总结这个数据框架的更好方法是什么？

Question

我有一个这样的数据框:

  message.id sender recipient
1          1      A         B
2          1      A         C
3          2      A         B
4          3      B         C
5          3      B         D
6          3      B         Q

我想通过发送方和收件人列中的值计数来总结它,以获得:

  address messages.sent messages.received
1       A             3                 0
2       B             3                 2
3       C             0                 2
4       D             0                 1
5       Q             0                 1

我有工作代码,但它很混乱,我希望有一种方法可以在一个magrittr链中完成所有操作而不是我在下面的操作:

df <- data.frame(message.id = c(1,1,2,3,3,3),
                 sender = c("A","A","A","B","B","B"),
                 recipient = c("B","C","B","C","D","Q"))
sent <- df %>% 
  group_by(sender) %>%
  summarise(messages.sent = n()) %>%
  mutate(address = sender) %>%
  select(address, messages.sent)

received <- df %>% 
  group_by(recipient) %>%
  summarise(messages.received = n()) %>%
  mutate(address = recipient) %>%
  select(address, messages.received)

df_summary <- merge(sent, received, all = TRUE) %>%
  replace(is.na(.), 0)

Answer 1

我们可以用 melt/dcast

library(reshape2)
dcast(melt(df1, id.var='message.id'), value~variable, 
                 value.var='message.id', length)

或者使用包装器 recast

recast(df1, id.var='message.id', value~variable, length)
#    value sender recipient
#1     A      3         0
#2     B      3         2
#3     C      0         2
#4     D      0         1
#5     Q      0         1

如果我们需要使用 dplyr/tidyr

library(dplyr)
library(tidyr)
gather(df1, messages, address, 2:3) %>%
          group_by(messages, address) %>%
          summarise(n=n()) %>% 
          spread(messages, n, fill=0)
#     address sender recipient
#     (chr)  (dbl)     (dbl)
#1       A      3         0
#2       B      3         2
#3       C      0         2
#4       D      0         1
#5       Q      0         1