Making Modules • formods

Introduction

This vignette will provide an overview of the formods framework for creating reproducable modules that interact with each other. Each module has its own namespace that is mantained by using a module short name as a prefix for functions. For example the figure generation module uses FG. If you want to create a module, please submit an issue at the formods github repository with the following information:

The short name you intend to use
A brief description of the module
A list of modules it depends on

Current modules

ASM - App State Management
DW - Data Wrangling
FG - Figure Generation
FM - Formods
UD - Upload Data
MC - This is a reserved word

The current modules in development:

NCA - Non-Compartmental Analysis
MB - Model Builder

formods framework

To get started you need to create some template files. The example below assumes you are creating this module for a package called mypackage and that you are running this command in a git repository. Say that this module is used to produce widgets, the short name is MM which stands for My Module:

use_formods(SN          = "MM",
            Module_Name = "My Module",
            element     = "widgets", 
            package     = "mypackage")

MM_module_components.R - An app that can be used for testing the module and highlighting the different UI elements that are used within the module (found in inst/templates).
MM_Server.R: A bare bones file containing the expected functions and their minimal inputs. (found in R).
MM.yaml: This module configuration file contains the the minimal elements expected, but you can add your own fields to suit your modules needs (found in inst/templates).
MM_funcs.R: This script contains example code for the different elements of the module that can be used in the function examples and also to quickly test the different module functions while you develop them (found in inst/test_apps).

Expected functions

The module template will create a standard set of functions for you. The MM below will be replaced with whatever short name you choose above when you create the templates. These functions can be customized for your specific module. Some are optional and can be deleted. For example the MM_fetch_ds function is only needed if your module creates datasets and provides them for other modules to use (like the DW module exports data views to be used by other modules). The modules are designed to create elements. For example the DW module creates data view elements, the FG module is used to create figure elements, etc.

MM_Server Shiny server function.
MM_init_state Creates an empty the formods state for the module.
MM_fetch_state Each module has a function to fetch the state. Within this function there should be no interactivity. Any access of the elements to the Shiny input should be isolated. Based on differences between the input elements (current state of the app) and the stored app state can be used to trigger different things.
MM_fetch_code This takes as it’s first argument the module state. When called with only this argument it should return a character object containing all of the code needed to generate the elements of this module represented in the app. You can assume that any modules this one depends on will be defined previously. For example the FG module will return only that code associated with generating figures. It will be appended to the code from the UD and DW modules that define loading the dataset and creation of the data views. For modules where no code is generated (e.g. ASM) just return NULL.
MM_append_report If a module generates reportable outputs, this function will be used to append those outputs to the overall reports generated by formods.
MM_fetch_ds If a module provides data sets to be used in other modules you need to create this function. It should return at least the following:
- hasds Boolean variable indicating if the modules currently has any exportable datasets.
- isgood General return status of the funciton. Set to FALSE if any errors were encoutered.
- msgs A character vector of any messages to pass back to the user.
- ds A module can provide multiple datasets. This is a list with the following elements for each dataset:
  - label Text label for the dataset
  - MOD_TYPE Short name for the type of module.
  - id module ID.
  - DS Dataframe containing the actual dataset.
  - DSMETA Metadata describing DS.
  - code Complete code to build dataset.
  - checksum Module checksum.
  - DSchecksum Dataset checksum.
MM_fetch_mdl If a module provides ODE models to be used in other modules you need to create this function. It should return at least the following:
- hasmdl Boolean variable indicating if the modules currently has any exportable models.
- isgood General return status of the funciton. Set to FALSE if any errors were encountered.
- msgs A character vector of any messages to pass back to the user.
- mdl A module can provide multiple models. This is a list with the following elements for each model:
  - label Text label for the model.
  - MOD_TYPE Short name for the type of module.
  - id module ID.
  - rx_obj The rxode2 object that holds the model.
  - rx_obj_name The rxode2 object name in generated code.
  - fcn_def Text to define the model
  - DSMETA Verbose metadata describing model.
  - code Complete code to build the model.
  - checksum Checksum of the module the model came from.
  - DSchecksum Checksum of the model.
MM_onload The ASM module provides the ability to load a previous analysis. Some elements of the module state may need to be changed on load. For example if you have compiled binary data in state, and the underlying libraries can change over time, you can use this function to recompile when the analysis is loaded. This function is optional and the template example can be deleted if you do not need it. The default in the template should just pass the state back unchanged.
MM_test_mksession When testing outside of Shiny it is useful to have a prepopulated session, intput, etc objects with actual data. Each module shoudl have a test_mksession function that populates these objects with useful data. If your module depends on a different module, you can use the test_mksession for the modyou yours depends on. For example the data wrangling module depends on the upload data module. So the DW_test_mksession function uses UD_test_mksession internally to load the test dataset and then builds on top of that. The function should return the following:
- isgood Boolean indicating the exit status of the function.
- session The value Shiny session variable (in app) or a list (outside of app) after initialization.
- input The value of the shiny input at the end of the session initialization.
- state App state.
- rsc The react_state components.
MM_new_element Creates a new module element.
MM_fetch_current_element Extracts the current element from the state object.
MM_set_current_element Sets the current element to the provided value.
MM_del_element Deletes the current active element.

