https://github.com/dangrossman/bootstrap-daterangepicker

This may be an issue, or more of a question about how to do something - it doesn't seem possible as I see it now.

I have an application that takes an input (a search term, for example), processes that input in the shinyServer() and returns output to the mainPanel().

Conceptually, I would like to further manipulate this output, by creating checkboxes/sliders/, based on the output - that can be modified, then reactively updating certain outputs. Normally (in R), I would do something like this:

for(element in as.character(elements)) {
element_id <- paste(element, 'id', sep = '_')
element_label <- paste('Change level of', element', ':')
sliderInput(element,
element_label,
value = as.numeric(outputData[element]),
min = 1,
max = 1000)
}

... this for-loop approach does not work - the only solution I have come up with is to hard-code a bunch of conditionalPanel()s, and activate the ones relevant to the given search/output in question, which compromises any sort of order I might desire to display these UI elements in.

Is there a way to deliver a variable number of UI elements based on the data held in the Output object that shinyUI() elements draw from?

BTW! ... thank you soooo much for developing this package. I have been waiting for exactly this sort of project for 3 years now. It may be new and have development to go, but what you have here is exceptional and I can't thank you enough :)

PS I have only been using this package since Sunday, so I may be missing something obvious. I'm also new to the sort of server-client communication structure (i.e., input$ and output$ vars)

ui.R:

library(shiny)
library(twitteR)

Define UI for miles per gallon application

shinyUI(pageWithSidebar(

Application title

headerPanel('FitnessTrack'),
sidebarPanel(
selectInput("subscriber", "Select Subscriber:",
choices = c("nutwition_log", "anotherAccount")),
numericInput("obs", "Number of observations to view:", 10)
),
mainPanel(
tableOutput("view")
)
))

server.R:

library(shiny)
library(twitteR)

Define server logic required to plot various variables against mpg

shinyServer(function(input, output) {

datasetInput <- reactive(function() {
userTweets <- userTimeline(input$subscriber)
do.call("rbind", lapply(userTweets, as.data.frame))
})

output$view <- reactiveTable(function() {
datasetInput()
})

})

...

when I call, "> runApp()", I would expect the, "tableOutput('view')", call to print the datasetInput data frame to the mainPanel() in the browser.

I can run the operations in the datasetInput() reactive function, from the usual R terminal and get the data frame I expect to see. I am curious why the tableOuput('view') call here is not working as I might expect it to, and print the same table to the browser.

I also checked that a hard-coded data frame would print to the browser and it did. I suspect, then, that tableOutput()'s failure has something unique to do with the "do.call()" call, but I really don't know... I've only been using the shiny package for a number of hours now - and am not a versed programmer.

PS I've been waiting 3 years for something like shiny. Thank you!

I'm playing with a little app design that has something like this:

numericInput("id", "ID:", 10000000000),

Unfortunately, that default displays as 1e+10. The function should probably have some sort of formatting or number-type argument that disables scientific notation, potentially disables the decimal point for integers, etc.

Related, it'd be useful to be able to disable the stepping part of that widget, in cases where it's not relevant (ID numbers, phone numbers, etc.)

An alternative to this would be a filter to textInput that could be used to restrict input to a certain regular expression or similar.

Would love the checkboxGroupInput to (optionally) have a select all/select none box associated with the group.

It would be useful to have a scrollbar in the sidebar. Right now, when there are many options, you may have to scroll down to set an option, then scroll back up to see the result.

when using headerPanel(a('Dashboard',href='http://x.com'))

the title get <a href="http://x.com">Dashboard</a> but not Dashboard

When I run install.packages("shiny"), I get this error:

Loading Tcl/Tk interface ... Error: .onLoad failed in loadNamespace() for 'tcltk', details:
    call: dyn.load(file, DLLpath = DLLpath, ...)
    error: unable to load shared object '/usr/lib/R/library/tcltk/libs/tcltk.so':
    libtk8.5.so: cannot open shared object file: No such file or directory

So I tried install.packages("tcltk2"), which gives me the same error. Is there a requirement for this package I am missing?

Arch Linux x86_64
R 2.15.2

Our slider control is based on the jslider component: https://github.com/egorkhmelev/jslider

There are several additional capabilities of this component which we should expose:

1. Ability to input two values simultaneously as a range
2. Non-linear scales (via the heterogeneity property)
3. Tick labels (note we'd need to add code to do snapping to ticks since it doesn't do this by default)
4. Additional display formats (e.g. currency)
5. Custom input types (e.g. time -- this requires a custom calculate function)

Is a fully functional 'fileInput' button far away? The button works well locally but causes a crash on glimmer, I assume because it is still experimental.

