Filter Panel for Developers

NEST CoreDev

Source: vignettes/filter-panel-for-developers.Rmd

Filter Panel API

Getting and setting filter states

All filter panel classes have dedicated methods to set and get the current filter state. These methods include:

  • get_filter_state - returns current state of filters as teal_slices object
  • set_filter_state - adds or modifies filters based on teal_slices object
  • remove_filter_state - removes particular filter states based on teal_slices object
  • clear_filter_states - removes all filter states

Setting and getting filter states are done through teal_slices object which is a collection of teal_slice objects. Think of a teal_slice as a quantum of information that fully describes the filter state of one variable.

In order to tell FilteredData to set a filter for a specific variable, one must call the set_filter_state method with a teal_slices object containing a teal_slice that refers to the variable of interest. To remove a particular FilterState object, one must call the remove_filter_state method using a teal_slices containing a teal_slice that refers to the respective variable.

Each teal_slice object contains all the information necessary to:

  1. Determine the column in the data set on which to apply the filter expression:

    • dataname - name of the data set
    • varname - name of the column
    • experiment (only for MultiAssayExperiment objects) - name of the experiment
    • arg (only for SummarizedExperiment objects, e.g within a MultiAssayExperiment) - name of the argument in the call to subset (subset of select)

  2. Express or store the current selection state:

    • selected - selected values or limits of the selected range
    • keep_inf - determines if Inf values should be dropped
    • keep_na - determines if NA values should be dropped
    • expr - explicit logical expression

  3. Control the behavior and appearance of the FilterState object:

    • choices - determines set of values or range that can be selected from
    • multiple (only for ChoiceFilterState) - allows multiple values to be selected
    • fixed - forbids changing state of the FilterState
    • anchored - forbids removing the FilterState
    • title - displayed title of the filter card

In addition, every teal_slice object has an id.

It is impossible to create FilteredData with slices with duplicated ids. This is because filter states are both created and modified with the set_filter_state method so if two consecutive calls to set_filter_state are passed a teal_slice with the same id, the first call will instantiate a FilterState, and the second call will modify it.

Creating teal_slices with slices with duplicated ids is forbidden and will raise an error.

1. Setting the filter state 

library(teal.slice)

datasets <- init_filtered_data(list(iris = iris, mtcars = mtcars))

set_filter_state(
  datasets = datasets,
  filter = teal_slices(
    teal_slice(dataname = "iris", varname = "Species", selected = "virginica", keep_na = FALSE),
    teal_slice(dataname = "mtcars", id = "4 cyl", title = "4 Cylinders", expr = "cyl == 4"),
    teal_slice(dataname = "mtcars", varname = "mpg", selected = c(20.0, 25.0), keep_na = FALSE, keep_inf = FALSE),
    include_varnames = list(iris = c("Species", "Sepal.Length")),
    exclude_varnames = list(mtcars = "cyl")
  )
)

2. Updating filter states. *Works only in the shiny reactive context. 

set_filter_state(
  datasets = datasets,
  filter = teal_slices(
    teal_slice(dataname = "mtcars", varname = "mpg", selected = c(22.0, 25.0))
  )
)

3. Getting the filter state 

## {
##   "slices": [
##     {
##       "dataname"       : "iris",
##       "varname"        : "Species",
##       "id"             : "iris Species",
##       "choices"        : ["setosa", "versicolor", "virgin...
##       "selected"       : ["virginica"],
##       "keep_na"        : false,
##       "fixed"          : false,
##       "anchored"       : false,
##       "multiple"       : true
##     },
##     {
##       "dataname"       : "mtcars",
##       "id"             : "4 cyl",
##       "expr"           : "cyl == 4",
##       "fixed"          : true,
##       "anchored"       : false,
##       "title"          : "4 Cylinders"
##     },
##     {
##       "dataname"       : "mtcars",
##       "varname"        : "mpg",
##       "id"             : "mtcars mpg",
##       "choices"        : [10.4, 34],
##       "selected"       : [22, 25],
##       "keep_na"        : false,
##       "keep_inf"       : false,
##       "fixed"          : false,
##       "anchored"       : false,
##       "multiple"       : true
##     }
##   ],
##   "attributes": {
##     "exclude_varnames" : {
##       "mtcars"         : "cyl"
##     },
##     "include_varnames" : {
##       "iris"           : ["Species", "Sepal.Length"]
##     },
##     "count_type"       : "none",
##     "allow_add"        : true
##   }
## }

