First draft of Shiny app

experiment-trait-data-visualizer/app.R

@@ -0,0 +1,163 @@
+library(shiny)
+library(traits)
+library(ggplot2)
+library(lubridate)
+library(DataCache)
+library(timevis)
+
+# set options for BETYdb API
+knitr::opts_chunk$set(echo = FALSE, cache = TRUE)
+options(betydb_key = readLines('~/.betykey', warn = FALSE),
+        betydb_url = "https://terraref.ncsa.illinois.edu/bety/",
+        betydb_api_version = 'beta')
+
+# get list of seasons for user to select from
+experiments <- as.data.frame(betydb_query(table='experiments'))
+seasons <- unique(experiments[c('start_date', 'end_date')])
+rownames(seasons) <- paste0('[', seasons$start_date, ']', ' - ', '[', seasons$end_date, ']')
+
+# set page UI
+ui <- fluidPage(
+
+  column(width=8, offset=2, 
+    title = "TERRA-REF Experimental Data",
+
+    h1('TERRA-REF Experimental Data'),
+
+    # season menu
+    selectInput('selectedSeason', 'Season', rownames(seasons)),
+
+    hr(),
+
+    h3('Trait Data'),
+
+    # variable menu to be rendered when variables for a given season are parsed in server()
+    uiOutput('selectVariable'),
+
+    plotOutput('traitPlot'),
+
+    hr(),
+    h3('Managements Data'),
+
+    timevisOutput('timeline')
+  )
+
+)
+
+# load trait data from BETYdb
+# function is used only by DataCache library to get and update data
+loadTraitData <- function(startDate, endDate) {
+
+  # set progress bar while API is queried
+  withProgress(message="Retrieving BETYdb Data", value=0, {
+
+    fullTraitData <- data.frame()
+
+    initialDateDiff <- as.numeric(difftime(endDate, startDate, units="days"))
+    currDate <- startDate
+    # loop through all days in a given season
+    while (endDate - currDate != 0) {
+      # get trait data for each day
+      currTraitData <- betydb_query(table='traits', date=paste0('~', currDate), limit='none')
+      fullTraitData <- rbind(fullTraitData, currTraitData[c('date', 'mean', 'variable_id', 'specie_id')])
+      currDate <- currDate + days(1)
+
+      # update progress bar
+      incProgress(1/initialDateDiff)
+    }
+  })
+
+  # format data as for usability with DataCache library
+  retData <- list(fullTraitData)
+  names(retData) <- 'fullTraitData'
+
+  return(retData)
+}
+
+getManagementsData <- function(startDate, endDate) {
+
+    fullMgmtData <- data.frame()
+
+    currDate <- startDate
+    while (endDate - currDate != 0) {
+      # get management data for each day
+      currMgmtData <- betydb_query(table='managements', date=paste0('~', currDate))
+      fullMgmtData <- rbind(fullMgmtData, currMgmtData[c('date', 'mgmttype')])
+      currDate <- currDate + days(1)
+    }
+
+    return(fullMgmtData)
+}
+
+# handle all app logic
+server <- function(input, output) {
+
+  # set reactive start date and end date variables to change when selected season changes
+  selectedSeasonRow <- reactive({ seasons[input$selectedSeason,] })
+  seasonStartDate <- reactive({ as.Date(selectedSeasonRow()$start_date) })
+  seasonEndDate <- reactive({ as.Date(selectedSeasonRow()$end_date) })
+
+  # set unique cache name for date range for usability with DataCache library
+  cacheName <- reactive({ paste0('TraitCache_', seasonStartDate(), '_', seasonEndDate()) })
+
+  # render menu for selecting variable to view data for
+  output$selectVariable <- renderUI({
+
+    # get access to 'fullTraitData' from cache
+    data.cache(cache.name=cacheName(), loadTraitData, startDate=seasonStartDate(), endDate=seasonEndDate(), frequency='daily')
+
+    # get unique variable ids from observations in current season
+    variableIds <- unique(as.numeric(fullTraitData$variable_id))
+
+    # query API for readable names for variable ids, set names
+    variableNames <- vector()
+    for (variableId in variableIds) {
+      varName <- betydb_query(table='variables', id=variableId)$name
+      varName <- gsub('_', ' ', varName)
+      variableNames <- c(variableNames, varName)
+    }
+    names(variableIds) <- variableNames
+
+    selectInput('selectedVariable', 'Variable', variableIds)
+  })
+
+  # render plot for selected variable
+  output$traitPlot <- renderPlot({
+
+    # get access to 'fullTraitData' from cache
+    data.cache(cache.name=cacheName(), loadTraitData, startDate=seasonStartDate(), endDate=seasonEndDate())
+
+    # only render plot of a variable is selected
+    if (!is.null(input$selectedVariable)) {
+      # get observations for selected variable
+      variableIdData <- betydb_query(table='variables', id=input$selectedVariable)
+      variableTraitData <- subset(fullTraitData, variable_id==as.numeric(variableIdData$id))
+
+      # get specie data for title
+      specieId <- unique(as.numeric(variableTraitData$specie_id))
+      specieData <- betydb_query(table='species', id=specieId)
+      title <- paste0('Mean ', gsub('_', ' ', variableIdData$name), ' for ', specieData$scientificname)
+
+      # generate timeseries of boxplots from mean value
+      ggplot(variableTraitData, aes(as.Date(date), mean)) + 
+      geom_boxplot(aes(group=cut_width(as.Date(date), 1))) +
+      labs(title=title,
+        x="Observation Dates", y=variableIdData$units) +
+        theme(text = element_text(size=20), axis.text.x = element_text(angle=45, hjust=1)) +
+        expand_limits(y=0) + xlim(seasonStartDate(), seasonEndDate())
+    }
+  })
+
+  # generate timeline visualization for managements data
+  output$timeline <- renderTimevis({
+    mgmtData <- getManagementsData(startDate=seasonStartDate(), endDate=seasonEndDate())
+    timelineData <- data.frame(
+      id=1:nrow(mgmtData),
+      content=paste0(mgmtData$mgmttype),
+      start=as.Date(mgmtData$date)
+    )
+    timevis(timelineData)
+  })
+}
+
+shinyApp(ui=ui, server=server)