If it isn't a priority at the moment I can probably make do with a simple 'paste your data' input option, but would prefer the upload option.

Thanks

Kevin

Here is the error log from the JavaScript error console:

file: Genepop_[KK_edit].gen [0%] shiny.js:1184
file: Genepop_[KK_edit].gen [83%] shiny.js:1184

Listening on port 44915
Error in switch(msg$method, init = { : EXPR must be a length 1 vector
Calls: runApp ... tryCatchList -> serviceApp -> service ->

Execution halted

edit: error information

In Linux, some objects are not antialiased when you use png output. For example, the default point shape in ggplot2 (#16) is jagged-looking.

It would be really useful to have an option to use CairoPNG() from the Cairo package for output. (Note that this differs from doing png(type="cairo")).

See this gist for examples rendered in Linux, with pictures:
https://gist.github.com/4036319

Need to give developers a way to specify what values are valid, at least for text boxes.

Checkboxes with id don't register any value when initially added. They need to use the InputBinding system like all the other controls.

it does not seem to worth a dependency; I guess it takes you five minutes to port an existing C routine to do base5 encoding (e.g. http://base64.sourceforge.net/b64.c, or see the orphaned package http://cran.r-project.org/package=base64)

I have an R version here: https://github.com/yihui/knitr/blob/master/R/utils-base64.R

It would be nice for reactives inside of hidden tabs in a tabset that depend on inputs in the active tab to not be called until the hidden tab is made active. This seems to happen for plots, but not for other things. I don't know from a design standpoint whether this would be a good default or not, but it would be nice to be able to specify somehow as an option.

Line-oriented deduplication can cause JavaScript code to be mangled, for example:

head(tags$script("
if (location.hash) {
  console.log(location.hash);
}
else {
  console.log('No hash');
}
"))

In this example, the latter close-brace will be stripped because an identical line appeared earlier.

We probably need a solution that operates on the direct-child level.

e.g. when I want to load mathjax, I can use tag('script'): https://github.com/yihui/appstat/blob/master/beta-hat/ui.R#L10 This trick works because mathjax allows itself to be loaded in the body section; for other JS libraries, it may not work, and it is also better for CSS to be added to the head section.

When creating a sliderInput through reactiveUI the slider is not rendered properly unless another slider has already been created in ui.R.

(This was seen under Fedora Core 16 with both Firefox 16.0.2 and Google Chrome 23.0.1271.64)

How to reproduce:

ui.R

library(shiny)

shinyUI(pageWithSidebar(
  headerPanel("Reactive UI test"),
  sidebarPanel(
              uiOutput("reactiveSlider")
              ),
  mainPanel ()
  ))

server.R

library(shiny)

shinyServer( function(input, output)
  {
  output$reactiveSlider <- reactiveUI(function()
    {
    sliderInput("reactiveSlider", "Choose a value:", 
                 min=1, max=10, step=0.5, value=2, ticks=T)   
    })
  })

This results in reactiveSlider being rendered as a text input rather than a slider.
Simply adding a slider to ui.R, e.g.

sliderInput("dummySlider", "some value", min=0, max=100, value=10) 

is sufficient to solve the issue.
A temporary workaround is then to add an invisible slider to ui.R

conditionalPanel
   (
   condition = '0==1',
   sliderInput("dummyslider", "", min=0, max=1, value=0)
   )

It seems this somehow derives from missing CSS instances, but I can't really understand why.
Even in the first case FireBug reports http://localhost:8100/shared/slider/css/jquery.slider.min.css as being loaded.
When a second slider is added in ui.R a second instance of the CSS appears (same file, just a second instance).
A third instance is not added if adding other sliders in ui.R but a new instance of the CSS is loaded for every reactiveUI slider component. If you add a few reactiveUI sliders you can see them rendered as text inputs and then change into sliders one by one as the CSS files get loaded.

So I guess what happens is that if a slider is present in ui.R the slider CSS is loaded.
Reactive UI components then load the CSS if needed.

Note that if you just add two reactiveUI sliders and no slider in ui.R the problem reappears, but two instances of the CSS are loaded anyway.

When I try to share a shiny app using a Gist or run the example Gist on a Windows XP machine I get the following error:

shiny::runGist('3239667')
cygwin warning:
  MS-DOS style path detected: C:\DOCUME~1\knowlje\LOCALS~1\Temp\Rtmp0eRcpP\shinygist16181a2427d.tar.gz
  Preferred POSIX equivalent is: /cygdrive/c/DOCUME~1/knowlje/LOCALS~1/Temp/Rtmp0eRcpP/shinygist16181a2427d.tar.gz
  CYGWIN environment variable option "nodosfilewarning" turns off this warning.
  Consult the user's guide for more details about POSIX paths:
    http://cygwin.com/cygwin-ug-net/using.html#using-pathnames
Error in setwd(appDir) : cannot change working directory

I am wondering if you need RTools installed to be able to run from a Gist on Windows machines?

