Function Review

library(tidyverse)
library(data.table)

合并函数到一个文档方便查看。

list.files('180901_靳晓松_函数',full.names = T) %>% 
    as_tibble() %>% 
    mutate(fun = map(value,read_file)) %>% 
    unnest() %>% 
    .$fun %>% 
    paste0(.,paste0(rep('\n',3),collapse = '')) %>% 
    str_flatten() %>% 
    write_file('function_sum.txt')

导入测试数据。

freq_sort

freq_nosort <- function(data, var, fill = "#1f77b4", title='变量频数查询') {
  data %>%
  dplyr::count(eval(parse(text = var))) %>%
  ggplot(aes(x = `eval(parse(text = var))`, y = n)) +
  geom_bar(stat = "identity", fill = fill) +
  geom_text(aes(label = n), vjust = -0.5, size = 3, family = "MYH") +
  labs(x = "", y = "", title = title)
  # eval(parse(text = mytheme_b))
}

data <- test_data
var <- 'bad_good'
fill <- "#1f77b4"
title <- '变量频数查询'
data %>%
  dplyr::count(eval(parse(text = var))) %>%
  ggplot(aes(x = `eval(parse(text = var))`, y = n)) +
  geom_bar(stat = "identity", fill = fill) +
  geom_text(aes(label = n), vjust = -0.5, size = 3, family = "MYH") +
  labs(x = "", y = "", title = title)

  # eval(parse(text = mytheme_b))

1. 每个bar给数据值非常好，但是我建议给有效数字，或者用,分开，见修改后函数。
2. 建议title和theme给default值，不然影响函数调用。比如，每个人都必须要输入title和theme才能调用函数，否则报错，但是这是个trivial的问题。
3. x和y可以不需要删除，建议给default值方便用户设计个性化。
4. 其他大部分函数可以按照这个来修改。

function()
data <- test_data

function()
data <- test_data

var <- 'bad_good'
fill <- "#1f77b4"
title <- '变量频数查询'
data %>%
  dplyr::count(eval(parse(text = var))) %>%
  ggplot(aes(x = `eval(parse(text = var))`, y = n)) +
  geom_bar(stat = "identity", fill = fill) +
  geom_text(aes(label = n), vjust = -0.5, size = 3, family = "MYH") +
  labs(x = "", y = "", title = title)

  # eval(parse(text = mytheme_b))

freq_nosort <- function(data, var, fill = "#1f77b4", title='变量频数查询', theme = 'theme_bw()', x = var, y = '频数') {
library(tidyverse)  
library(formattable)
data %>%
    count(eval(parse(text = var))) %>%
    ggplot(aes(x = `eval(parse(text = var))`, y = n)) +
    geom_bar(stat = "identity", fill = fill) +
    geom_text(aes(label = accounting(n,digits = 0)), vjust = -0.5, size = 3, family = "MYH") +
    labs(x = x, y = y, title = title) +
    eval(parse(text = theme))
}

freq_nosort(test_data,'bad_good')

xvar_pass_cumsum_line

xvar_pass_cumsum_line <- function(data, xvar, yvar, fill = "#1f77b4", title) {
  data %>% 
  dplyr::count(eval(parse(text = xvar)), eval(parse(text = yvar))) %>%
  spread(`eval(parse(text = yvar))`, n) %>%
  mutate(bad = ifelse(bad %in% NA, 0, bad),
         badcumsum  = cumsum(bad),
         goodcumsum = cumsum(good),
         percumsum  = round(badcumsum / (badcumsum + goodcumsum) * 100, 1)) %>%
  ggplot(aes(x = `eval(parse(text = xvar))`, y = percumsum, color = "red")) +
  geom_line(group = 1, size = 1) +
  geom_point(shape = 19, size = 2) +
  geom_text(aes(label = paste0(percumsum, "%")), 
            vjust = -0.5, size = 3, family = "MYH", color = "black") +
  scale_color_manual(values = fill) +
  labs(x = "", y = "", title = title) +
  eval(parse(text = mytheme_b)) +
  theme(axis.text.x = element_text(angle = 60, hjust = 1), legend.position = "none")
}

data <- test_data
var <- 'bad_good'
fill <- "#1f77b4"
data %>% 
    count(type1,bad_good) %>% 
    spread(bad_good,n) %>% 
    mutate(
        bad = ifelse(bad %in% NA, 0, bad)
        ,badcumsum  = cumsum(bad)
        ,goodcumsum = cumsum(good)
        ,percumsum  = round(badcumsum / (badcumsum + goodcumsum) * 100, 1)) %>%
  ggplot(aes(x = type1, y = percumsum, color = "red")) +
  geom_line(group = 1, size = 1) +
  geom_point(shape = 19, size = 2) +
  geom_text(aes(label = paste0(percumsum, "%")), 
            vjust = -0.5, size = 3, family = "MYH", color = "black") +
  scale_color_manual(values = fill)

这种滚动逾期率的思路我可以借鉴下，这个函数看起来没什么问题。 但是可以简化很多，代码如下。

1. cumsum没有na.rm参数，因此需要对good和bad都执行替换，这里最简单的方式是加一条代码mutate_if(is.double,~ifelse(is.na(.),0,.)) %>%
2. is.character替换成as.factor

data <- test_data
var <- 'bad_good'
fill <- "#1f77b4"
data %>% 
    mutate_if(is.character,as.factor) %>% 
    mutate_if(is.double,~ifelse(is.na(.),0,.)) %>% 
    count(type1,bad_good) %>% 
    spread(bad_good,n) %>% 
    mutate(
        badcumsum  = cumsum(bad)
        ,goodcumsum = cumsum(good)
        ,percumsum  = round(badcumsum / (badcumsum + goodcumsum) * 100, 1)) %>%
    ggplot(aes(x = type1, y = percumsum)) +
    geom_line(group = 1, size = 1) +
    geom_point(shape = 19, size = 2) +
    geom_text(aes(label = paste0(percumsum, "%")), 
            vjust = -0.5, size = 3, family = "MYH", color = "black") +
    scale_color_manual(values = fill)