse_band
	that implements these methods by calling cholmod_band.

2006-05-15  Martin Maechler  <>

	* R/Tsparse.R ("["): column or row subsetting; @Dimnames[k] got
	erased when it was NULL.  This led to invalid subselections!

2006-04-25  Douglas Bates  <>

	* R/dtCMatrix.R: avoid coercion of dtCMatrix object to dgCMatrix
	in method for "t" so as not to lose the unit diagonal property.

2006-04-19  Douglas Bates  <>

	* R/lmer.R, R/AllGeneric.R, NAMESPACE: Remove the postVar generic
	and methods.  This is now an option to the ranef method for the
	mer class.

	* src/cs_utils.c: Ensure that the nz component is -1 for a
	compressed column-oriented matrix.  Minor formatting cleanup.

	* man/lmer-class.Rd: Document the qqmath method for ranef.lmer

2006-04-19  Martin Maechler  <>

	* R/Auxiliaries.R (diagU2N): new for the solve() methods in
	* R/dtCMatrix.R: where dgC -> dgT coercion now preserves diag = "U".

2006-04-15  Douglas Bates  <>

	* src/cs.[ch],src/cs_utils.[ch] : Added Tim Davis' CSparse library
	in cs.[ch] and utilities to interface to that code in cs_utils.[ch].

	* R/dtCMatrix.R, src/dtCMatrix.[ch] : CSparse-based solve methods
	for the dtCMatrix class.

2006-04-12  Douglas Bates  <>

	* R/pedigree.R, R/AllClass.R, NAMESPACE: added a pedigree class
	and methods for it.

2006-04-12  Martin Maechler  <>

	* R/dgCMatrix.R: add  storage.mode(.) <- "double" for "matrix"
	arguments, such that M %*% 1:6 now works
	* Tests/matprod.R: test the above

2006-04-03  Douglas Bates  <>

	* R/lmer.R (qqmath,ranef.lmer-method): added a qqmath method for
	the ranef.lmer class.

	* R/AllClass.R, NAMESPACE, R/lmer.R: Added a postVar generic and
	methods to extract the posterior variances from the bVar slot.

2006-03-30  Martin Maechler  <>

	* R/dtCMatrix.R: allow coercion from  dgC* to triangular (dtC*)
	* R/dsCMatrix.R: and symmetric (dsC*) 'Csparse' matrices.

	* R/Tsparse.R: Tsparse* -> Csparse* coercion now works and is tested in
	* tests/simple: (extended)

	* R/sparseMatrix.R (isTriangular): now using much improved
	* R/Auxiliaries.R (isTriC): new triangularity check for *CMatrix

2006-03-23  Douglas Bates  <>

	* src/dsyMatrix.c (dsyMatrix_as_dspMatrix): Propagate DimNames
	(problem report from Franklin Parlamis).  This should be done

2006-03-21  Douglas Bates  <>

	* R/AllClass.R,lmer.R: Change the name of the lmer.ranef class to
	ranef.lmer (like summary.lmer).  Add the coef.lmer class and
	update the plot methods.

2006-03-20  Douglas Bates  <>

	* R/lmer.R (resid and ranef methods):  Added methods for the
	"residuals" and "resid" generic but only for linear mixed model
	fits.  Changed the ranef method to return a list of data frames so
	that the plot methods now work.

2006-03-16  Douglas Bates  <bates@bates2>

	* src/dpoMatrix.c (dpoMatrix_chol): Require n > 0 in call to
	dpotrf - otherwise the BLAS on Mac OS X complains and quits.

	* DESCRIPTION (Date): New release

2006-03-15  Martin Maechler  <>

	* DESCRIPTION (Version): 0.995-6  -- to be released to CRAN

	* data/KNex.rda: replacing 'mm' and 'y' by KNex <- list(mm=mm, y=y)
	* man/KNex.Rd: and other help files
	* tests/*.R: several ones needed adaption
	* inst/doc/Comparisons.Rnw: ditto

2006-03-11  Martin Maechler  <>

	* R/dgCMatrix.R (replCmat): "[<-" methods for dgCMatrix

	* tests/indexing.R: tests for new [<- methods for sparse matrices.