4. Removing filter states 

remove_filter_state(
  datasets = datasets,
  filter = teal_slices(
    teal_slice(dataname = "iris", varname = "Species")
  )
)

5. Clearing the filter state

Controlling settings of the filter panel

In addition to controlling individual filter states through set_filter_state, one can also manage some general behaviors of the whole filter panel. This can be done with arguments of the teal_slices function:

  1. include_varnames defines which columns in the used data sets are allowed to be filtered on. In the following example only two columns of iris and two columns of mtcars will be able to have filters set.
set_filter_state(
  datasets,
  teal_slices(
    include_varnames = list(
      iris = c("Species", "Sepal.Length"),
      mtcard = c("cyl", "mpg")
    )
  )
)
  1. exclude_varnames defines which columns in the used data sets are not allowed to be filtered on. In the following example all variables except the four will be available to choose from.
set_filter_state(
  datasets,
  teal_slices(
    exclude_varnames = list(
      iris = c("Species", "Sepal.Length"),
      mtcard = c("cyl", "mpg")
    )
  )
)
  1. count_type defines how observation counts are displayed in filter cards
“none” “all”
Distribution in unfiltered data Filtered vs. unfiltered distribution
  1. allow_add determines whether the “Add Filter Variables” module will be displayed to allow the user to add new filters.

Filter panel as a module

All the instructions herein can be utilized to build a shiny app.

## 
## Attaching package: 'bslib'
## The following object is masked from 'package:utils':
## 
##     page
# initializing FilteredData
datasets <- init_filtered_data(list(iris = iris, mtcars = mtcars))

# setting initial filters
set_filter_state(
  datasets = datasets,
  filter = teal_slices(
    teal_slice(dataname = "iris", varname = "Species", selected = "virginica", keep_na = FALSE),
    teal_slice(dataname = "mtcars", id = "4 cyl", title = "4 Cylinders", expr = "cyl == 4"),
    teal_slice(dataname = "mtcars", varname = "mpg", selected = c(20.0, 25.0), keep_na = FALSE, keep_inf = FALSE),
    include_varnames = list(iris = c("Species", "Sepal.Length")),
    exclude_varnames = list(mtcars = "cyl"),
    count_type = "all",
    allow_add = TRUE
  )
)

ui <- bslib::page_fluid(
  bslib::layout_column_wrap(
    style = htmltools::css(grid_template_columns = "2fr 1fr"),
    tags$div(
      tags$div(
        actionButton("add_species_filter", "Set iris$Species filter"),
        actionButton("remove_species_filter", "Remove iris$Species filter"),
        actionButton("remove_all_filters", "Remove all filters")
      ),
      verbatimTextOutput("rcode"),
      verbatimTextOutput("filter_state")
    ),
    datasets$ui_filter_panel("filter_panel")
  )
)

server <- function(input, output, session) {
  # calling filter panel module
  datasets$srv_filter_panel("filter_panel")

  # displaying actual filter states
  output$filter_state <- renderPrint(print(get_filter_state(datasets), trim = FALSE))

  # displaying reproducible filter call
  output$rcode <- renderText(
    paste(
      sapply(c("iris", "mtcars"), datasets$get_call),
      collapse = "\n"
    )
  )

  # programmatic interaction with FilteredData
  observeEvent(input$add_species_filter, {
    set_filter_state(
      datasets,
      teal_slices(
        teal_slice(dataname = "iris", varname = "Species", selected = c("setosa", "versicolor"))
      )
    )
  })

  # programmatic removal of the FilterState
  observeEvent(input$remove_species_filter, {
    remove_filter_state(
      datasets,
      teal_slices(
        teal_slice(dataname = "iris", varname = "Species")
      )
    )
  })
  observeEvent(input$remove_all_filters, clear_filter_states(datasets))
}

if (interactive()) {
  shinyApp(ui, server)
}