Expected UI components

ui_mm_compact This is a UI output that contains a compact view of your module that can be called from the main ui functions for the App. It is composed of the individual UI elements that are shown in the MM_module_components.R file. This allows the user a quick way to utilize a model (using the ui_mm_compact), and the ability to customize the module UI by manually arranging the pieces found in MM_module_components.R.

Module interaction

Say you are using the UD module to feed data into the DW module and the user goes back to the upload form and uploads a different data set. This will need to trigger a reset of the Data Wrangling module as well as tell your larger app that something has changed.

Module state and reacting to changes

Changes in module states are detected with the react_state object. For a given module of type "MM" with a module id of "ID" you would detect changes by reacting to react_state[["ID"]] and looking for changes in the checksum element below:

react_state[["ID"]][["MM"]][["checksum"]]

checksum A checksum that can be used to detect changes in this module. For example in the UD module this will change if the uploaded file or the sheet selected from a currently uplo:waded file changes.

Helper functions in formods

FM_le() - Creates log entries (le) that are displayed in the console.
FM_tc() - This can be used to evaluate code, trap errors, and process results.
has_changed()
set_hold() - Used to set a hold on one or more UI elements. This prevents internal updating of that UI element based on the current value in the App.
fetch_hold() - This will retrieve the hold status of a UI element.
remove_hold() - This will remove any holds set on a UI element.
‘FM_build_comment()’ - This creates comments from strings so they will form sections when viewed in RStudio.
FM_add_ui_tooltip() - Attaches a tooltip to a UI element.
FM_init_state() - Called at the top of your module state initialization function to create a skeleton of a module state that you can then build upon.
FM_set_notification() - Within you code you can create notifications and attach them to a module state.
FM_notify() - Used in observeEvent() to show notifications that have not yet been displayed.
FM_set_mod_state()
FM_fetch_mod_state()
FM_set_ui_msg()
FM_pretty_sort() - Used as a general sorting function that will try to make the sorted results prettier.
FM_pause_screen() - Pauses the screen when doing something on the server side that takes a while.
FM_resume_screen() - Resumes activity (unpauses the screen) when you’re done with the pause.
FM_fetch_data_format() - Creates formatting information for display for a given data frame.

The examples below require a Shiny session variable and a formods state object. Here we create some examples and other objects needed to demonstrate the functions below.

library(formods)
#> Loading formods
# This creates the state and session objects
sess_res = UD_test_mksession(session=list())
#> → UD: including file
#> → UD:   source: file.path(system.file(package="onbrand"), "templates", "report.docx")
#> → UD:   dest:   file.path("config","report.docx")
#> → UD: including file
#> → UD:   source: file.path(system.file(package="onbrand"), "templates", "report.pptx")
#> → UD:   dest:   file.path("config","report.pptx")
#> → UD: including file
#> → UD:   source: file.path(system.file(package="onbrand"), "templates", "report.yaml")
#> → UD:   dest:   file.path("config","report.yaml")
#> → UD: State initialized
#> → UD: module checksum updated:897d952fecbc804999396a96f9df4b20
state    = sess_res$state
session  = sess_res$session

