NAME

       PDL::CCS::Ufunc - Ufuncs for compressed storage sparse PDLs

SYNOPSIS

        use PDL;
        use PDL::CCS::Ufunc;

        ##---------------------------------------------------------------------
        ## ... stuff happens

FUNCTIONS

   ccs_accum_prod
         Signature: (
           indx ixIn(Ndims,NnzIn);
           nzvalsIn(NnzIn);
           missing();
           indx N();
           indx [o]ixOut(Ndims,NnzOut);
              [o]nzvalsOut(NnzOut);
           indx [o]nOut();
           )

       Accumulated product over values $nzvalsIn() associated with vector-valued keys $ixIn().
       On return, $ixOut() holds the unique non-"missing" values of $ixIn(), $nzvalsOut() holds
       the associated values, and $nOut() stores the number of unique non-missing values
       computed.

       If $N() is specified and greater than zero, then the quantity:

        $missing ** ($N - (rlevec($ixIn))[0])

       is multiplied into $nzvalsOut: this is probably What You Want if you are computing the
       product over a virtual dimension in a sparse index-encoded PDL (see PDL::CCS::Nd for a
       wrapper class).

       Returned PDLs are implicitly sliced such that NnzOut==$nOut().

       In scalar context, returns only $nzvalsOut().

       ccs_accum_prod processes bad values.  The state of the bad-value flag of the output
       ndarrays is unknown.

   ccs_accum_dprod
         Signature: (
           indx ixIn(Ndims,NnzIn);
           nzvalsIn(NnzIn);
           missing();
           indx N();
           indx [o]ixOut(Ndims,NnzOut);
           double [o]nzvalsOut(NnzOut);
           indx [o]nOut();
           )

       Accumulated double-precision product over values $nzvalsIn() associated with vector-valued
       keys $ixIn().  On return, $ixOut() holds the unique non-"missing" values of $ixIn(),
       $nzvalsOut() holds the associated values, and $nOut() stores the number of unique non-
       missing values computed.

       If $N() is specified and greater than zero, then the quantity:

        $missing ** ($N - (rlevec($ixIn))[0])

       is multiplied into $nzvalsOut: this is probably What You Want if you are computing the
       product over a virtual dimension in a sparse index-encoded PDL (see PDL::CCS::Nd for a
       wrapper class).

       Returned PDLs are implicitly sliced such that NnzOut==$nOut().

       In scalar context, returns only $nzvalsOut().

       ccs_accum_dprod processes bad values.  The state of the bad-value flag of the output
       ndarrays is unknown.

   ccs_accum_sum
         Signature: (
           indx ixIn(Ndims,NnzIn);
           nzvalsIn(NnzIn);
           missing();
           indx N();
           indx [o]ixOut(Ndims,NnzOut);
              [o]nzvalsOut(NnzOut);
           indx [o]nOut();
           )

       Accumulated sum over values $nzvalsIn() associated with vector-valued keys $ixIn().  On
       return, $ixOut() holds the unique non-"missing" values of $ixIn(), $nzvalsOut() holds the
       associated values, and $nOut() stores the number of unique non-missing values computed.

       If $N() is specified and greater than zero, then the quantity:

        $missing * ($N - (rlevec($ixIn))[0])

       is added to $nzvalsOut: this is probably What You Want if you are summing over a virtual
       dimension in a sparse index-encoded PDL (see PDL::CCS::Nd for a wrapper class).

       Returned PDLs are implicitly sliced such that NnzOut==$nOut().

       In scalar context, returns only $nzvalsOut().

       ccs_accum_sum processes bad values.  The state of the bad-value flag of the output
       ndarrays is unknown.

   ccs_accum_dsum
         Signature: (
           indx ixIn(Ndims,NnzIn);
           nzvalsIn(NnzIn);
           missing();
           indx N();
           indx [o]ixOut(Ndims,NnzOut);
           double [o]nzvalsOut(NnzOut);
           indx [o]nOut();
           )

       Accumulated double-precision sum over values $nzvalsIn() associated with vector-valued
       keys $ixIn().  On return, $ixOut() holds the unique non-"missing" values of $ixIn(),
       $nzvalsOut() holds the associated values, and $nOut() stores the number of unique non-
       missing values computed.

