Shiny render glm.fit is not working, the categorical-response cannot be truly readable

Question

I created a shiny app for logistic regression model.

My idea is allow user define formular by typing in textInput box ex. 'Response ~ Pred1 + Pred2' and then convert to formula by function 'as.formula(paste0(input$formula))'. It works in RStudio, but not work in shinydashboard render. The error is 'y values must be 0 <= y <= 1'. Any body help?

This work well in Rstudio

s <- 'Direction ~ Lag1+ Lag2'
model <- glm(as.formula(paste0(s)),
               data = Smarket,
               family = binomial,
               subset = train_data$id)

But in shinydashboard, the formula render from textInput box is not working. My code below

My ui.R

library(tidyverse)
library(shiny)
library(shinydashboard)
library(ISLR2)
ui <- dashboardPage(
  dashboardHeader(title = 'Classification'),
  dashboardSidebar(width = '275',
    sidebarMenu(
      menuItem('Instruction', tabName = 'instruction', icon = icon('hammer', lib = 'font-awesome')),
      menuItem('Table & Correlation', tabName = 'table', icon = icon('table', lib = 'font-awesome')),
      menuItem('DS', tabName = 'ds', icon = icon('desktop', lib = 'font-awesome')),
      
      fileInput(inputId = 'f_input', label = 'Upload File'),
      div(class='col-lg-12 col-md-12',
          hr()),
      
      # Correlation
      uiOutput(outputId = 'corre'),
      actionButton(inputId = 'button1', label = 'Correlation'),
      div(class='col-lg-12 col-md-12',
          hr()),

      # Train-Test
      sliderInput("data_split", label = "Percent Train", min = 0,
                  max = 100, value = 70, step = 5),

      # Fit Model
      textInput(inputId = 'formula',
                label = 'Creat formula',
                placeholder = 'y ~ X1 + X2 + ...'),
      actionButton(inputId = 'button2', label = 'Run Model'),
      div(class='col-lg-12 col-md-12',
          hr())
    )
  ),
  dashboardBody(
    tabItems(
      tabItem(tabName = 'instruction'),
      tabItem(tabName = 'table',
              fluidRow(
                DT::dataTableOutput('view_table')
              ),
              fluidRow(
                plotOutput('corr_plot')
              )),
      tabItem(tabName = 'ds',
              fluidRow(
                box(verbatimTextOutput('summary')),
                box(plotOutput('conf.matrix'))
              )
              )
    )
  )
)

My server.R

server <- function(input, output, session){
  ### RAW MATERIALS
  #upload csv file
  data <- reactive({
    file <- input$f_input 
    data <- if(!is.null(file)){
      read.csv(file$datapath)
    } %>%
      as_tibble()%>%
      mutate(id = row_number())
  })

  #render RAW table
  output$view_table <- DT::renderDataTable(
    data(),
    options = (list(scrollX = TRUE))
  )
  ###-----------------------------------------
  ### CORRELATION
  #user select Xi to generate correlation plot
  output$corre <- renderUI({
    data <- data()
    if(!is.null(data)){
      CHOICES <- names(data)
      shinyWidgets::pickerInput('var',
                                label = 'Select to generated correlation of variables',
                                choices = CHOICES,
                                options = list(`actions-box` = TRUE),
                                multiple = TRUE)
    }
  })
  
  #filter table matched selection from pickerInput
  df <- eventReactive(input$button1, {
    data <- data()
    if(!is.null(data)){
      data %>%
        select(dplyr::matches(input$var)) 
    }
  })
  
  #corr plot
  output$corr_plot <-renderPlot({
    df() %>%
      ggcorr()
  })
  ###-----------------------------------------
  ### TRAIN-TEST DATA SPLIT
  i <- reactive({
    set.seed(112)
    data() %>%
      rsample::initial_split(prop = as.numeric(input$data_split)/100)

  })
  ###-----------------------------------------
  ### FIT MODEL
  model <- eventReactive(input$button2, {
    train_data <- rsample::training(i())
    test_data <- rsample::testing(i())
    
    ##Logistic Regression
    glm(as.formula(paste0(input$formula)),
        data = data(),
        family = binomial,
        subset = train_data$id) #parse index of train
  })
  
  output$summary <- renderPrint({
    if(!is.null(model())){
      summary(model())
    }
  })
}
shinyApp(ui = ui, server = server)