匹配来自两个数据框的两列，并提供不同的列

Question

我想根据“索引”列匹配两个具有不同尺寸（df1和df2）的数据框。 然后，根据匹配将两列从df2（shift和shiftdate）添加到df1。 但是我需要遵循多个规则。

df1 <- data.frame("Index" = c("Adams10-1", "Adams10-1", "Adams10-2", "Adams10-2", "Ball10-1", "Ball10-2", "Cash10-1", "Cash10-2", "David10-1", "David10-2"),
                    "CaseDate" = c("2005-10-01", "2005-10-01", "2005-10-02", "2005-10-02", "2005-10-01", "2005-10-02", "2005-10-01", "2005-10-02", "2005-10-01", "2005-10-02"),
                    "Type" = c("heart", "local", "knee", "nose", "heart", "foot", "shin", "foot", "spine", "delivery"),
                  "StartTime" = c(1640, 1755, 0112, 0300, 2145, 0233, 2123, 0326, 858, 1024))

df2 <- data.frame("Index" = c("Adams10-1", "Adams10-1", "Ball10-1", "Cash10-1", "David10-1", "David10-1", "David10-3"),
                     "ShiftDate" = c("2005-10-01", "2005-10-01", "2005-10-01", "2005-10-01", "2005-10-01", "2005-10-01", "2005-10-03"),
                  "Shift" = c("OB", "CV", "Night", "Super", "OB", "Day", "OB"),
                  "Multiple" = c("yes", "yes", "no", "no", "yes", "yes", "no"))

规则：

1. 如果df1 $ Index和df2 $ Index AND之间存在匹配项：

   • 如果df2 $ Multiple ==“ no”，则将df2 $ Shift和df2 $ ShiftDate添加到df1

   • 如果df2 $ Multiple ==“ yes”，则给出NA（除非df1 $ Type ==“ heart”＆df2 $ Shift ==“ CV”（在这种情况下，将CV移位和shiftdate从df2添加到df1））

2. 如果df1 $ Index和df2 $ Index之间没有匹配项，则给出NA

   • 除非df1 $ StartTime> 0000和<0700（在这种情况下，请从df1 $ CaseDate前一天的df2 $ shiftdate中添加df2 $ shift和df2 $ shiftdate）

   • 除非df1 $ Type ==“ delivery”＆df2 $ Shift =“ OB”（在这种情况下，请从df1 $ CaseDate之后一天的df2 $ shiftdate中添加df2 $ shift和df2 $ shiftdate）

我想得到以下结果。

df3 <- data.frame("Index" = c("Adams10-1", "Adams10-1", "Adams10-2", "Adams10-2", "Ball10-1", "Ball10-2", "Cash10-1", "Cash10-2", "David10-1", "David10-2"),
                     "CaseDate" = c("2005-10-01", "2005-10-01", "2005-10-02", "2005-10-02", "2005-10-01", "2005-10-02", "2005-10-01", "2005-10-02", "2005-10-01", "2005-10-02"),
                     "Type" = c("heart", "local", "knee", "nose", "heart", "foot", "shin", "foot", "spine", "delivery"),
                     "StartTime" = c(1640, 1755, 0112, 0300, 2145, 0233, 2123, 0326, 858, 1024),
                     "Shift" = c("CV", NA, NA, NA, "Night", "Night", "Super", "Super", NA, "OB"),
                      "ShiftDate" = c("2005-10-01", NA, NA, NA, "2005-10-01", "2005-10-01", "2005-10-01", "2005-10-01", NA, "2005-10-03"))

即使我不能基于所有这些规则来执行此操作，仅在匹配方面获得帮助也会有所帮助。 先感谢您！

Answer 1

我将基于df1和df2的结构以及问题中提供的目标数据集的内容在此做出一些假设。

• Index只是人员标识符（在此为姓名）与班次或病例日期的组合。 所以我们真的很想参加人和约会。
• df1 Multiple仅表示该人一天是否有多次轮班。 （我假设第一个David10-1记录的“ no ”是一个错字。）因此，规则1的确是关于该人一天是否有多次轮班。

如果这两个是正确的，我们可以执行以下操作。 这段代码在很多地方都是多余的。 它可能会收紧很多。 但是它非常明确地显示了规则的逻辑。

library(dplyr)
library(lubridate)

# First, let's do make two changes: (1) convert the dates to real dates, and
# (2) replace Index with Name.
df1 = df1 %>%
  mutate(CaseDate = ymd(CaseDate),
         Name = gsub("[^A-Za-z]", "", Index)) %>%
  select(Name, CaseDate, Type, StartTime)
df2 = df2 %>%
  mutate(ShiftDate = ymd(ShiftDate),
         Name = gsub("[^A-Za-z]", "", Index)) %>%
  select(Name, ShiftDate, Shift)

# Start with df1.
df3 = df1 %>%
  # Bring in matching records in df2.  Filter df2 to records that are either
  # (1) the only record for that person, or (2) CV shifts.
  left_join(df2 %>%
              group_by(Name, ShiftDate) %>%
              mutate(num.shifts = n()) %>%
              filter(num.shifts == 1 | Shift == "CV"),
            by = c("Name", "CaseDate" = "ShiftDate")) %>%
  # We want to keep Shift and ShiftDate for records from df2 that are either
  # (1) the only record for that person, or (2) CV shifts that join to a
  # "heart" type in df1.
  mutate(Shift = case_when(num.shifts == 1 ~ Shift,
                           Type == "heart" & Shift == "CV" ~ Shift,
                           T ~ NA_character_),
         ShiftDate = case_when(num.shifts == 1 ~ CaseDate,
                               Type == "heart" & Shift == "CV" ~ CaseDate)) %>%
  select(Name, CaseDate, Type, StartTime, Shift, ShiftDate) %>%
  # Bring in records in df2 that match on person and whose shift date is the
  # day before the case date.
  left_join(df2 %>%
              group_by(Name, ShiftDate) %>%
              filter(n() == 1) %>%
              mutate(ShiftDateOneDayLater = ShiftDate + 1),
            by = c("Name", "CaseDate" = "ShiftDateOneDayLater")) %>%
  # Keep Shift and ShiftDate only if StartTime is between 0000 and 0700.
  mutate(Shift = case_when(!is.na(Shift.x) ~ Shift.x,
                           StartTime > 0 & StartTime < 700 ~ Shift.y),
         ShiftDate = case_when(!is.na(ShiftDate.x) ~ ShiftDate.x,
                               StartTime > 0 & StartTime < 700 ~ ShiftDate.y)) %>%
  select(Name, CaseDate, Type, StartTime, Shift, ShiftDate) %>%
  # Bring in records in df2 that match on person and whose shift date is the
  # day after the case date.
  left_join(df2 %>%
              group_by(Name, ShiftDate) %>%
              filter(n() == 1) %>%
              mutate(ShiftDateOneDayBefore = ShiftDate - 1),
            by = c("Name", "CaseDate" = "ShiftDateOneDayBefore")) %>%
  # Keep Shift and ShiftDate only if this is a "delivery" case and an "OB"
  # shift.
  mutate(Shift = case_when(!is.na(Shift.x) ~ Shift.x,
                           Type == "delivery" & Shift.y == "OB" ~ Shift.y),
         ShiftDate = case_when(!is.na(Shift.x) ~ ShiftDate.x,
                               Type == "delivery" & Shift.y == "OB" ~ ShiftDate.y)) %>%
  select(Name, CaseDate, Type, StartTime, Shift, ShiftDate)