       If $N() is specified and greater than zero, then the quantity:

        $missing * ($N - (rlevec($ixIn))[0])

       is added to $nzvalsOut: this is probably What You Want if you are summing over a virtual
       dimension in a sparse index-encoded PDL (see PDL::CCS::Nd for a wrapper class).

       Returned PDLs are implicitly sliced such that NnzOut==$nOut().

       In scalar context, returns only $nzvalsOut().

       ccs_accum_dsum processes bad values.  The state of the bad-value flag of the output
       ndarrays is unknown.

   ccs_accum_or
         Signature: (
           indx ixIn(Ndims,NnzIn);
           nzvalsIn(NnzIn);
           missing();
           indx N();
           indx [o]ixOut(Ndims,NnzOut);
              [o]nzvalsOut(NnzOut);
           indx [o]nOut();
           )

       Accumulated logical "or" over values $nzvalsIn() associated with vector-valued keys
       $ixIn().  On return, $ixOut() holds the unique non-"missing" values of $ixIn(),
       $nzvalsOut() holds the associated values, and $nOut() stores the number of unique non-
       missing values computed.

       If $N() is specified and greater than zero, $missing() is logically (or)ed into each
       result value at each output index with a run length of less than $N() in $ixIn().  This is
       probably What You Want.

       Returned PDLs are implicitly sliced such that NnzOut==$nOut().

       In scalar context, returns only $nzvalsOut().

       ccs_accum_or processes bad values.  The state of the bad-value flag of the output ndarrays
       is unknown.

   ccs_accum_and
         Signature: (
           indx ixIn(Ndims,NnzIn);
           nzvalsIn(NnzIn);
           missing();
           indx N();
           indx [o]ixOut(Ndims,NnzOut);
              [o]nzvalsOut(NnzOut);
           indx [o]nOut();
           )

       Accumulated logical "and" over values $nzvalsIn() associated with vector-valued keys
       $ixIn().  On return, $ixOut() holds the unique non-"missing" values of $ixIn(),
       $nzvalsOut() holds the associated values, and $nOut() stores the number of unique non-
       missing values computed.

       If $N() is specified and greater than zero, $missing() is logically (and)ed into each
       result value at each output index with a run length of less than $N() in $ixIn().  This is
       probably What You Want.

       Returned PDLs are implicitly sliced such that NnzOut==$nOut().

       In scalar context, returns only $nzvalsOut().

       ccs_accum_and processes bad values.  The state of the bad-value flag of the output
       ndarrays is unknown.

   ccs_accum_bor
         Signature: (
           indx ixIn(Ndims,NnzIn);
           nzvalsIn(NnzIn);
           missing();
           indx N();
           indx [o]ixOut(Ndims,NnzOut);
              [o]nzvalsOut(NnzOut);
           indx [o]nOut();
           )

       Accumulated bitwise "or" over values $nzvalsIn() associated with vector-valued keys
       $ixIn().  On return, $ixOut() holds the unique non-"missing" values of $ixIn(),
       $nzvalsOut() holds the associated values, and $nOut() stores the number of unique non-
       missing values computed.

       If $N() is specified and greater than zero, $missing() is bitwise (or)ed into each result
       value at each output index with a run length of less than $N() in $ixIn().  This is
       probably What You Want.

       Returned PDLs are implicitly sliced such that NnzOut==$nOut().

       In scalar context, returns only $nzvalsOut().

       ccs_accum_bor processes bad values.  The state of the bad-value flag of the output
       ndarrays is unknown.

   ccs_accum_band
         Signature: (
           indx ixIn(Ndims,NnzIn);
           nzvalsIn(NnzIn);
           missing();
           indx N();
           indx [o]ixOut(Ndims,NnzOut);
              [o]nzvalsOut(NnzOut);
           indx [o]nOut();
           )

       Accumulated bitwise "and" over values $nzvalsIn() associated with vector-valued keys
       $ixIn().  On return, $ixOut() holds the unique non-"missing" values of $ixIn(),
       $nzvalsOut() holds the associated values, and $nOut() stores the number of unique non-
       missing values computed.

       If $N() is specified and greater than zero, $missing() is bitwise (and)ed into each result
       value at each output index with a run length of less than $N() in $ixIn().  This is
       probably What You Want.

       Returned PDLs are implicitly sliced such that NnzOut==$nOut().

