1. What is reshape2 and why would you use it?

reshape2 is an R package that lets you change the shape of any dataframe, i.e. to pivot and to “unpivot”.

Keep in mind that if your favourite R package for dataframes manipulation is data.table, functions dcast and melt are already in this package and work exactly the same as those in reshape2.

2. A few simple examples

In fact there are only two functions worth mentioning: dcast, which is equivalent to MS Excel pivot table, and melt, which does the opposite or unpivots a table.

Here’s an example dataframe:

d <- data.frame(
  account_no = paste(rep(7, 5), 1:5, sep=""),
  Jan = rnorm(5, 10, 1),
  Feb = rnorm(5, 10, 2),
  Mar = rnorm(5, 10, 3)
)
print(d)

##   account_no       Jan       Feb       Mar
## 1         71 10.780595 11.831565 13.751443
## 2         72  9.838766  8.252641  8.160033
## 3         73 10.688876 10.686336  5.731527
## 4         74  8.287228  8.267534  6.156336
## 5         75  9.139528 11.398885 10.314572

Transormation into a normalized table (unpivot):

dn <- reshape2::melt(
  data = d, 
  id.vars = "account_no", 
  variable.name = "month", 
  value.name = "revenue"
)
print(dn)

##    account_no month   revenue
## 1          71   Jan 10.780595
## 2          72   Jan  9.838766
## 3          73   Jan 10.688876
## 4          74   Jan  8.287228
## 5          75   Jan  9.139528
## 6          71   Feb 11.831565
## 7          72   Feb  8.252641
## 8          73   Feb 10.686336
## 9          74   Feb  8.267534
## 10         75   Feb 11.398885
## 11         71   Mar 13.751443
## 12         72   Mar  8.160033
## 13         73   Mar  5.731527
## 14         74   Mar  6.156336
## 15         75   Mar 10.314572

And back to the previous format using a pivot:

reshape2::dcast(
  data = dn, 
  formula = account_no ~ month, 
  value.var = "revenue"
)

##   account_no       Jan       Feb       Mar
## 1         71 10.780595 11.831565 13.751443
## 2         72  9.838766  8.252641  8.160033
## 3         73 10.688876 10.686336  5.731527
## 4         74  8.287228  8.267534  6.156336
## 5         75  9.139528 11.398885 10.314572

3. Links

A pretty nice and much longer tutorial is available here.