IE 9 and earlier do not support the WebSockets protocol, which we currently have a hard dependency on.

Is there any way to allow pass through of the tar variable to the untar call on runGist? On Windows, sometimes the untar function does not behave so well (http://www.worthinstalling.com/2006/06/command-line-compression-for-windows.html).

maybe

runGist <- function (gist, port = 8100L, launch.browser = getOption("shiny.launch.browser",
interactive(), ...))
...
dirname <- untar(filePath, list = TRUE, ...)[1]
...

Hello,

If one moves < tabPanel("Plot", plotOutput("plot"))> from the first position to the second or third in the list of arguments of tabsetPanel() function the histogram plot does not show up. Consequently. only one plot in the first position is allowed.

Thank you

Andrzej

mainPanel(
tabsetPanel(
tabPanel("Plot", plotOutput("plot")), # plot
tabPanel("Summary", verbatimTextOutput("summary")),
tabPanel("Table", tableOutput("table"))
)
)

sessionInfo()
R version 2.15.1 (2012-06-22)
Platform: x86_64-pc-mingw32/x64 (64-bit)

locale:
[1] LC_COLLATE=English_United States.1252 LC_CTYPE=English_United States.1252
[3] LC_MONETARY=English_United States.1252 LC_NUMERIC=C
[5] LC_TIME=English_United States.1252

attached base packages:
[1] stats graphics grDevices utils datasets methods base

other attached packages:
[1] shiny_0.1.2 xtable_1.7-0 RJSONIO_0.98-1 websockets_1.1.4 digest_0.5.2 caTools_1.13
[7] bitops_1.0-4.1

loaded via a namespace (and not attached):
[1] tools_2.15.1

This will allow Shiny hosting environments to pass custom messages from the server and handle them on the client.

After looking at the Twitter Bootstrap controls you are developing from/with, there are several I would probably use (e.g., accordian, tooltip)...one thing that is missing (as far as I can tell) is a method to input a vector of variable length. This is something I need for several input items to gsDesign...any method or workarounds would be welcome.

Given a Shiny directory you should be able to pass a list of inputs and get back a list of output values, for automated testing purposes.

I want to create a multi-tab multi-dataset package. As the different datasets are not overly related, I want the user to be able to select which they want to look at and for that to change what filtering options are available for them to then use.

So I want something like this,

selectInput("variable", "Variable:",
list(""Cylinders" = "cyl",
"Transmission" = "am"),

if (selectInput == "Transmission") {

sliderInput("integer", "Integer:",
min=0, max=1, value=0) },

else{

Decimal interval with step value
sliderInput("decimal", "Decimal:", 
            min = 0, max = 1, value = 1) }

how do I do a conditional in Shiny?

I tried using the gist example 4050398 (https://gist.github.com/4050398#comments) without success.
See my comments on this gist page.

Best regards,

Yann

Trying to use in the same ui a submitButton and a fileInput, it seems that when a file selection is done, the submitButton is not taken into account, that is to say, the output are processed immediatly.

Even if it could be taken as a bug, I think it should be more convenient to have an explicit event handling method for the submitButtton.
Something like a boolean which says if the last event was the press on this button (false if it was any other widget which was updated) ... But I don't know if it is possible with the shiny events policy...
It should also allow to use several buttons on the same ui.

Best regards,
Yann

• Previous and next going to top of page
• Source code should be a bit more narrow (or we should take less room for the sidebar)

Current, options for print.xtable are ignored. I believe the fix is to change this code:

return(paste(capture.output(print(xtable(data, ...), 
            type = "html", html.table.attributes = paste("class=\"", 
                htmlEscape(classNames, TRUE), "\"", sep = ""))), 
            collapse = "\n"))

to

return(paste(capture.output(print(xtable(data, ...), 
            type = "html", html.table.attributes = paste("class=\"", 
                htmlEscape(classNames, TRUE), "\"", sep = ""), ...)), 
            collapse = "\n"))

In particular this change would allow reactiveTable(fn, include.rownames=FALSE) to work.

Following is a simple, limited interface to the gsDesign R package (ui and server code below).
This requires the gsDesign package available at CRAN.
This project has been open source since its inception in 2006 and is NOT for profit.
I am trying to replace an interface developed in Qt so that it can be maintained and extended by an R programmer.
See http://biopharmnet.com/wiki/GsDesign_Package if you want to see the Qt interface.
Any comments on this are welcome!