       In scalar context, returns only $nzvalsOut().

       ccs_accum_band processes bad values.  The state of the bad-value flag of the output
       ndarrays is unknown.

   ccs_accum_maximum
         Signature: (
           indx ixIn(Ndims,NnzIn);
           nzvalsIn(NnzIn);
           missing();
           indx N();
           indx [o]ixOut(Ndims,NnzOut);
              [o]nzvalsOut(NnzOut);
           indx [o]nOut();
           )

       Accumulated maximum over values $nzvalsIn() associated with vector-valued keys $ixIn().
       On return, $ixOut() holds the unique non-"missing" values of $ixIn(), $nzvalsOut() holds
       the associated values, and $nOut() stores the number of unique non-missing values
       computed.

       If $N() is specified and greater than zero, and if $missing() is greater than any listed
       value for a vector key with a run-length of less than $N(), then $missing() is used as the
       output value for that key.  This is probably What You Want.

       Returned PDLs are implicitly sliced such that NnzOut==$nOut().

       In scalar context, returns only $nzvalsOut().

       ccs_accum_maximum processes bad values.  The state of the bad-value flag of the output
       ndarrays is unknown.

   ccs_accum_minimum
         Signature: (
           indx ixIn(Ndims,NnzIn);
           nzvalsIn(NnzIn);
           missing();
           indx N();
           indx [o]ixOut(Ndims,NnzOut);
              [o]nzvalsOut(NnzOut);
           indx [o]nOut();
           )

       Accumulated minimum over values $nzvalsIn() associated with vector-valued keys $ixIn().
       On return, $ixOut() holds the unique non-"missing" values of $ixIn(), $nzvalsOut() holds
       the associated values, and $nOut() stores the number of unique non-missing values
       computed.

       If $N() is specified and greater than zero, and if $missing() is less than any listed
       value for a vector key with a run-length of less than $N(), then $missing() is used as the
       output value for that key.  This is probably What You Want.

       Returned PDLs are implicitly sliced such that NnzOut==$nOut().

       In scalar context, returns only $nzvalsOut().

       ccs_accum_minimum processes bad values.  The state of the bad-value flag of the output
       ndarrays is unknown.

   ccs_accum_maximum_nz_ind
         Signature: (
           indx ixIn(Ndims,NnzIn);
           nzvalsIn(NnzIn);
           missing();
           indx N();
           indx [o]ixOut(Ndims,NnzOut);
           indx [o]nzvalsOut(NnzOut);
           indx [o]nOut();
           )

       Accumulated maximum_nz_ind over values $nzvalsIn() associated with vector-valued keys
       $ixIn().  On return, $ixOut() holds the unique non-"missing" values of $ixIn(),
       $nzvalsOut() holds the associated values, and $nOut() stores the number of unique non-
       missing values computed.

       Output indices index $nzvalsIn, -1 indicates that the missing value is maximal.

       Returned PDLs are implicitly sliced such that NnzOut==$nOut().

       In scalar context, returns only $nzvalsOut().

       ccs_accum_maximum_nz_ind processes bad values.  The state of the bad-value flag of the
       output ndarrays is unknown.

   ccs_accum_minimum_nz_ind
         Signature: (
           indx ixIn(Ndims,NnzIn);
           nzvalsIn(NnzIn);
           missing();
           indx N();
           indx [o]ixOut(Ndims,NnzOut);
           indx [o]nzvalsOut(NnzOut);
           indx [o]nOut();
           )

       Accumulated minimum_nz_ind over values $nzvalsIn() associated with vector-valued keys
       $ixIn().  On return, $ixOut() holds the unique non-"missing" values of $ixIn(),
       $nzvalsOut() holds the associated values, and $nOut() stores the number of unique non-
       missing values computed.

       Output indices index $nzvalsIn, -1 indicates that the missing value is minimal.

       Returned PDLs are implicitly sliced such that NnzOut==$nOut().

       In scalar context, returns only $nzvalsOut().

       ccs_accum_minimum_nz_ind processes bad values.  The state of the bad-value flag of the
       output ndarrays is unknown.

   ccs_accum_nbad
         Signature: (
           indx ixIn(Ndims,NnzIn);
           nzvalsIn(NnzIn);
           missing();
           indx N();
           indx [o]ixOut(Ndims,NnzOut);
           int+ [o]nzvalsOut(NnzOut);
           indx [o]nOut();
           )

       Accumulated number of bad values over values $nzvalsIn() associated with vector-valued
       keys $ixIn().  On return, $ixOut() holds the unique non-"missing" values of $ixIn(),
       $nzvalsOut() holds the associated values, and $nOut() stores the number of unique non-
       missing values computed.

