根据sql、r或python中的不同参数计算列之间的结果

Question

大家好，我有这个问题，我正在计算日期之间的天数，前提如下：

• 状态A是基准日期，所有计算都必须以这个日期为参考（按ID分组）
• 对于状态 B、C、D，我必须选择较早的日期
• 我必须计算天数并显示在不同的列中

例如

用 R 生成表格

ColID = c(1, 1, 1, 1, 1, 2, 2, 2)        
ColStatus = c("A", "B", "B", "C", "D", "A", "C", "C")
ColDate = c("01/01/2018","02/03/2018", "05/04/2018", "04/05/2018", "04/05/2018", "02/01/2018", "04/03/2018", "05/04/2018")
data.frame(ColID, ColStatus, ColDate)

我如何进行计算

For ColID = 1

Status A = 01/01/2018
Status B (I Have to select the older one) = 02/03/2018
Status C = 04/05/2018
Status D = 04/05/2018

ResultColB = 02/03/2018 - 01/01/2018 = 60
ResultColC = 04/05/2018 - 01/01/2018 = 123
ResultColD = 04/05/2018 - 01/01/2018 = 123

结果表（以天为单位）

用 R 生成表格

ColID = c(1,2)
ResultColStatusB = c(60,0)
ResultColStatusC = c(123,61)
data.frame(ColID, ResultColStatusB, ResultColStatusC, ResultColStatusB)

这个问题可以用 R、Python 或 SQL 解决，有什么建议我来解决这个问题？

Answer 1

这是 R 中的 tidyverse 解决方案：

library(lubridate)
library(tidyverse)

df %>%
  group_by(ColID, ColStatus) %>% 
  summarise(min_date = min(parse_date_time(ColDate, "%d/%m/%Y"))) %>%
  group_by(ColID) %>%
  summarise(a_b = as.period(interval(min_date[ColStatus=="A"], 
                                     min_date[ColStatus=="B"])) %/% days(1) - 1,
            a_c = as.period(interval(min_date[ColStatus=="A"], 
                                     min_date[ColStatus=="C"])) %/% days(1) - 1,
            a_d = as.period(interval(min_date[ColStatus=="A"], 
                                     min_date[ColStatus=="D"])) %/% days(1) - 1) %>%
mutate_all(funs(if_else(is.na(.), 0, .)))

输出：

  ColID   a_b   a_c   a_d
  <dbl> <dbl> <dbl> <dbl>
1    1.   60.  123.  123.
2    2.    0.   61.    0.

解释：

1. 将ColDate转换为日期类型，获取每个ColStatus的最早日期
2. 计算A 和彼此ColStatus 值之间的时间间隔（以天为单位）。
3. 将NA 值转换为0。

---

R 中使用data.table 的另一种方法：

setDT(df)
#convert string Dates to integer
df[, ColDate := as.numeric(as.Date(ColDate, format="%d/%m/%Y"))]
cols <- c("B","C","D")
#pivot table without A into a wide table
dcast(df[ColStatus!="A"], ColID ~ ColStatus, min, fill=NA_integer_, value.var="ColDate")[ 
    #join with table containing only A and do the differencing
    df[ColStatus=="A"], on=.(ColID),
    (cols) := lapply(.SD, function(x) x - i.ColDate), .SDcols=cols]

---

或者使用baseR:

df$ColDate <- as.integer(as.Date(df$ColDate, format="%d/%m/%Y"))
cols <- c("B","C","D")
by(df, df$ColID, function(x) {
    aDate <- x$ColDate[x$ColStatus=="A"]
    vapply(cols, 
        function(id) if(any(x$ColStatus==id)) min(x$ColDate[x$ColStatus==id]) - aDate 
            else NA_integer_, 
        integer(1))
})

---

数据：

ColID = c(1, 1, 1, 1, 1, 2, 2, 2)        
ColStatus = c("A", "B", "B", "C", "D", "A", "C", "C")
ColDate = c("01/01/2018","02/03/2018", "05/04/2018", "04/05/2018", "04/05/2018", "02/01/2018", "04/03/2018", "05/04/2018")
df <- data.frame(ColID, ColStatus, ColDate)

Answer 2

Python：

import pandas as pd 
Dic = {'ColID': [1, 1, 1, 1, 1, 2, 2, 2],
       'ColStatus': ["A", "B", "B", "C", "D", "A", "C", "C"], 
       'ColDate': ["01/01/2018", "02/03/2018", "05/04/2018", "04/05/2018",
        "04/05/2018", "02/01/2018", "04/03/2018", "05/04/2018"]} 
df = pd.DataFrame(Dic) 
df.ColDate = pd.to_datetime(df.ColDate, format='%d/%m/%Y') 
conditions = [df.ColID==n for n in df.ColID.unique()] 
choices = [df[(df.ColID==n) & (df.ColStatus=='A')]['ColDate'].min() for n in df.ColID.unique()] 
df['Amin'] = pd.np.select(conditions, choices) 
df['days'] = df.ColDate - df.Amin 
df = df[df['days'].dt.days>0]