2006-03-10  Martin Maechler  <>

	* R/dgTMatrix.R (replTmat): "[<-" methods for dgTMatrix

	* R/Tsparse.R (.ind.prep): fix out-of-range indexing

2006-03-08  Martin Maechler  <>

	* R/dMatrix.R: enable things like  M [ M < 10 ]
	* R/dgeMatrix.R: implement it

2006-03-06  Martin Maechler  <>

	* R/AllClass.R: define "summary.mer" and "*.lmer"
	* R/lmer.R (summary): summary(<mer>) computes & returns the above;
	* R/lmer.R (show): now works with summary()

2006-03-04  Martin Maechler  <>

	* R/dgCMatrix.R: finally direct "Arith"  dgC o dgC
	* R/Auxiliaries.R (WhichintersectInd): and other utilities to
		support the above

2006-02-07  Douglas Bates  <>

	* R/lmer.R (lmer): fix initial values of offset and weights for
	Use glmFit$prior.weights for weights in a glmm.
	Allow an option usePQL = FALSE to skip the PQL steps for the
	Laplace method (and, in time, the AGQ method).

	* src/lmer.c (mer_factor): Move downdating and factoring of XtX
	into a separate function internal_mer_Xfactor to be able to call
	it from internal_bhat.

2006-01-23  Martin Maechler  <>

	* tests/Class+Meth.R (tstMatrixClass): function for much better
	testing; now again of all actual classes.

		use macros and define make_d_matrix_* and make_i_matrix_*
		where _i_ is for the ldense routines:

	* src/ldense.c (ltrMatrix_as_lgeMatrix): provide functions
	* src/ldense.c (lsyMatrix_as_lgeMatrix):

	* R/ldenseMatrix.R: use the above in setAs(*,"lgeMatrix")

2006-01-16  Martin Maechler  <>

	* R/Matrix.R (Matrix): has become much "smarter" now auto-producing
		many different kinds of matrices.
	* R/*.R: quite a few new methods were needed for R CMD check with
		new Matrix().  Very good for users playing around.

2006-01-15  Martin Maechler  <>

	* src/dgeMatrix.c (dMatrix_validate): new
	* src/Mutils.c (dense_nonpacked_validate): new

	* src/dtrMatrix.c (dtrMatrix_validate): improved/fixed

2006-01-14  Douglas Bates  <>

	* R/AllClass.R (compMatrix), (generalMatrix): new virtual classes

2006-01-07  Douglas Bates  <>

	* DESCRIPTION (Version): 0.99-6 released to CRAN

	* src/dgBCMatrix.c (cscb_trcbsm): Fix due to Peter Dalgaard for
	segfault in cases with multiple non-nested grouping factors.

2006-01-03  Martin Maechler  <>

	* DESCRIPTION (Version): 0.99-4 to be released to CRAN
		(Depends): also on 'utils'

	* R/AllClass.R (diagonalMatrix): new class with "ddi*" and "ldi*"
	* R/diagMatrix.R (Diagonal): constructor and methods for
	diagonal matrices

	* R/ltTMatrix.R: new "minimal methods"

2005-12-12  Martin Maechler  <>

	* R/AllGeneric.R (tcrossprod): 2-argument version; here, and for
	all methods (and help files).

2005-12-09  Martin Maechler  <>

	* R/Auxiliaries.R (dimNamesCheck): fixed thinko -> bug
		for case (dimn.	 op  no_dimn.)

2005-11-14  Douglas Bates  <>

	* DESCRIPTION (Version): 0.99-2	 released to CRAN

2005-10-21  Douglas Bates  <>

	* R/lmer.R (simulate method): Fixed a drop=FALSE problem reported
	by Julian Faraway.

2005-10-06  Martin Maechler  <>

	* R/Auxiliaries.R (try_as): new utility
	* R/sparseMatrix.R: use try_as() in coercion to original class

2005-09-30  Martin Maechler  <>

	* src/dgCMatrix.c (double_to_csc): and dgeMatrix_to_csc()

2005-09-29  Martin Maechler  <>

	* R/Auxiliaries.R (dimNamesCheck): added

	* R/Matrix.R (as.array), (as.vector): new

2005-09-28  Martin Maechler  <>

	* R/Matrix.R (Matrix): get logical argument 'sparse' with a smart
	* R/AllClass.R: move 'factors' slot toplevel "Matrix";
		"pMatrix" now contains "sparseMatrix"