Comments:

1. shiny is VERY accessible for an R programmer who is a statistician (read: NOT a professional programmer!).
   ~8 hours to learn shiny and do the interface so far is not bad!
   This could make somewhat complex packages such as gsDesign easily accessible!
2. I am running this using Chrome 21.0.1180.79 m, R 2.15.0 and a version of shiny downloaded on August 18, 2012.

Initial issues with the current interface:

1. I don't understand why the plot subtab under the "Group Sequential Design" tab is not functioning.
   Also, if I want to use ggplot2, how do I get a plot to output (I pretty recent release of ggplot2)?
2. Look at the sample size tab in the UI. It would be ideal if I could use radio buttons or tabs
   (instead of both) to get the functionality implemented here (I would prefer just tabs, but on the server side,
   I don't have access to which tab is selected, so I had to use a radio buttons; radio buttons, on the other hand,
   do not allow the dynamic ui side that tabs do).
3. Supporting variable length vector input (issue #14) is needed for this application. In the timing tab,
   I need to have a vector of the length specified there...and will want increasing numbers in
   that vector between 0 and 1.
   This functionality will be used in two other instances in a more complete interface.
4. One of the nice things about the current Qt interface is that I can compare multiple designs.
   This would require storing all of the parameters for each design, as well as some way to
   navigate between them.
5. See the "Timing" input on the spending function tab. It would be nice to have control over
   input increments in numericInput.

ui.r

library(shiny)
library(gsDesign)
shinyUI(pageWithSidebar(

  # Application title
  headerPanel("gsDesign Explorer"),

  # Sidebar with a slider input for spending fn parameter and
  # time to evaluate spending function
  sidebarPanel(
    tabsetPanel(
      tabPanel("Sample Size",
               numericInput("alpha",label="alpha (1-sided):",value=.025,min=.00001,max=.39999),
               numericInput("Power",label="Power:",value=.9,min=.4,max=.999),
               radioButtons("dist",  "Endpoint type", list("User defined" = "User", "Binomial" = "Binomial")),
               tabsetPanel(
                 tabPanel("User input", 
                          numericInput(inputId="nfix",label="N for Fixed Design", value=1, min=1)),
                 tabPanel("Binomial", 
                          numericInput(inputId="p1",label="Control event rate", value=.15, min=.005,max=.995),
                          numericInput(inputId="p2",label="Experimental event rate", value=.1, min=.005,max=.995),
                          numericInput(inputId="delta0",label="H0 difference",value=0)))),                  
      tabPanel("Timing",
        numericInput(inputId="km1", label="Number of interim analyses", value=2, min=1)),
      tabPanel("Spending Functions",
        tabsetPanel(
          tabPanel("Upper spending",
            sliderInput(inputId="rho", label="Rho:", min = 0.1, max = 15.0, value = 4.0, step=.05),
            numericInput(inputId="t", label="Timing:", value=.5, min=0, max=1)),
          tabPanel("Lower spending",
            sliderInput(inputId="lrho", label="Rho:", min = 0.1, max = 15.0, value = 2.0, step=.05),
            numericInput(inputId="lt", label="Timing:", value=.5, min=0, max=1)))))
  ),
  # Show a plot of the generated distribution
  mainPanel(
    tabsetPanel(
      tabPanel("Upper spending",plotOutput("sfPlot")),
      tabPanel("Fixed design sample size",verbatimTextOutput("dist")),
      tabPanel("Group sequential design",
        tabsetPanel(
          tabPanel("Text",verbatimTextOutput("design")),
          tabPanel("Plot",plotOutput("plot"))
))))))

