Items from PAGE 2023 course

[RichardHooijmaijers]

During the LAPP course at page 2023 a couple of things for the app were noticed:

• The app uses shell/system comands which are not suported on all linux distributions. See if a switch to callr makes things more stable
• Double check the use of project directories. It seems that sometimes the current working directory is used instead of the project directory; perform some checks to make sure that in all cases the project directory is used
• Some errors were noticed in cases that a model is copied manually into the directory. Noticed that sometimes the new model is present in the model overview but cannot be selected in the editor widget
• Add more loading screens for certain tasks (e.g. duplicate model). When this is not implemented there is a risk that people click multiple times which could potentially break the app
• See if the concept of project directories can be made more clear. At least put the project directory somewhere in the header of the app but also check the usage of a batch file opposed to starting the app from Rstudio
• The update inits functionality does currently not update the random effects; see if this can be added (maybe discuss with Matt where this info is located)
• Check if we need to implement more checks (e.g. check if data defined in model is present, check if est method is valid, etc.)

[RichardHooijmaijers]

The above was tested and the following was observed:

• Did some testing with callr. Although this works, the redirecting of output is not handled as expected. As this is key for the shiny app, do not want to change it now
• Checked the project directory vs working directory; handling of directories is now greatly improved as it is set once and passed to different modules. This should no longer be an issue
• manual copying of models; this is no longer an issue in the latest version of shinyMixR
• Added loading screens for duplicating model, this should cover the actions with the most risk
• For project directories; added the folder name in the top of the app so people know where they are working. Also added an Rstudio gadget/addin to start shinyMixR from a certain location. The latter is more robust than a batch file, as this only works on windows and implementation on (other) operating systems are highly variable
• Adapted update_inits to also take into account random effect updating
• Implemented checks when running models, mostly a alert is given in shiny. For some other checks the error is specified in the run log. Should be easier to see what is going on when a mistake is made