2005-09-26  Martin Maechler  <>

	* tests/Class+Meth.R: new tests; t(t(m)) == m

	* src/dtCMatrix.c (tsc_transpose): add forgotten "diag" slot

	* src/dsTMatrix.c (dsTMatrix_as_dsCMatrix): bad typo (segfault!)
	* src/dtTMatrix.c (dtTMatrix_as_dtCMatrix): new

	* R/dspMatrix.R: typo in "t" method

2005-09-18  Douglas Bates  <>

	* R/AllClass.R (TsparseMatrix), CsparseM* and RsparseM*
	* R/Tsparse.R: instead of R/gTMatrix.R
	* R/Csparse.R: new
	* src/Tsparse.c (Tsparse_to_Csparse): new; -> cholmod_()
	* src/Tsparse.c: new; many trivial methods calling cholmod_()
	* src/Csparse.c (Csparse_to_Tsparse), transpose, (mat|cross)prod:
			via cholmod

2005-09-16  Martin Maechler  <>

	* R/Auxiliaries.R (non0ind): new function using new C code
	* src/dgCMatrix.c (compressed_non_0_ij): new utility

2005-09-15  Douglas Bates  <bates@localhost.localdomain>

	* src/chm_common.h: header file required by all C sources that
	call CHOLMOD functions.	 It defines a cholmod_common structure
	called 'c' whose address is passed as the last argument to
	(virtually) every CHOLMOD function.

	* src/Pattern.c: Simple example of the use of CHOLMOD.

	* src/init.c: initialize and finalize the cholmod_common structure.

	* src/Makefile: Added source packages CHOLMOD UMFPACK AMD COLAMD
	LDL and CCOLAMD from the U. of Florida sparse matrix library.

2005-09-08  Martin Maechler  <>

	* inst/test-tools.R: new file collecting the utility functions used
		in ./tests/*.R

	* R/ddenseMatrix.R (cbind2): new methods for "numeric" and "matrix"
	* R/Matrix.R (cbind2): methods for NULL and missing

2005-08-31  Martin Maechler  <>

	* R/AllClass.R: new "index" class for "[" and "[<-":
		First cut at "symmetricMatrix" and "triangularMatrix"
	* R/gTMatrix.R (.ind.prep): new function;
		Logical and character indexing now work too.
	* R/Matrix.R: cheap "[<-" methods for denseMatrix now work
	* tests/indexing.R: new, including *
	* tests/dgTMatrix.R: new

2005-08-29  Douglas Bates  <>

	* src/dgTMatrix.c (graphNEL_as_dgTMatrix): Corrected the position
	indicator pos not being updated.  Also enforced upper triangular
	for symmetric case.  Need to coerce edges component of elements of
	edge list - grr! (Why don't they define their classes cleanly?)

2005-08-26  Martin Maechler  <>

	* R/Matrix.R:  added first "[<-" methods; not yet functional
	* R/denseMatrix.R: ditto
	* man/Subassign-methods.Rd: new help file for these

2005-08-25  Martin Maechler  <>

	* DESCRIPTION (Version): 0.98-6

	* R/denseMatrix.R: new file for "[" fallback methods for all dense

2005-08-19  Martin Maechler  <>

	* src/lgCMatrix.c (lcsc_to_matrix): new; need for
		as( <logical>, "matrix")

	* R/pMatrix.R: coercion to "lgTMatrix"

	* R/gTMatrix.R: new virtual class to define "[" methods for.
	* man/gTMatrix-class.Rd:
	* General slight re-organization of where "[" methods are defined.
	  more to come.

2005-08-18  Douglas Bates  <>

	* DESCRIPTION (Version): 0.98-5	 released to CRAN

	* src/dgTMatrix.c (graphNEL_as_dgTMatrix): add first graphNEL methods
	* .....

2005-08-18  Douglas Bates  <bates@localhost.localdomain>

	* R/lmer.R: Corrected naming scheme in mcmcsamp to work with lmer
	or glmer objects.

2005-08-17  Martin Maechler  <>

	* DESCRIPTION (Version): 0.98-4 : upload to CRAN

2005-08-16  Douglas Bates  <bates@localhost.localdomain>

	* R/HBMM.R: finish re-writing R-only code.

2005-08-15  Douglas Bates  <bates@localhost.localdomain>

	* man/externalFormats.Rd: move documentation for writeHB and
	writeMM here.

