Argthtjtr

I'm trying to build a shiny app very closely based on a previous one I have (which works), albeit with a different (though similarly structured) data object underneath. After changing all the variable and object names, I got the following error:

Warning: Error in filter_impl: Evaluation error: $ operator is invalid
for atomic vectors.

I found a couple of seemingly relevant pages in my searches online:

$ operator is invalid for atomic vectors :: R shiny

https://github.com/rstudio/shiny/issues/1823

Unfortunately the $ operator error seems to be quite a general error message, and neither of these seemed to specifically address my problem. After some tinkering paring back various elements, I found I could render the plots and tables in the app, provided I didn't use attempt to filter on any of the input fields.

For instance, the following ouptut worked, fine, including the switch that allowed me to turn the whole plot on and off, and alter the heading of the graph with a text input called filter1.

output$emoPlot <- renderPlotly({



if(input$prefilterswitch == "OFF"){

  df <- dtm_EMO %>% 

    clusterer(input$clusters)



  plot <- df %>% 

    left_join(EMO_ALL, by = "Work_Order") %>% 

    ggplot(aes(x = date, y = as.factor(cluster), col = MILL, shape = as.factor(PART), text = Equipment_Description_Line_1, text2 = Work_Order))+

    geom_point()+

    guides(col = "none")+

    ggtitle(label = input$filter1)

  ggplotly(plot, tooltip = c("x", "y", "col", "text", "text2", "size"))



}



})

However, if I add line that seeks to filter on one of those inputs, like this:

output$emoPlot <- renderPlotly({





if(input$prefilterswitch == "OFF"){

  df <- dtm_EMO %>% 

    filter(str_detect(combined, input$prefilterswitch %>% tolower()) ==T) %>%

    clusterer(input$clusters)



  plot <- df %>% 

    left_join(EMO_ALL, by = "Work_Order") %>% 

    ggplot(aes(x = date, y = as.factor(cluster), col = MILL, shape = as.factor(PART), text = Equipment_Description_Line_1, text2 = Work_Order))+

    geom_point()+

    guides(col = "none")+

    ggtitle(label = input$filter1)

  ggplotly(plot, tooltip = c("x", "y", "col", "text", "text2", "size"))



}



})

Then I get my error, and no plot:

Warning: Error in filter_impl: Evaluation error: $ operator is invalid
for atomic vectors.

I've had a play with rending a table output, using renderTable aswell, and get the same problem. I can use inputs, but not with functions like filter(), and mutate() also has the same problem.

I suspected that perhaps the issue was to do with the inputs, but all of the switches and text fields render find in the app, and they seem to work, just not with those functions.

That's about as much as I've been able to narrow it down. It's a little frustrating since the ability to apply multiple filters is quite important to the purpose of the app. Any help would be appreciated!

I am pretty sure that the problem is the use of non standard evaluation in filter. In

 filter(str_detect(combined, input$prefilterswitch %>% tolower()))

you have input$prefilterswitch ... which normally in filter would be invalid because it is not just giving a column name from combined (even though we actually know that it is. I think there is probably a check in filter that automatically throws an error if there is a $. My usual solution to this is to create the object prior to starting your piping, so something like

 prefilterswitch <- input$prefilterswitch

And then reference that in the filter statement. Also with a Boolean you don't need ==T.
Alternatively you can go into rlang.