NEWS

FIMS 0.9.3

Reduce verbose FIMS logging and refine messages
Cleans up integration tests after switching log_devs to random effects
Revert changes that eliminated compiler warnings because they were causing tests to fail but not R CMD check

Adds vignette on FIMSFrame and data_big
Adds logging information to fims-demo and updates the logging vignette to be dynamic instead of relying on a saved image
Adds a vignette on how to add new C++ modules to FIMS
Adds WHAM-inspired convergence checks to FIMS models
Relies on SparseNUTS for Bayesian test
Refactors FIMSFrame plot method to provide a summary of the data

Adds time-varying weight-at-age
Adds time-varying selectivity
Makes recruitment deviations random effects by default when using the wrapper functions
Adds a glossary of mathematical symbols to the doxygen documentation
Brings doxygen to the pkgdown site
Fixes hidden bug in dimension_folded_to_tibble() in while loop
Removes get_log_module
Removes calculation of uncertainty within the C++ code
Changed package data name to data_big and added helper data
Changes m_*() functions to model_*()

Fixes two bugs in fit_fims when getting random effect names and not wanting measurements of uncertainty
Enhance the documentation by adding a Cheatsheet, updating the README, increasing information for some functions, added Rcpp training vignette, adding additional equations to doxygen

Allow for projections based on fixed fishing mortality, catches, or in theory using priors on any parameter or derived quantity that you want such as the ratio of current spawning biomass to initial biomass.
Fixes create_default_Population() to define age and years for log_M

Fixes json output when there are non numeric values, infinity values, and -999
Allows the output columns to be merged with reshape and dplyr when character values are combined with strings
Fixes spawning biomass output in print method to report value
Fixes get_output() when there is no sdreport with added argument do_sd_report

Turns on the code for model families that allows for more model types in the future than just the catch-at-age model that is currently implemented through the use of FisheryModelBase, which is a unified parent class for all model families. Moved derived quantities (e.g., numbers at age, spawning biomass, catch at age) from Fleet and Population objects into FisheryModelBase. Refactored Population and Fleet classes to be trait only objects
Converts nested list of lists used for model control to a tibble with one line per parameter that can be un-nested and nested.
Increased the documentation of FIMS by using formulas in doxygen, building the doxygen on the NOAA-FIMS github.io site instead of its own web page, writing more information in the testing README, etc.
Removes seasonality logic given it was not being used
Removes date columns in input data in exchange for timing
Uses {stockplotr} for the figures in the "fims-demo" vignette

Fixes bugs in fims_math.hpp when using TMBad library to use the standard library functions (exp, log, cos, sqrt, pow) when necessary.

Fix json output to use full stop rather than underscore for names, a bug introduced in v0.5.0. Also fixes reshape_tmb_estimates() where the splits were incorrectly referenced because module_id comes before the label.

Allows for CPUE data and landings within a single fleet and changes the composition data to be separate from the catch. If there are catches, compositions are based on F, if you have survey compositions those are scaled by q, and you can also have composition data without catch or CPUE and those match the population.
Fixes bug in finalize() where variables were going out of scope by using a shared pointer and RealVector, which allows for estimates tibble to return more. Output is integrated between json and TMB using reshape_tmb_estimates(), reshape_json_estimates(), and reshape_json_derived_quantities(). ParameterVector are now set using resize() and purrr::walk() rather than methods::new().
Increases code coverage
- Restructures R tests using use_testthat_template()
- Starts to use Rcpp::Rcerr instead of Rcpp::Rcout
- Skips parallel and distribution rcpp test
- Adds FIMS_dmultinom() distribution to match TMB
Fixes bug in get_number_of_parameters().
Uses static_cast([-0-9.]+) to wrap constant variables.
Fixes cmake_minimum_required to use a range and updates googletest version.
Adds minimal fims-demo vignette.
Removes unused estimated_log_M and estimated_log_init_naa from inst/include/interface/rcpp/rcpp_objects/rcpp_population.hpp.
Adds gdb to the devcontainer.json file.
Fixes broken model comparison project links.
Adds dependabot.yml file to update GitHub actions.
Combine initialize_comp() functions.

Pluralism in cli was incorrect in FIMSFrame.
Composition modules are initialized using the composition times the uncertainty, which was failing for -999 values because -999 * uncertainty does not equal -999 and was leading to the likelihood being evaluated.

Fits to length data using an age-to-length-conversion matrix, data1 includes the necessary information needed to fit to both ages and lengths.
Adds C++ ParameterVector to allow for the estimation of time-varying parameters.
Implements R wrapper functions to facilitate
- creating the input model specifications with create_default_*(), update_parameters(), and initialize_*();
- adding -999 to the missing fleet, year, age, length, etc. combinations;
- running the model with a user-supplied argument of n_of_loops, where the default is three, to restart the optimizer from the previous run of nlminb;
- summarizing the output with the FIMSFit() function and class.
Implements a switch for global verbosity within FIMS through the use of {cli} messages and warnings.
Updates the logging system complete with a vignette about how to use it, the logging system can be used for both R and C++ errors, warnings, and information.
Creates the initial infrastructure to implement random effects with density functions.
Implements a helper function to get the parameter names from the C++ code and populate the results with those names.
Makes lpdf_vec return 0 if data is missing.

Added a pkgdown site for FIMS
Add code cov tests and site link
Remove recruitment bias correction and adjustment because we plan to use TMB (https://github.com/NOAA-FIMS/FIMS/issues/185)
Improve logging coverage and separate logs into multiple files
Put recruitment deviations on the log scale when input
Report more values with standard errors
Code cleanup, documentation, and consistency improvements
Remove rapidjson dependence
Add start_year and end_year as slots to FIMSFrame
Make log_obs_error a vector
Load the Rcpp module upon package install so components don't need to be prefaced by fims$
Refactor vectors to use fims::Vector class
Add Newton steps to FIMS integration test to improve convergence in integration test