	* src/mmio.c: replace inclusion of <malloc.h> by <stdlib.h>
	(suggested by Paul Roecker).

	* tests/validObj.R (assertError): Comment out test that is failing
	after recent changes in r-devel.

2005-08-11  Martin Maechler  <>

	* R/AllClass.R: intermediate virtual class "denseMatrix"
	* man/denseMatrix-class.Rd
	* NAMESPACE: export it, and also export
	* man/unused-classes.Rd: "iMatrix", "zMatrix" and "ldenseMatrix"

2005-08-10  Douglas Bates  <>

	* DESCRIPTION (Version): 0.98-3 to CRAN

	* src/dtrMatrix.c (dtrMatrix_validate): fixed up validation and
	matrix_solve code (which was really, really wrong).

2005-08-07  Douglas Bates  <>

	* DESCRIPTION (Version): 0.98-2

	* R/HBMM.R (readHB), (readMM): read Matrix Market formats

	* R/lmer.R (abbrvNms): new
	* R/lmer.R (mcmcsamp): allow transformed parameters
	* src/HBMM.c (Matrix_writeMatrixMarket): Added read/write routines
	for the Harwell-Boeing and the MatrixMarket formats.

2005-08-04  Martin Maechler  <>

	* man/dtrMatrix-class.Rd: add examples
	* man/dtpMatrix-class.Rd: ditto; plus note about PROBLEM

	* TODO: note the dtpMatrix (docu) bug

	* R/zzz.R (.onLoad): assignInNamespace("as.matrix", *, "base")
	in order to ensure that new as.matrix() is used by old functions,
	e.g., svd(), qr(), eigen(), dist(),..;
	      apply(), also matplot() or pairs().

2005-08-03  Martin Maechler  <>

	* R/lmer.R: add 'fixme' comments and move the linear vs glm check;
	  add comments about 'control' / lmerControl() arguments

2005-07-27  Douglas Bates  <>

	* man/sleepstudy.Rd: Added the sleep data set.
	* DESCRIPTION (Version): 0.98-1	 released to CRAN

2005-07-12  Douglas Bates  <>
	* man/sleepstudy.Rd: Added the sleep data set.

	* R/lmer.R (glmmMCMC): Added PACKAGE = "Matrix" in a couple of
	.Call calls that were producing spurious output.

2005-07-05  Douglas Bates  <>

	* R/lmer.R (lmer): stored updated variance component estimates in
	mer object for the generalized model.  (Bug reported by Renaud

2005-07-03  Douglas Bates  <>

	* src/lmer.c (glmer_devAGQ): Added AGQ for single grouping factor,
	unidimensional case.

2005-06-08  Douglas Bates  <>

	* DESCRIPTION (Version): 0.96-1

	* moved	 lmer-class' R and C code moved from lme4 to here

2005-06-04  Douglas Bates  <>

	* R/dgCMatrix.R: Call to csc_matrix_mm used undefined arguments
	(reported by Guissepe Ragusa <>)

2005-06-02  Douglas Bates  <>

	* src/ Forgot to update this when Makefile changed.

2005-05-11  Douglas Bates  <>

	* src/dgCMatrix.c (csc_transpose): Simplified function fixing a
	bug reported by Kurt Hornik and Michael Hahsler.

2005-05-10  Douglas Bates  <>

	* src/lgCMatrix.c (Matrix_lgClgCmm): Implementation of methods for
	logical sparse matrices.  These will also be used in the symbolic
	analysis for lmer objects.

	* src/dsCMatrix.c (dsCMatrix_matrix_solve): Copied the dimensions
	of b to the result.  Fixes bug reported by

2005-05-06  Douglas Bates  <>

	* src/dgeMatrix.c (dgeMatrix_colsums): Added an implementation of
	colMeans, colSums, rowMeans and rowSums.

2005-04-18  Douglas Bates  <>

	* src/lgCMatrix.[ch]: code for _validate method and stub for
	multiplication operation.

	* src/dgeMatrix.c (dgeMatrix_matrix_solve): Passing wrong argument
	to dgetrs.

	* src/init.c: Fix cut-and-paste error in definition of

	* src/{many files}: Tighten code by using ALLOC_SLOT.

2005-04-15  Douglas Bates  <>

	* R/AllClass.R: Add lgTMatrix and lgCMatrix classes

	* DESCRIPTION: Eliminate import of stats.