server.r

library(shiny)
library(ggplot2)
shinyServer(function(input, output){

  # Function that generates a plot of a spending function. The function
  # is wrapped in a call to reactivePlot to indicate that:
  #
  #  1) It is "reactive" and therefore should be automatically 
  #     re-executed when inputs change
  #  2) Its output type is a plot 
  #
  output$sfPlot <- reactivePlot(function(){
    # generate a power spending function
    tt <- (0:100)/100
    alpha <- input$alpha
    spend <- sfPower(alpha,input$t,input$rho)$spend
    plot(tt, sfPower(alpha,tt,input$rho)$spend,
         xlab="Timing", ylab="Spend", main="Power Spending Function", type="l")
    points(input$t,spend,cex=2)
    if (spend > alpha/2) delt <- -alpha/20
    else delt <- alpha/20
    text(input$t,spend+delt,as.character(round(spend,4)))
  })
  N <- reactive(function(){
                   switch(input$dist, 
                      "User" = input$nfix,
                      "Binomial" = nBinomial(p1=input$p1,p2=input$p2,delta0=input$delta0,alpha=input$alpha,beta=1-input$Power)
                   )})
  output$dist <- reactivePrint(function(){N()})
  output$design <- reactivePrint(function(){
    gsDesign(k=input$km1+1,n.fix=N(),sfu=sfPower,sfl=sfPower,sfupar=input$rho,sflpar=input$lrho,alpha=input$alpha,
             beta=1-input$Power)})
  output$plot <- reactivePlot(function(){
    plot(gsDesign(k=input$km1+1,n.fix=N(),sfu=sfPower,sfl=sfPower,sfupar=input$rho,sflpar=input$lrho,alpha=input$alpha,
                  beta=1-input$Power),base=TRUE)
  })
})

I've tried custom HTML UI with example 8 and also with my own bare bones example. I get the following:

runExample("08_html")

Listening on port 8100
Error: argument "handler" is missing, with no default

I'm using shiny 0.1.1, websockets 1.1.4 on R 2.15.1 on Mac OS X 10.6.8.

When setting up radioButtons, for example:

radioButtons("first", "Plot Type 1", c("p"="p","l"="l","g"="g"), selected="g")

the buttons shown in the interface does not change to "g", but continues to default to the first item in the list.

We've had multiple user requests for submit buttons that only affect certain controls.

singleton and tags$head are designed to execute while the page is rendering for the first time and don't currently work in reactiveUI mode. Would need to put explicit support within reactiveUI to fix tags$head, where reactiveUI output that resulted in stuff being written to the head could be pushed there. And for singleton, maybe calculate a hash for each singleton and make that part of the HTML of the page somehow. The init call back to the server could include a list of all singleton hashes that have been calculated so far, this could be stored in shinyapp.

Hi -
example2 and cars/miles-per-gallon fail on some platforms
with e.g. "object 'mtcars' not found"

works OK with explicit "data(mtcars)"

Problem platform = the "Raspbian wheezy" http://www.raspbian.org/ (2012-09-18) on this hardware
http://www.element14.com/community/groups/raspberry-pi

Apart from that minor gotcha: impressive stuff, thanks!

Cheers, Adam

I've got an app where I'm displaying the same data different ways on different tags. Below the output graphs, I've got this trivial item:

h3(textOutput("n"))

defined by:

  output$n <- reactiveText(function() {
    paste("n =", nrow(this_dat()))
  })

Works great if I have a single tab. If I have two tabs, each of which have the above (but with different plotOutput objects), then the Javascript crashes because of the duplicate object names.

The workaround is to duplicate the above with n2. Not a big deal, once I figured out the problem, but it'd be great if there were some way to have the same output objects on multiple tabs without error. At a minimum, a better error message would be very helpful.

Thanks!

The tutorial page http://rstudio.github.com/shiny/tutorial/#building-outputs contains the following bug:

< used instead of <- below:

reactiveTimeSeries <- function(func) {
reactive(function() {
val < func()
list(start = tsp(val)[1],

Should read

reactiveTimeSeries <- function(func) {
reactive(function() {
val <- func()
list(start = tsp(val)[1],

In this example, neither output is rendered at all.

shinyUI(pageWithSidebar(
  headerPanel("Header Panel"),
  sidebarPanel(    
    uiOutput("choiceControl")),
  mainPanel(
    verbatimTextOutput("main"))
))

shinyServer(function(input, output) {  
  output$choiceControl <- reactiveUI(function() {
    choice.names <- c("λ=0.1","λ=0.5","λ=0.9")
    choice.values <- as.list(1:3)
    names(choice.values) <- choice.names
    selectInput("choice", "Choose λ", choice.values)
  })
  output$main <- reactivePrint(function() {
    "Main Panel"
  })
})

JSON unexpected character error when empty data frame is supplied to custom outputs.

When a data frame is supplied that has neither rows NOR columns.

render called instead of error?

workaround: check for empty data frames and supply a data frame with the correct columns but no rows.

Note: can we get documentation on how best to throw errors in both the R and javascript sides?

Thanks for this AWESOME tool. My life just got way better.

Any chance you will create a capability to have tabs in the input window to allow a more complex input structure?
-I would love to consider re-writing gsDesignExplorer interface (http://biopharmnet.com/wiki/GsDesign_Package) using Shiny (previously done using Qt...but apparently Qt is not working with R for Windows?????).

Comment: it has been pretty easy to do a simple screen! Thanks!