NEWS

nopaco 1.0.9 (2024-10-16)

Bug fix: Fixed an issue where getPsi would results in NaN if the inputmatrix contained any unknowns (NA or NaN)

Bug fix: Fixed an issue where concordance.test would result in errors for larger datasets due to a bug in the getVar function.
Compilation Compatibility: Updated the package to allow compilation with R_NO_REMAP defined. This change anticipates the default setting expected in future R versions, specifically starting from R 4.5.0.

The default nopaco.nCPU setting previously set by parallel::detectCores(), is now maximized to a value of 2. If higher values are required, it can be set manually.
Count ties in both matrices (if given), instead of only matrix 'x'.
Can now handle infinite values in the input matrices. Previously Inf and -Inf were set to NA.
Changed the direction of 'delta' in the difference test.
The options concordance.nCPU, concordance.seed and concordance.verbose, are replaced by nopaco.nCPU, nopaco.seed and nopaco.verbose respectively. The option concordance.nDraws is replaced by nopaco.nDraws.CI (for CI estimation).
In the concordance.test function, method = simulation has been renamed as bootstrap.

Fixed memory protection inconsitentsies in c++ code that were introduced in the last bugfix

In order to resolve the linking error on Solaris, the c++ headers are reordered in line with the description in 'Writing R Extensions': R headers after system headers.
Fixed memory protection inconsitentsies in c++ code
Added ORCID to DESCRIPTION file
Updated CITATIONS file

Avoids non-stable CI estimates by excluding replicates with less than 3 values in the concordance test for differences.
Now obtains confidence intervals by bootstrapping instead of assuming asympototic normality.
Debuged code: Confidence intervals were not correct if (psi1 > psi2) in the concordance.test function
Removed the Latex package headers for language and font encodings: 'english', 'utf8x', and 'T1'

Added a vignette explaining the basic use of the package. The vignette works out an expample which data is given in the newly added data scores.rda (also requiring newly added data documentation)
A major change was made by replacing the Google dense-hash (and corresponding configuration script) by an c++11 std unordered_map in order to solve the poor portability of the package.
According to rcheck tool by Thomas Kalibera (see github.com/kalibera/rchk) the samplePsi.cpp file did contain some unprotected variables which are now protected.
Concordance.test could sometimes give p values slightly larger (in the order of machine precision) than 1. This is resolved.
Implementation of tied values is handled in this version.
The .samplePsi function did not correctly release the random.seed within the global.env causing random number generators to be non random anymore.
Fixed a warning (Found no call to: ‘R_useDynamicSymbols’) by adding a call to R_useDynamicSymbols(info, TRUE); in register.cpp

The previous version still did not compile on solaris (and possible also OSX). I found the cause in the configuration file which was not enforcing c++11. This is fixed in this version.

The previous version contained memory leaks. These have been corrected by: replacing a delete by free (in getPsi.cpp:68) freeing pointers pStateLimits and pState (in exactdistribution202.cpp:296-297) No more leaks are found by the valgrind tool.
After submission to CRAN, Kurt Hornik reported a failure to install on Debian Linux. Although the error could not be reproduced, it is assumed to be related to the c++11 dependency. The current version includes 'CXX_STD = CXX11' in the Makevars file in order to force compilation and linking with a C++11 compiler.
- The previous nopaco package did not compile on some systems because of non supported c++ headers. These headers were used in the external code (sparsehash @google). The current version includes a configure script that came with the sparsehash code hopefully solving the issue.
- remove some non-used files shipped with the external sparsehash code.
- in addition to the 'cph' role for google, I added 'aut' in the description file.
- more specifically identified the files to which the copyright applies