2005-04-06  Douglas Bates  <>

	* R/AllClass.R : add logical sparse matrix classes

2005-04-01  Martin Maechler  <>

	* R/dgTMatrix.R: add "[" method for triplet matrices
	* R/sparseMatrix.R: and other sparse ones; --> add show() for sparse

2005-03-31  Douglas Bates  <>

	* DESCRIPTION (Version): release 0.95-5 to CRAN

	* R/dMatrix.R: add %*%, crossprod and solve "fallback" methods
	* R/sparseMatrix.R: %*%, crossprod()
	* R/dgeMatrix.R: more "fallback" methods for numeric/dense matrices
	* man/*.Rd: move method definitions to 'Matrix' and 'dMatrix'
	* src/lmer.c (lmer_fitted): fix thinko

2005-03-26  Martin Maechler  <>

	* R/AllClass.R: add two virtual sparse classes ``on top''

2005-03-24  Martin Maechler  <>

	* R/AllClass.R (setClass): use "VIRTUAL" for the virtual classes;
		correspondingly fix examples and tests/ since new()
		doesn't work for virtual classes.

2005-03-17  Martin Maechler  <>

	* R/Matrix.R (as.matrix): method and one for unname()
	* tests/dpoMatrix.R: tests should now be less platform dependent;
	  also run for R 2.1.0; using as.matrix()

2005-03-15  Douglas Bates  <>

	* R/pMatrix.R: "pMatrix" class added
	* ....

2005-03-14  Douglas Bates  <>

	* R/dtpMatrix.R: Add unpack method and an example.

	* src/dsyMatrix.c (dsyMatrix_trf): Add BunchKaufman factorization
	of general symmetric matrices and associated S4 methods.

2005-03-10  Martin Maechler  <>
2005-03-05  Martin Maechler  <>

	* R/dgeMatrix.R (setAs): and many other files: more coercion,
	  crossprod() and "%*%" methods added; tests, too.
	* tests/matprod.R: new, for testing these

2005-03-03  Douglas Bates  <>

	* src/lmer.c (lmer_fitted): Added.

2005-03-02  Douglas Bates  <>

	* R/dsTMatrix.R: Conversion from dsTMatrix to dsCMatrix

2005-02-28  Douglas Bates  <>

	* src/*.c,po/,inst/po: Internationalization and localization of
	the package.

	* src/ldl.[ch]: Removed these as their contents are referenced in the
	R_ldl.c file.

	* src/flame.[ch]: Removed these source files.

	* src/dtrMatrix.c (make_array_triangular): Move to Mutils

	* src/LU.[ch],src/init.c: absorb in factorizations

	* src/Mutils.h: prepare for internationalization

	* src/cblas.h: move the enum definitions to Mutils.h and remove
	this file

2005-02-26  Martin Maechler  <>

	* R/dgeMatrix.R: provide "dimnames" and "dimnames<-" methods

	* R/dtrMatrix.R: fix t() method

	* R/dgeMatrix.R: define group methods "Arith", "Math", "Math2"
	* NAMESPACE: export them (and import generics from "methods")
	* tests/group-methods.R : and test them.

	* src/dtrMatrix.c (dtrMatrix_as_dgeMatrix): prevent seg.fault in
	border case

2005-02-24  Douglas Bates  <>

	* DESCRIPTION (Version): 0.95-2	 released to CRAN

	* src/dgBCMatrix.c:
	* src/lmer.c: many changes
	* ...

2005-02-04  Martin Maechler  <>

	* R/Matrix.R: add more sophisticated show() method.

2005-02-02  Douglas Bates  <>

	* */* : almost complete reorganization of classes.

2005-01-26  Douglas Bates  <>

	* R/AllGeneric.R: Added matrix exponential generic expm and a method
	for the geMatrix class.

2005-01-24  Douglas Bates  <>

	* src/Makefile (clean): Remove *.a and *.so

	* man/cscBlocked-class.Rd: Remove reference to the lmer-class.

2005-01-23  Douglas Bates  <>

	* src/lmer.c (Lind): Definition of Lind was backwards.	This only
	had an effect in cases with more than 2 grouping factors.

