Creating Large Data Frames

Question

Let's say that I want to generate a large data frame from scratch.

Using the data.frame function is how I would generally create data frames. However, df's like the following are extremely error prone and inefficient.

So is there a more efficient way of creating the following data frame.

df <- data.frame(GOOGLE_CAMPAIGN=c(rep("Google - Medicare - US", 928), rep("MedicareBranded", 2983),
                                   rep("Medigap", 805), rep("Medigap Branded", 1914),
                                   rep("Medicare Typos", 1353), rep("Medigap Typos", 635),
                                   rep("Phone - MedicareGeneral", 585),
                                   rep("Phone - MedicareBranded", 2967),
                                   rep("Phone-Medigap", 812),
                                   rep("Auto Broad Match", 27),
                                   rep("Auto Exact Match", 80),
                                   rep("Auto Exact Match", 875)),                   
                 GOOGLE_AD_GROUP=c(rep("Medicare", 928), rep("MedicareBranded", 2983),
                                   rep("Medigap", 805), rep("Medigap Branded", 1914),
                                   rep("Medicare Typos", 1353), rep("Medigap Typos", 635),
                                   rep("Phone ads 1-Medicare Terms",585),
                                   rep("Ad Group #1", 2967), rep("Medigap-phone", 812),
                                   rep("Auto Insurance", 27),
                                   rep("Auto General", 80),
                                   rep("Auto Brand", 875)))

Yikes, that is some 'bad' code. How can I generate this 'large' data frame in a more efficient manner?

Answer 1

If your only source for that information is a piece of paper, then you probably won't get much better than that, but you can at least consolidate all that into a single rep call for each column:

#I'm going to cheat and not type out all those strings by hand
x <- unique(df[,1])
y <- unique(df[,2])

#Vectors of the number of times for each    
x1 <- c(928,2983,805,1914,1353,635,585,2967,812,27,955)
y1 <- c(x1[-11],80,875)

dd <- data.frame(GOOGLE_CAMPAIGN = rep(x, times = x1), 
                 GOOGLE_AD_GROUP = rep(y, times = y1))

which should be the same:

> all.equal(dd,df)
[1] TRUE

But if this information is already in a data structure in R somehow and you just need to transform it, that could possibly be even easier, but we'd need to know what that structure is.

Answer 2

Manually, (1) create this data frame:

> dfu <- unique(df)
> rownames(dfu) <- NULL
> dfu
           GOOGLE_CAMPAIGN            GOOGLE_AD_GROUP
1   Google - Medicare - US                   Medicare
2          MedicareBranded            MedicareBranded
3                  Medigap                    Medigap
4          Medigap Branded            Medigap Branded
5           Medicare Typos             Medicare Typos
6            Medigap Typos              Medigap Typos
7  Phone - MedicareGeneral Phone ads 1-Medicare Terms
8  Phone - MedicareBranded                Ad Group #1
9            Phone-Medigap              Medigap-phone
10        Auto Broad Match             Auto Insurance
11        Auto Exact Match               Auto General
12        Auto Exact Match                 Auto Brand

and (2) this vector of lengths:

> lens <- rle(as.numeric(interaction(df[[1]], df[[2]])))$lengths
> lens
 [1]  928 2983  805 1914 1353  635  585 2967  812   27   80  875

From these two inputs (dfu and lens) we can reconstruct df (here called df2):

> df2 <- dfu[rep(seq_along(lens), lens), ]
> rownames(df2) <- NULL
> identical(df, df2)
[1] TRUE