NEWS | R Documentation |
Converted test suite to tinytest
framework to reduce
dependencies.
Added DOI and CFF for citation purposes
Obtained access to original Fritsch paper and used supplied Pade expansion (equation 5) for values close to 0, eliminating need for Halley step close to 0 and speeding up code slightly
Used series expansion around 0 for more accurate return value for x ≤ 1e-16 (thanks to Jerry Lewis)
Contracted region around -1/e which returns -1
to Machine$double.eps
Converted continuous integration to Github actions
Corrected minor issues tripping CRAN checks
Adjusted linking to other packages in help documentation
Moved repository to Github
Added travis, appveyor, and codecovr
Added check for package version and behavior at -1/e
explicit calls to std::
in the C++ code
Corrected regression which treated PosInf and NegInf the same on principle branch (thanks to CRAN & Uwe Ligges)
Added test to ensure PosInf and NegInf are properly identified accurately
Expanded existing tests
Use R magic words and Rcpp sugar instead of calls to STD (caused regression)
Added more tests near asymptotes and for real vs. integer
Added ORCID to DESCRIPTION
Reduced max interations in some loops as convergence acheived more quickly
Refactored old nomenclature
Cleaned code for whitespace and unnecessary comments
Spellcheck and minor cleanup of help file
Added new package native routine registration.
Now using RcppParallel for calculation. Speedup when calculating more than one value at a time should be noticeable, e.g. around a 5X speedup on an 8 core machine. There should be no noticeable change when calculating single values at a time.
Added testing for proper behavior near 0 due to log in Fritsch step implementation
Edited NEWS.Rd so that it shows properly on CRAN
Switched from exclusively using Halley iteration to usually using Fritsch iteration resulting in a significant increase in speed in both branches
Added explicit exportation of headers for use in the LambertW package (Thanks to Georg M. Goerg).
Update to initial release
Adjust NEWS to more closely follow keep a changelog suggestions
Removed mention of LambertW and its importing the gsl
package
from the documentation as it now imports this package instead.
Have both branches of the function return their respective limits.
The main branch returns Inf
when it is passed Inf
and the
secondary branch returns -Inf
when it is passed 0 (Thanks to Georg
M. Gorg)
Make unit tests more robust
Update code for tests for equality for changes in testthat
package
Remove extra parenthesis which was causing boolean to be passed to
fabs
(thanks to Professor Brian Ripley)
Added URL and Bugreports to DESCRIPTION
Implemented minor optimizations to reduce number of needed calculations
Added comments to source explaining Halley step and provenance of numeric coefficients in Pade approximant
Added more robust unit testing
Enhanced and cleaned documentation
Cleaned up source code from unnecessary comments
Initial release