Shiny flexdashboard 中的 facet_grid 给出错误“分面变量必须至少有一个值”

Question

我在获取facet_grid用于评估系统的 ggplot2图时遇到了一些麻烦。绘图呈现良好，但我在浏览器和控制台中收到以下错误：

错误：分面变量必须至少有一个值

每次我根据 input 切换品牌条目时都会发生这种情况input$brand。应用程序不会崩溃，但错误消息很烦人。

我准备了这个可重现的例子：

---
title: "Power ranking for mtcars"
runtime: shiny
output:
  flexdashboard::flex_dashboard:
    orientation: rows
    source_code: embed
---



```{r rows.print = 25}
library(dplyr)
library(ggplot2)

mtcars_tidy <- mtcars %>% 
    tibble::rownames_to_column() %>% 
    rename(model = rowname) %>% 
    mutate(brand = gsub( " .*$", "", model )) %>% 
    mutate(model = model) %>% 
    select(brand, model,  everything())  %>% 
    tidyr::gather(key = 'measure', value = "value", mpg:carb) %>%
    mutate(ranking = as.factor(sample(x = c(1, 2, 3), size = n(), replace = TRUE))) %>%

    mutate(power = case_when(
        .$measure == "hp" & value > 200 | (.$measure == "cyl" & value == 8) ~ "high",
        .$measure == "hp" & value < 200 | (.$measure == "cyl" & value == 8) ~ "medium",
        .$measure == "hp" & value > 100 | (.$measure == "cyl" & value == 6) ~ "high",
        .$measure == "hp" & value < 100 | (.$measure == "cyl" & value == 6) ~ "medium",
        .$measure == "hp" & value > 50  | (.$measure == "cyl" & value == 6) ~ "high",
        .$measure == "hp" & value < 50  | (.$measure == "cyl" & value == 6) ~ "medium",

        .$measure == "hp" & value > 200 | (.$measure == "carb" & value >  4) ~ "high",
        .$measure == "hp" & value < 200 | (.$measure == "carb" & value <= 4) ~ "medium",
        .$measure == "hp" & value > 100 | (.$measure == "carb" & value >  2.8) ~ "high",
        .$measure == "hp" & value < 100 | (.$measure == "carb" & value <= 2.8) ~ "medium",
        .$measure == "hp" & value > 50  | (.$measure == "carb" & value > 2) ~ "high",
        .$measure == "hp" & value < 50  | (.$measure == "carb" & value <= 2) ~ "medium",
        TRUE ~ "low"
    )) 
```

# Sidebar {.sidebar data-width="350"}

```{r}
selectInput("brand", "Brand of the car", 
            choices = unique(mtcars_tidy$brand))

renderUI({
    selectInput("model", "Car model",
                choices = mtcars_tidy$model[mtcars_tidy$brand == levels(mtcars_tidy$brand)[1]])
})

br()

observe({
    brand <- input$brand
    updateSelectInput(session, "model", 
                      choices = mtcars_tidy$model[mtcars_tidy$brand == brand])
})    


# when switching the brand of the car, input$brand this error pops up:
# Error in : Faceting variables must have at least one value
```


# Main

##

### Plot power ranking for each measure

```{r}
nameorder <- make.unique(mtcars_tidy$measure[order(mtcars_tidy$power, mtcars_tidy$ranking)])
mtcars_tidy$measure <- factor(mtcars_tidy$measure, levels=nameorder, 
                                   ordered = TRUE)

dataset <- reactive({
    subset(mtcars_tidy, brand == input$brand & model == input$model) 
})


renderPlot({
    ggplot(dataset(), aes(x = ranking, y = measure)) +
        geom_segment(aes(yend = measure), xend=0, color = "grey50") +
        geom_point(size = 3, aes(colour = power)) +
        scale_colour_brewer(palette="Set1", limits = c("high","medium", "low")) +
        theme_bw() +
        theme(panel.grid.major.y = element_blank()) +   # No horizontal grid lines
        facet_grid(power ~ ., scales="free_y", space="free_y") +
        ggtitle(paste0("Brand: ", input$brand, ", Model: " , input$model))
})    
```

编辑1：我改变facet_grid到facet_wrap，但错误依然存在。

编辑2：根据建议，我切换到facet_wrap这个公式：p <- p + facet_wrap(power ~ .)。还是一样的错误。我也试过这个其他公式 p <- p + facet_wrap(power ~ ranking)。错误仍然存​​在。

编辑 3：在facet_wrap函数上，我也尝试使用这些公式：

• facet_wrap(~power )
• facet_wrap(vars(power ))
• facet_wrap(vars(power , ranking)).

错误仍然相同（相同）。没有变化（Error in : Faceting variables must have at least one value）。

编辑 4：如果我尝试使用facet_wrap(power)，错误会更糟，因为用这一口使 Shiny 崩溃：

Error: Column `function (lambda = 1) \n{\n    if (!is.numeric(lambda) || is.na(lambda)) \n        stop("invalid argument 'lambda'")\n    if (lambda <= 0) \n        return(make.link("log"))\n    if (lambda == 1) \n        return(make.link("identity"))\n    linkfun <- function(mu) mu^lambda\n    linkinv <- function(eta) pmax(eta^(1/lambda), .Machine$double.eps)\n    mu.eta <- function(eta) pmax((1/lambda) * eta^(1/lambda - \n        1), .Machine$double.eps)\n    valideta <- function(eta) all(is.finite(eta)) && all(eta > \n        0)\n    link <- paste0("mu^", round(lambda, 3))\n    structure(list(linkfun = linkfun, linkinv = linkinv, mu.eta = mu.eta, \n        valideta = valideta, name = link), class = "link-glm")\n}` must be a 1d atomic vector or a list

Answer 1

当其中dataset()没有行时会发生此错误。当我运行您的代码（带有 的当前版本facet_grid(power ~ .,）时，它实际上工作正常。当我选择一个新品牌时，在input$model列表更新时显示此错误的间隔很短。一旦这样做，以及组合brand和model返回行，该图就会显示得很好。

您可以通过使用req延迟渲染图直到满足某些要求来防止这种差距。只需在您的顶部插入以下代码renderPlot

req(nrow(dataset()) > 0)

renderPlot如果dataset()不包含至少一行，这将阻止运行。在这种情况下，绘图将只是空白（删除可怕的错误消息），直到数据准备好使用。添加该行后，您的应用程序似乎运行良好（顺便说一下，看起来还不错）。

您可以通过在shiny上下文之外测试您的代码来查看该错误消息的来源。这是您的情节的最小示例：

ggplot(dataset, aes(x = ranking, y = measure)) +
        geom_segment(aes(yend = measure), xend=0, color = "grey50") +
        geom_point(size = 3, aes(colour = power)) +
        facet_grid(power ~ ., scales="free_y", space="free_y")

当我dataset使用此调用时：

dataset <- subset(mtcars_tidy, brand == 'Honda' & model == 'Honda Civic')

绘图正确呈现。当我使用subset不返回任何行的 a 时：

dataset <- subset(mtcars_tidy, brand == 'Honda' & model == 'Civic')

我得到你同样的错误：

Error: Faceting variables must have at least one value