[R] Matrix library error: "should never happen; please report"

[Martin Maechler]

Hi Jose,

>>>>> "JQ" == Jose Quesada <" <quesada at gmail.com>>
>>>>>     on Tue, 19 Jun 2007 21:12:53 +0200 writes:

    JQ> Hi, I got the following error. Sorry but this time I
    JQ> couldn't reproduce it with a simple chunk of code:

>> .TM.repl.i.2col(): drop 'matrix' case ...
>> Error in .nextMethod(x = x, i = i, j = j) :
>>          'i' has no integer column number should never happen; please report
>> In addition: Warning messages:
>> 1: Ambiguous method selection for "%*%", target "ddiMatrix#dgCMatrix" (the  
>> first of the signatures shown will be used)
>>      diagonalMatrix#CsparseMatrix
>>      ddenseMatrix#CsparseMatrix
>>   in: .findInheritedMethods(classes, fdef, mtable)
>> 

    JQ> I got 4 other copies of the same warning. Will play
    JQ> around a bit more...  This is really strange.

Yes, but

- the Matrix library is the file Matrix.so or Matrix.dll which
 is part of the installed (aka "binary") Matrix *package*
 Maybe you really need to read the result of
    fortune("package.*Maechler")  # after installing package 'fortunes'

- "please report" was not meant to say to report to R-help,
  but to the package maintainers,

- since you cannot reproduce it yet, we cannot do much about it.
  It may be a bug in the Matrix package (and Jose has told me
  that he's using the latest released version 0.99875-2),
  but in theory it could even be your own mistake, namely by
  wrongly manipulating the slots of a Matrix object.

Please try to produce an R script - even if not small -- with a
reproducible example;
[and then do report to  Matrix-authors at r-project.org].

    JQ> Thanks -- Jose Quesada, PhD.

Best regards,
Martin Maechler, ETH Zurich