reshaping data frame r

Question

Simple reshaping, I have the following data:

df<-data.frame(Product=c("A","A","A","B","B","C"), Ingredients=c("Chocolate","Vanilla","Berry","Chocolate","Berry2","Vanilla"))
df
Product Ingredients
1   A   Chocolate 
2   A     Vanilla
3   A       Berry
4   B   Chocolate
5   B      Berry2
6   C     Vanilla

I want a column for each unique value of "ingredient", for example:

df2
Product Ingredient_1 Ingredient_2 Ingredient_3
A       Chocolate       Vanilla        Berry
B       Chocolate       Berry2         NULL
C       Vanilla         NULL           NULL

Seems trivial enough, I tried reshape but I keep getting counts (not the actual values of "ingredients"). Ideas?

Answer 1

Here's a possible solution using data.table package

library(data.table)
setDT(df)[, Ingredient := paste0("Ingredient_", seq_len(.N)), Product]
dcast(df, Product ~ Ingredient, value.var = "Ingredients")
#    Product Ingredient_1 Ingredient_2 Ingredient_3
# 1:       A    Chocolate      Vanilla        Berry
# 2:       B    Chocolate       Berry2           NA
# 3:       C      Vanilla           NA           NA

---

Alternavely, we could do this with the sexy dplyr/tidyr combination

library(dplyr)
library(tidyr)
df %>% 
  group_by(Product) %>%
  mutate(Ingredient = paste0("Ingredient_", row_number())) %>%
  spread(Ingredient, Ingredients)

# Source: local data frame [3 x 4]
# 
#   Product Ingredient_1 Ingredient_2 Ingredient_3
# 1       A    Chocolate      Vanilla        Berry
# 2       B    Chocolate       Berry2           NA
# 3       C      Vanilla           NA           NA

Answer 2

In the spirit of sharing alternatives, here are two more:

Option 1 : split the columns and use stri_list2matrix to create your wide form.

library(stringi)
x <- with(df, split(Ingredients, Product))
data.frame(Product = names(x), stri_list2matrix(x))
#   Product        X1        X2      X3
# 1       A Chocolate Chocolate Vanilla
# 2       B   Vanilla    Berry2    <NA>
# 3       C     Berry      <NA>    <NA> 

Option 2 : Use getanID from my "splitstackshape" package to generate an ".id" column, then dcast it. The "data.table" package is loaded with "splitstackshape", so you can directly call dcast.data.table to do the reshaping.

library(splitstackshape)
dcast.data.table(getanID(df, "Product"), 
                 Product ~ .id, value.var = "Ingredients")
#    Product         1       2     3
# 1:       A Chocolate Vanilla Berry
# 2:       B Chocolate  Berry2    NA
# 3:       C   Vanilla      NA    NA

Answer 3

With base R reshape

df$Count<-ave(rep(1,nrow(df)),df$Product,FUN=cumsum)
reshape(df,idvar="Product",timevar="Count",direction="wide",sep="_")

#  Product Ingredients_1 Ingredients_2 Ingredients_3
#1       A     Chocolate       Vanilla         Berry
#4       B     Chocolate        Berry2          <NA>
#6       C       Vanilla          <NA>          <NA>