2005-01-03  Douglas Bates  <>

	* src/lmeRep.c (lmer_variances): change from lmeRep to lmer

2004-12-23  Douglas Bates  <>

	* src/init.c (R_init_Matrix): Reorder calls to R_registerRoutines
	and R_useDynamicSymbols (suggested by B.D.Ripley).

2004-12-14  Douglas Bates  <>

	* R/sscMatrix.R: Add determinant methods

	* src/triplet.[ch],src/init.c (triplet_to_matrix): Add a coercion
	for tripletMatrix to matrix.

2004-12-13  Douglas Bates  <>

	* R/AllClass.R (.onLoad): Eliminate the bbCrosstab class, which is
	no longer used.

	* src/R_ldl.c: Created an R-specific version of the ldl.[ch] files
	with dynamic allocation of scratch arrays.

	* src/ssclme.c (ssclme_copy_ctab): Fixed bug in creation of ZtZ
	for multivariate random effects with multiple grouping factors.
	Fixes part but not all of #15.

2004-12-03  Douglas Bates  <>

	* src/lmeRep.c (lmeRep_factor): order of operations for multiple
	scalar grouping factors corrected.

2004-11-29  Douglas Bates  <>

	* src/bCrosstab.c: remove diag_update which is no longer used

2004-11-16  Douglas Bates  <>

	* src/Metis_utils.c: Move metis.h include to C sources so that the
	.h file can be included.

2004-11-12  Douglas Bates  <>

	* src/LU.c,geMatrix.c,trMatrix.c, etc.:
	Complete allocation of slots in NEW_OBJECT.

	* src/Mutils.h: Moved list of symbols to an include file

2004-11-11  Douglas Bates  <>

	* src/geMutils.c (Matrix_init): remove unused function

2004-11-10  Douglas Bates  <>

	* src/cscMatrix.c (csc_to_imagemat): removed unused function

2004-11-05  Douglas Bates  <>

	* src/ (SOURCES_C): Keep consistent with Makefile

2004-10-27  Douglas Bates  <>

	* R/pdmatrix.R: remove PACKAGE="Matrix" in .Call calls

2004-10-04  Douglas Bates  <>

	* src/init.c: Created R_init_Matrix and added registration of C

2004-10-02  Douglas Bates  <>

	* R/tripletMatrix.R: Force a require(lattice) for the image methods.

2004-06-15  Douglas Bates  <>

	* man/trMatrix-class.Rd: Escape the % chars in .Rd files.

2004-04-20  Douglas Bates  <>

	* src/ ($(SHLIB)): Modifications per Uwe Ligges.

2004-04-19  Douglas Bates  <>

	* src/ssclme.c (ssclme_update_mm): fix logic error in

2004-04-18  Douglas Bates  <>

	* src/ssclme.c (ssclme_coef, ssclme_coefGets): Create consistency
	in the order of unconstrained and constrained parameters.
	(ssclme_gradient): Added the gradients (not yet correct for
	multidimensional, unconstrained case).

2004-04-14  Douglas Bates  <>

	* src/ssclme.c (ssclme_EMsteps): Fix logic in REML update

	* src/ Remove unneeded ranlib call

2004-04-12  Douglas Bates  <>

	* DESCRIPTION (Version): New release

	* src/ Update to umfpack removal.

2004-04-05  Douglas Bates  <bates@bates2_home>

	* src/triplet_to_col.c: Create triplet_to_col as a native
	function, not the version from umfpack.	 There were problems with
	the configuration of UMFPACK for 64-bit processors and there was
	only one umfpack routine being used so I moved it here.

2004-04-04  Douglas Bates  <bates@bates2_home>

	* src/ssclme.c (ssclme_variances): New function.

2004-03-28  Douglas Bates  <bates@bates2_home>

	* src/ssclme.c (ssclme_fitted): Added function.

2004-03-27  Douglas Bates  <bates@bates2_home>

	* src/ssclme.c (ssclme_transfer_dimnames): Add new function to
	store the dimnames in the XtX and bVar slots
	(ssclme_update_mm): Change the dimensions of the bVar slot
	components and the returned value from ssclme_ranef.

2004-03-18  Douglas Bates  <>

	* R/{pdMat.R,pdIdent.R,pdLogChol.R,pdMatrixLog.R,pdNatural.R},
	Moved the pdMat classes from the lme4 package.