# Here we load an example dataset into the df object.
data_file_local =  system.file(package="formods", "test_data", "TEST_DATA.xlsx")
sheet           = "DATA"

df = readxl::read_excel(path=data_file_local, sheet=sheet)

Setting holds on UI elements

The mechanics of the fetch state functions mean that each time a fetch state is called, all of the UI elements in the App are pulled and placed in the app state. This generally works well with some exceptions. The main exception is when you want to have a UI element that changes another UI element. Say for example you have a selection box with a UI id of my_selection. You want that selection to alter a text input with an id of my_text. However if you just poll the ui elements you may update my_text based on changes to my_selection then have those overwritten by the current value of my_text. To prevent this, you need to do two things:

When processing my_selection you need to set a hold on my_text (done with set_hold()).
When processing my_text you need to do that only if there is no hold set. This is checked with fetch_hold()

Lastly you need to remove the hold. This is done after the UI has refreshed with the new text value populated in to my_text (with the appropriate reactions set). This is done with an observeEvent that is triggered after everything else (with a priority of -100 below):

remove_hold_listen  <- reactive({
  list(input$my_selection)
})
observeEvent(remove_hold_listen(), {
  # Once the UI has been regenerated we
  # remove any holds for this module
  state = MM_fetch_state(id              = id,
                         input           = input,
                         session         = session,
                         FM_yaml_file    = FM_yaml_file,
                         MOD_yaml_file   = MOD_yaml_file,
                         react_state     = react_state)

  FM_le(state, "removing holds")
  # Removing all holds
  for(hname in names(state[["MM"]][["ui_hold"]])){
    remove_hold(state, session, hname)
  }
}, priority = -100)

The remove_hold_listen object should contain all of the inputs that create holds.

Dataframe formatting information

If you want to tables and pulldown menues based on the types of data in each column you can use the FM_fetch_data_format() function.

hfmt = FM_fetch_data_format(df, state)

# Descriptive headers 
head(as.vector(unlist( hfmt[["col_heads"]])))
#> [1] "<span style='color:#3C8DBC'><b>ID</b><br/><font size='-3'>num</font></span>"      
#> [2] "<span style='color:#3C8DBC'><b>TIME_DY</b><br/><font size='-3'>num</font></span>" 
#> [3] "<span style='color:#3C8DBC'><b>TIME_HR</b><br/><font size='-3'>num</font></span>" 
#> [4] "<span style='color:#3C8DBC'><b>NTIME_DY</b><br/><font size='-3'>num</font></span>"
#> [5] "<span style='color:#3C8DBC'><b>NTIME_HR</b><br/><font size='-3'>num</font></span>"
#> [6] "<span style='color:#3C8DBC'><b>TIME</b><br/><font size='-3'>num</font></span>"

# Subtext
head(as.vector(unlist( hfmt[["col_subtext"]])))
#> [1] "num: 1,⋅⋅⋅,360"  "num: 0,⋅⋅⋅,84"   "num: 0,⋅⋅⋅,2016" "num: 0,⋅⋅⋅,42"  
#> [5] "num: 0,⋅⋅⋅,1008" "num: 0,⋅⋅⋅,2016"

The custom headers can be used with the rhandsontable package.

hot = rhandsontable::rhandsontable(
  head(df),
  width      = "100%",
  height     = "100%",
  colHeaders = as.vector(unlist(hfmt[["col_heads"]])),
  rowHeaders = NULL
  )

To add subtext to a selection widget in Shiny you need to use the shinyWidgets package.

sel_subtext = as.vector(unlist( hfmt[["col_subtext"]]))
library(shinyWidgets)
shinyWidgets::pickerInput(
    inputId    = "select_example",
    choices    = names(df),
    label      = "Select with subtext",
    choicesOpt = list(subtext = sel_subtext))

To alter the formats shown here you need to edit the formods.yaml configuration file and look at the FM$\rightarrow$data_meta section.

