Heroku部署Shiny App：用base R/dplyr替代rowr::cbind.fill方案咨询

Since you're hitting R version compatibility issues with rowr on Heroku, let's walk through how to replicate the behavior of rowr::cbind.fill using tools that are more widely supported (base R or dplyr/purrr).

First, let's recap what rowr::cbind.fill does: it takes multiple vectors (or data frames) of different lengths, pads the shorter ones with a fill value (default NULL, which becomes NA in practice) to match the longest input, then combines them column-wise into a data frame.

Option 1: Base R Implementation #

You can replicate this logic directly in base R without any extra packages. Here's a function that mirrors the core behavior of cbind.fill:

cbind_fill_base <- function(..., fill = NA) {
  # Collect all input vectors/data frames into a list
  inputs <- list(...)
  
  # Calculate the maximum length across all inputs
  max_length <- max(sapply(inputs, function(x) if(is.data.frame(x)) nrow(x) else length(x)))
  
  # Pad each input to the maximum length with the fill value
  padded_inputs <- lapply(inputs, function(x) {
    if (is.data.frame(x)) {
      # If input is a data frame, pad rows with fill
      pad_rows <- max_length - nrow(x)
      if (pad_rows > 0) {
        padded_rows <- as.data.frame(lapply(x, function(col) rep(fill, pad_rows)))
        rbind(x, padded_rows)
      } else {
        x
      }
    } else {
      # If input is a vector, pad with fill to max length
      c(x, rep(fill, max_length - length(x)))
    }
  })
  
  # Combine all padded inputs into a single data frame
  do.call(cbind.data.frame, padded_inputs)
}

Example Usage: #

# Test with vectors of different lengths
vec1 <- 1:3
vec2 <- 4:6
vec3 <- 7:8

cbind_fill_base(vec1, vec2, vec3)
#   vec1 vec2 vec3
# 1    1    4    7
# 2    2    5    8
# 3    3    6   NA

This function handles both vectors and data frames as inputs, just like rowr::cbind.fill. It’s lightweight and requires no additional packages—perfect for avoiding compatibility headaches on Heroku.

Option 2: dplyr + purrr Implementation #

If you’re already using the tidyverse in your Shiny app, this concise version uses dplyr and purrr to replicate the behavior:

library(dplyr)
library(purrr)

cbind_fill_dplyr <- function(..., fill = NA) {
  inputs <- list(...)
  
  # Get the maximum length across all inputs
  max_length <- inputs %>% 
    map_int(function(x) if(is.data.frame(x)) nrow(x) else length(x)) %>% 
    max()
  
  # Pad each input and convert to a tibble column
  padded_inputs <- inputs %>% 
    map(function(x) {
      if (is.data.frame(x)) {
        pad_rows <- max_length - nrow(x)
        if (pad_rows > 0) {
          x %>% add_row(!!!set_names(rep(fill, ncol(x)), colnames(x)))
        } else {
          x
        }
      } else {
        tibble(col = x) %>% add_row(col = rep(fill, max_length - length(x)))
      }
    })
  
  # Combine all columns into one tibble
  bind_cols(padded_inputs)
}

Example Usage: #

# Same test vectors as before
cbind_fill_dplyr(vec1, vec2, vec3)
# # A tibble: 3 × 3
#   col1  col2  col3
#   <int> <int> <int>
# 1     1     4     7
# 2     2     5     8
# 3     3     6    NA

This version returns a tibble (tidyverse’s enhanced data frame), but you can convert it to a base data frame with as.data.frame() if needed.

Key Notes for Heroku Deployment #

• Both implementations avoid the rowr package entirely, so you won’t run into R version compatibility issues with Heroku’s buildpacks.
• The base R version is dependency-free, keeping your deployment lightweight.
• The dplyr/purrr version integrates seamlessly if you’re already using tidyverse tools in your Shiny app.

内容的提问来源于stack exchange，提问作者J.Doe