       Should handle missing values appropriately.

       Returned PDLs are implicitly sliced such that NnzOut==$nOut().

       In scalar context, returns only $nzvalsOut().

       ccs_accum_nbad processes bad values.  The state of the bad-value flag of the output
       ndarrays is unknown.

   ccs_accum_ngood
         Signature: (
           indx ixIn(Ndims,NnzIn);
           nzvalsIn(NnzIn);
           missing();
           indx N();
           indx [o]ixOut(Ndims,NnzOut);
           indx [o]nzvalsOut(NnzOut);
           indx [o]nOut();
           )

       Accumulated number of good values over values $nzvalsIn() associated with vector-valued
       keys $ixIn().  On return, $ixOut() holds the unique non-"missing" values of $ixIn(),
       $nzvalsOut() holds the associated values, and $nOut() stores the number of unique non-
       missing values computed.

       Should handle missing values appropriately.

       Returned PDLs are implicitly sliced such that NnzOut==$nOut().

       In scalar context, returns only $nzvalsOut().

       ccs_accum_ngood processes bad values.  The state of the bad-value flag of the output
       ndarrays is unknown.

   ccs_accum_nnz
         Signature: (
           indx ixIn(Ndims,NnzIn);
           nzvalsIn(NnzIn);
           missing();
           indx N();
           indx [o]ixOut(Ndims,NnzOut);
           indx [o]nzvalsOut(NnzOut);
           indx [o]nOut();
           )

       Accumulated number of non-zero values over values $nzvalsIn() associated with vector-
       valued keys $ixIn().  On return, $ixOut() holds the unique non-"missing" values of
       $ixIn(), $nzvalsOut() holds the associated values, and $nOut() stores the number of unique
       non-missing values computed.

       Should handle missing values appropriately.

       Returned PDLs are implicitly sliced such that NnzOut==$nOut().

       In scalar context, returns only $nzvalsOut().

       ccs_accum_nnz processes bad values.  The state of the bad-value flag of the output
       ndarrays is unknown.

   ccs_accum_average
         Signature: (
           indx ixIn(Ndims,NnzIn);
           nzvalsIn(NnzIn);
           missing();
           indx N();
           indx [o]ixOut(Ndims,NnzOut);
           float+ [o]nzvalsOut(NnzOut);
           indx [o]nOut();
           )

       Accumulated average over values $nzvalsIn() associated with vector-valued keys $ixIn().
       On return, $ixOut() holds the unique non-"missing" values of $ixIn(), $nzvalsOut() holds
       the associated values, and $nOut() stores the number of unique non-missing values
       computed.

       If $N() is specified and greater than zero, then the quantity:

        $missing * ($N - (rlevec($ixIn))[0]) / $N

       is added to $nzvalsOut: this is probably What You Want if you are averaging over a virtual
       dimension in a sparse index-encoded PDL (see PDL::CCS::Nd for a wrapper class).

       Returned PDLs are implicitly sliced such that NnzOut==$nOut().

       In scalar context, returns only $nzvalsOut().

       ccs_accum_average processes bad values.  The state of the bad-value flag of the output
       ndarrays is unknown.

TODO / NOT YET IMPLEMENTED

       extrema indices
           maximum_ind, minimum_ind: not quite compatible...

       statistical aggregates
           daverage, medover, oddmedover, pctover, ...

       cumulative functions
           cumusumover, cumuprodover, ...

       other stuff
           zcover, intover, minmaximum

ACKNOWLEDGEMENTS

       Perl by Larry Wall.

       PDL by Karl Glazebrook, Tuomas J. Lukka, Christian Soeller, and others.

KNOWN BUGS

       Probably many.

AUTHOR

       Bryan Jurish <moocow@cpan.org>

   Copyright Policy
       Copyright (C) 2007-2022, Bryan Jurish. All rights reserved.

       This package is free software, and entirely without warranty.  You may redistribute it
       and/or modify it under the same terms as Perl itself.

SEE ALSO

       perl(1), PDL(3perl)