Notifications

Notifications are created using the shinybusy package and are produced with two different functions: FM_set_notification() and FM_notify(). This is done in a centralized fashion where notifications are added to the state object as user information is processed. This will set a notification called Example Notification. Along with that a timestamp is set:

   state = FM_set_notification(state, "Something happened", "Example Notification")

That timestamp is used to track and prevent the notification from being shown multiple times. Next you need to setup the reactions to display the notifications. Here you can create a reactive expression of the inputs that will lead to a notification:

    toNotify <- reactive({
      list(input$input1,
           input$input2)
    })

Next you use observeEvent() with that reactive expression to trigger notifications. You need to use the fetch state function for that molecule to get the state object with the notifications. Then FM_notify() will be called an any unprocessed notifications will be displayed:

    observeEvent(toNotify(), {
      state = MM_fetch_state(id              = id,
                             input           = input,
                             session         = session,
                             FM_yaml_file    = FM_yaml_file,
                             MOD_yaml_file   = MOD_yaml_file,
                             react_state     = react_state)

      # Triggering optional notifications
      notify_res =
      FM_notify(state    = state,
                session  = session)
    })

Adding tooltips

Tooltips are created interally using the suggested prompter package. To add a tool tip to a ui element you would use the FM_add_ui_tooltip() function. For example to add the tool tip, You need to type harder! to a text input you would do the following:

uiele = shiny::textInput(
          inputId = "some_text", 
          label   = "You need to type harder!")
uiele = FM_add_ui_tooltip(state, uiele, 
      tooltip  = "This is a tooltip",
      position = "left")

Pausing the screen

To pause the screen the shinybusy package is also used. This is controlled with two functions: FM_pause_screen() is used to pause the screen and/or update the pause message, and FM_resume_screen() is used end the pause and resume interaction with the user.

FM_pause_screen(state, session)
FM_resume_screen(state, session)

formods state objects

When you create a formods state object it can have the following fields:

yaml- Contents of the formods configuration file.
MC - Contents of the module configuration file.
MM - MM here is the short name of the current module. MOD_TYPE below), this is where you would store any app information. (see below).
MOD_TYPE - Short name of the module.
id - ID of the module.
FM_yaml_file - formods configuration file.
MOD_yaml_file - Module configuration file.
notifications - Contains notifications set by the user through FM_set_notification().

App information in MM

This field state$MM is relatively free form but there are some reserved elements. These reserved keyword are:

button_counters - Counter that tracks button clicks
ui_hold - List of hold elements that is populated with set_hold()
isgood - Boolean variable indicating the state of the module.
ui_msg - Messaages returned to the UI with captured errors populated with FM_set_ui_msg()

Other than those fields you can store whatever else you need for your module.

