NEWS.md
To comply with CRAN policies some breaking changes were required:
Breaking: - Functions that write files no longer default to the current working directory. The path argument now defaults to NULL and must be supplied explicitly when file output is requested. Affected functions are: write_in_out_json, write_ModelProperties_json, write_fileMetadata_json, calculateFitStat, calculateROCStat, calculateLiftStat, diagnosticsJson, and create_scoreSample. This may cause errors if a path is not explicitly assigned. These functions now also return explicitly instead of invisibly. - Codegen add_target_name default changed to FALSE (comment: it was confusing and unecessary most of the time)
Changes: - fix: json functions no longer return values invisibly. - fix: Functions that writes to path requires it defined explictly now - fix: codegen has been updated with some fixes and quality of life. (The user should still verify the generated code and files) - fix: The R scoring code sample and generated codes no loger uses .GlobalEnv, it relies on R context inheritance to load models. (comment: it was required to add sasctl to CRAN, it can still be used if the user wants.) - fix: Updated docs and readme to use file.path instead of paste0(...) when referencing system files.
output_as_df argument, by default the generated scorecode it returns a data.frame instead of a list. Required for the MM gateway background.add_target_name. By default a column with the name of the target will be returned with the predicted values.FALSE. As of Viya 2025.6, PMML 4.4 is natively supported by Model Manager.register_model wrongly identifying which version of Viya it is interacting with, which lead to use the wrong upload format.file.path instead of paste in the generated code when referencing file paths.scr_batch argument to format_data_json, allowing to have all data in a single json payload for Viya 2024.7 SCR batch scoring feature.openBrowser using a function instead of the argument.codegen for Tidy Models workflows guessing the predictors from the pre processing pipeline instead of the model workflow recipe.verify_ssl argument to session to be used in all api calls.openBrowser argument to session to open the browser automatically when auth_code = TRUE.publish_model exact argumentpublish_model fails.delete_* functions now return invisible(), to avoid polluting the terminal.register_model()
session to connect to Viya 2020.xcodegen function in experimental state. Works for simple lm, glm models and tidymodels workflow with regression or classification model mode`targetName and targetPredicted arguments to be named, now it avoids variable positioning errors.NEWS.md file to track changes to the package.Small bug fixes a quality of life improvements.
refresh_session functionsession(... auth_code = TRUE) to work on Jupyter Notebook and remove require to add client_secret parameterbase64enc and dplyr dependencies, for refresh_token and to make sure furrr doesn’t fail when called through masScore(...)