2004-03-02  Douglas Bates  <>

	* man/ssclme-class.Rd: Update definition and documentation of the
	ssclme class to include the DIsqrt slot.

	* src/ssclme.c (ssclme_deviance): Modify order of computation
	(much faster using dsyrk, a level 3 BLAS routine).

	* src/Makefile (SUBLIBS): Change definition (K. Hornik)

2004-02-28  Douglas Bates  <>

	* tests/ssclme.R: Modify the test to account for the permutation
	of the levels of the grouping factors.

2004-02-23  Douglas Bates  <>

	* R/ssclme.R,src/ssclme.c (ssclme): Move slots of sscCrosstab slot
	directly into the ssclme class definition.

2004-02-22  Douglas Bates  <>

	* DESCRIPTION (Date): New release

	* man/ssclme-class.Rd: new file.

	* src/ssclme.c (ssclme_loglik): major revisions in design.  It
	works and it's fast!

2004-02-17  Douglas Bates  <>

	* src/taucs/ (lib): Change "ar" to "$(AR)" (B.Ripley)

2004-02-16  Douglas Bates  <>

	* DESCRIPTION (Date): New release

	* NAMESPACE: Don't export ssclme.

	* data/ScotsSec.rda, man/ScotsSec.Rd: Add Scottish secondary
	school data.

2004-02-11  Douglas Bates  <>

	* src/sscCrosstab.c (sscCrosstab): Added a row to the incidence to
	keep track of the fixed-effects and the response.  Counts also
	gets an extra element, which is always one.

	* src/ldl.c: Include these routines from Tim Davis' LDL package.

2004-02-10  Douglas Bates  <>

	* src/cscMatrix.c (csc_transpose): new function

	* src/Mutils.c (csc_sort_columns): perm/iperm confusion corrected
	(csc_components_transpose): new function

2004-02-06  Douglas Bates  <>

	* src/triplet.c (triplet_validate): Fix Dim slot on generated

2004-01-30  Douglas Bates  <>

	* R/sscCrosstab.R (sscCrosstab): Added sscCrosstab generator function.

	* src/LU.h (MATRIX_LU_H): Add #ifndef #define ... #endif to this
	and all other .h files in src.

	* src/ This and other files contributed
	by Brian Ripley.

2004-01-27  Douglas Bates  <>

	* R/syMatrix.R: Added methods for "%*%".

	* R/Hilbert.R (Hilbert): Changed Hilbert function to return a
	poMatrix object.

2004-01-26  Douglas Bates  <>

	* man/sscChol-class.Rd,man/mm.Rd,man/y.Rd: Added man pages.

2004-01-25  Douglas Bates  <bates@bates2_home>

	* inst/doc/Introduction.Rnw,Comparisons.Rnw: Added vignettes.

	* R/csc.R: Convert all cscMatrix classes to use Dim slot instead
	of nrow.

2003-12-31  Douglas Bates  <>

	* src/taucs/taucs.h: Moved taucs.h, amd.h, and umfpack.h into

2003-12-08  Douglas Bates  <>

	* src/taucs.h: Accidently referred to global header files
	instead of local files.

2003-12-04  Douglas Bates  <>

	* R/AllClass.R: Lots of changes.  Removed all the lapack++ code
	and methods and replaced all classes with S4 classes.

2003-04-19  Douglas Bates  <>

	* R/det.R,man/det.Rd: Change name of det generic to determinant

	* src/ Change method of calculating determinants

2003-02-03  Douglas Bates  <>

	* DESCRIPTION (Version): removed empty data directory as requested
	by CRAN maintainers.  Changed version number and date.

2002-10-23  Douglas Bates  <>

	* src/laindex.h: Applied patches from Brian Ripley for compilation
	under Windows.

	* Added and src/ as requested by Brian

2002-05-03  Douglas Bates  <>

	* src/lamatrix.h: Removing pre-1.2.0 compatibility code per Kurt

2002-04-24  Douglas Bates  <>

	* Replaced with
	contributed by Kurt Hornik.

	* aclocal.m4 (ac_clean_files): Replaced this with Kurt Hornik's
	version for R-1.5.0

2001-12-10  Douglas Bates  <>

	* man/eigen.Rd: Removed the .Alias in the example
ViewVC Help
Powered by ViewVC 1.0.0  
Thanks to:
Vienna University of Economics and Business Powered By FusionForge