Configuration file

YAML configuration files

# https://www.glyphicons.com/sets/basic/
#General formods (FM) configuration across modules
FM:
  include:
    # This is where you can put files to include files in the working directory
    # of the app. For the files listed below you shouldn't change the 'dest'
    # portion but you can change the source to use custom report templates.
    # If relative paths are used they will be relative to
    # the user directory (either the temp formods directory running in shiny
    # or the top level of the zip file structure when saving the app state).
    files:
    - file:
        source: 'file.path(system.file(package="onbrand"), "templates", "report.docx")'
        dest:   'file.path("config","report.docx")'
    - file:
        source: 'file.path(system.file(package="onbrand"), "templates", "report.pptx")'
        dest:   'file.path("config","report.pptx")'
    - file:
        source: 'file.path(system.file(package="onbrand"), "templates", "report.yaml")'
        dest:   'file.path("config","report.yaml")'
  # Some features (e.g. copy to clipboard) don't work when deployed
  deployed: FALSE
  #General code options for the modules
  code:
    theme:           "vibrant_ink"
    showLineNumbers: TRUE
    # File name of the R script to contain generation code
    gen_file: run_analysis.R
    # This is the preamble used in script generation. It goes on the
    # top. Feel free to add to it if you need to. Note that packages should be
    # listed in the packages section at the same level.
    gen_preamble: |-
      # formods automated output ------------------------------------------------
      # https://formods.ubiquity.tools/
      rm(list=ls())
    # Each module should have a packages section that lists the packages
    # needed for code genereated for that module.
    packages: ["onbrand", "writexl"]
  notifications:
    config:
      # You can put any arguments here that would be arguments for
      # config_notify(). See ?shinybusy::config_notify() for more information
      success:
        useFontAwesome: FALSE
        useIcon:        FALSE
        background:     "#5bb85b"
      failure:
        useFontAwesome: FALSE
        useIcon:        FALSE
        background:     "#d9534f"
      info:
        useFontAwesome: FALSE
        useIcon:        FALSE
        background:     "#5bc0de"
      warning:
        useFontAwesome: FALSE
        useIcon:        FALSE
        background:     "#f0ac4d"
  reporting:
    # enabled here controls reporting for the app. Individual modules can be
    # controlled in their respective configuration files
    enabled: TRUE
    # The content_init section is used to initialize reports. You shouldn't
    # change the xlsx rpt but the docx and pptx rpt can be altered to
    # preprocess reports if you need to such as adding default.
    content_init:
      xlsx: |-
           rpt = list(summary = NULL,
                      sheets  = list())
      docx: |-
           rpt  = onbrand::read_template(
             template = file.path("config", "report.docx"),
             mapping  = file.path("config", "report.yaml"))
      pptx: |-
           rpt  = onbrand::read_template(
             template = file.path("config", "report.pptx"),
             mapping  = file.path("config", "report.yaml"))
    # Word template can contain placeholders. This is where you can put the
    # default values for placeholders. There are some nuances associated with
    # creating placeholders and you should see the documentation about them in
    # the onbrand package to better understand that:
    #
    #   https://onbrand.ubiquity.tools/articles/Creating_Templated_Office_Workflows.html#placeholder-text
    #
    # Each element below contains the placeholder used and should contain a
    # location and a value. For example the default document contains a header
    # placeholder in the upper right that looks like:
    #
    #  ===HEADERRIGHT===
    # The placeholder below is HEADERRIGHT, the location is header and by
    # default we will replace it with nothing.
    #
    # These defaults can be overwritten in the save section.
    #
    phs:
      - name:      "HEADERRIGHT"
        location:  "header"
        value:     ""
      - name:      "HEADERLEFT"
        location:  "header"
        value:     ""
      - name:      "FOOTERLEFT"
        location:  "footer"
        value:     ""
  ui:
    # See ?actionBttn for styles
    button_style: "fill"
    # Max size for picker inputs
    select_size:  10
    color_green:  "#00BB8A"
    color_red:    "#FF475E"
    color_blue:   "#0088FF"
    color_purple: "#bd2cf4"

  data_meta:
    # This controls the overall format of headers and the select subtext for
    # data frames with the following placeholders surrouned by ===:
    # COLOR  - font color
    # NAME   - colum name
    # LABEL  - type label
    # RANGE  - this depends on the nature of the data in the column:
    #        - If there are between 1 and 3 values then those values are shown.
    #        - If there are more than 3 values then the min and max are show.
    data_header:  "<span style='color:===COLOR==='><b>===NAME===</b><br/><font size='-3'>===LABEL===</font></span>"
    subtext:      "===LABEL===: ===RANGE==="
    # Separator when showing more than three in a column. For exmaple if you
    # had a dataset with 1,2,3,4,5,6 and many_sep was ",...," then it would
    # appear as "1,...,6"
    many_sep: ",⋅⋅⋅,"
    # This controls the differences for different data types. Take the output
    # of typeof(df$colname) and put an entry for that output here.
    data_types:
      character:
        color:    "#DD4B39"
        label:    "text"
      double:
        color:    "#3C8DBC"
        label:    "num"
      integer:
        color:    "#3C8DBC"
        label:    "num"
      other:
        color:    "black"
        label:    "other"
  labels:
    # JMH remove this once the datset stuff has been moved over
    default_ds:   "Original data set"
  user_files:
    use_tmpdir:     TRUE
  logging:
    enabled:        TRUE
    timestamp:      TRUE
    timestamp_fmt: "%Y-%m-%d %H:%M:%S"
    log_file:      "formods_log.txt"
    console:       TRUE