NEWS.md
Removed reshape2 imports and replaced with tidyr v1.0.0.
Added check_installation for users to be able to verify installations (fixes #18).
All closed form ADVAN/TRANS combinations added to the code library. Modified $ERROR so that log transformed DV and M3 models can be obtained by uncommenting lines.
The run monitor addin now uses object highlighting - useful for when objects are not in .GlobalEnv (e.g. they’re in a data.frame).
Changed run_nm() to have threads argument and make use of a job_time_spacing option to stagger multiple jobs and prevent overloading of a server (resolves #28).
Reorders arguments for decision so that auto decisions can be specified without explicitly referring to arguments
Removed future package from suggests and async arguments from nm_render*() functions.
Manual edits can now be performed on vector nm objects
Included *map*_nm() functions as equivalents to purrr::*map*_*() functions. These return nm_list objects.
Improvements to the subroutine() function and included package tests.
Added path argument to write_ctl() so control file can be written to other locations.
Added package hex sticker.
Added covariance_matrix() for extracting covariance matrices.
Added completed_nm() for handling completed nonmem runs.
Added sge_parallel_execute_batch for SGE cluster environments.
Added support for “name [unit] :trans” type parameter comments.
Fixed resolve manual edit failure when directory change was responsible for failure.
RANMETHOD=3S2 changed to RANMETHOD=3S2P in code library and examples to improve parallel processing consistency of runs.
Replaced caching mechanism of nm_render() and nm_list_render() with xfun.
Fixed bug when OMEGAS have same names as THETAS.
Added ability to use NMproject-specific code completion snippets setup_code_completion(). Typing new_nm will set up parent object, child will set up child object.
fixed bug in Windows where R markdown templates weren’t being produced.
fixed minor windows bug where addins were executing models.
run_nm() checks for psn installation in first use of R session
clean_tempfiles() replaces clean_run()
Fixed failing CRAN checks in macOS systems.
Added run monitor RStudio Addin.
Revamped the manual edit interface. This is backwards incompatible. merge conflicts will now be flagged.
Resolve manual edit ‘addin’ for handling merge conflicts in manual edits (#17)
view_patch() prints to console instead of opening window
Better instructions on RStudio Addin usage.
Fixed bug were nm_create_analysis_project would fail in windows if using the home directory due to a mismatch between how ‘~’ is interpreted by R and usethis.
The first post CRAN release
Revamped the NMproject project creation template and nm_create_analysis_project() for greater customisability ability and removed dependence on options() which was causing strange behaviour.
Expanded analysis README template for new NMprojects to run analysis level tests on build. Useful for compliance checking.
Added nm_pre_commit_hook() and nm_pre_commit_hook() for custom NMproject analysis projects
Expanded model diagnostic rmarkdown template to include more ‘xpose’ plots.
Included an experimental %f>% pipe for applying functions elementwise to vector valued nm objects.
Included is_nmproject_dir(), intended for package owners of dependent packages
Renamed nm_default_dir() to nm_dir()
Removed nm.cmd_default option in favour of the more flexible nm_default_fields. See ?nm_default_fields() for help.
decision() uses usethis::ui_yeah() for prompting rather than readline().
stage() has find_replace_dir_names argument to handle custom NMproject directory structures.
Made run_id argument mandatory in child().
Added example code for first object creation in code library dialog boxes when a model file is imported
new_nm() will now fail if user tries to use a run based on one in the Models. This is for safety as the Models directory should only contain NMproject generated code.
The code library will not display .Rroj or README files.
This is the first CRAN release of NMproject. Here are the major changes here relative to 0.5.1:
Removal and deprecation of tidyproject as an NMproject dependency.
Project libraries are now handled with renv.
Analysis project creation has been moved from tidyproject to NMproject with light wrappers around usethis project creation functionality and RStudio project templates. Analysis projects have a package-like structure.
Expanded demo based on a theophylline pharmacometric analysis and new vignette based around same demo.
The code library has been moved to NMproject. New shiny interface to code library available via RStudio ‘Addins’. The code library has been restructured to match NMproject analysis directory structure.
new_nm() function to create parent nm_list objects. This is mostly for creation of your first model. Subsequent models are generated with the child() function.
New manual edit interface via RStudio ‘Addins’. NMproject is now the only NONMEM interface (known to the author) that tracks manual edits to NONMEM code in the form of reusable patches.
Fully vectorised functions and compatibility of nm_list objects with dplyr.
Custom NMproject implementations of PsN’s bootstrap, cross-validation, stepwise covariate selection, simulation-re-estimation, and PPCs have been added.
“overwrite behaviour” RStudio ‘Addin’ has been created to allow different behaviours when modifying and rerunning a previously executed run. The default is to ask for confirmation.