Summing over rows containing particular strings in R - r

Question

I have a dataframe where the first column contains names of campaigns. I need to sum up all rows where the campaign names contain certain strings (it can appear in different places within the name, i.e. sometimes in the beginning sometimes in the end). The dataframe looks something like this:
Campaign Impressions
1 Local display 1661246
2 Local text 1029724
3 National display 325832
4 National Audio 498900
5 Audio local 597339
6 TV Regional 597339
...
So in this case I want to sum up all rows containing "local" in to one row, "national" into one, "regional" into one etc, like this:
Campaign Impressions
1 Local 939293929
2 National 9232423423
2 Regional 1123123123
How can this be achieved? I've been trying with ddply without success....

Answer 1

You could use grep to find the rows that match the Campaign column categories ('Local', 'National', 'Regional') in a loop (lapply). Subset the dataset ('df') based on grep and sum the 'Impressions' column and rbind the list elements.
res1 <- do.call(rbind,lapply(c('Local', 'National', 'Regional'),
function(x) {
x1 <- df[grep(x, df$Campaign, ignore.case=TRUE),]
data.frame(Campaign= x, Impressions=sum(x1$Impressions))}))
Or use data.table. Keep only the 'Local', 'National', 'Region' in the 'Category' using sub and use that as "grouping" variable to sum the column 'Impressions'.
library(data.table)
setDT(df)[, list(Impressions=sum(Impressions)),by=
list(Category=sub('.*?(Local|National|Region).*','\\U\\1', Campaign,
ignore.case=TRUE, perl=TRUE))]
data
df <- structure(list(Campaign = c("Local display", "Local text",
"National display",
"National Audio", "Audio local", "TV Regional"), Impressions =
c(1661246L, 1029724L, 325832L, 498900L, 597339L, 597339L)), .Names =
c("Campaign", "Impressions"), class = "data.frame", row.names = c("1",
"2", "3", "4", "5", "6"))

Answer 2

I guess you should use the grep function : say your data.frame is called mydata then
Local = grep(mydata$Campaign, pattern = "Local")
National = grep(mydata$Campaign, pattern = "National")
Regional = grep(mydata$Campaign, pattern = "Regional")
mydata_sum = data.frame(Campaign = c("Local", "National", "Regional"), Impressions = c(sum(mydata$Impressions[Local]), sum(mydata$Impressions[National]), sum(mydata$Impressions[Regional])))

Answer 3

Here's my approach using dplyr:
library(dplyr)
library(stringr)
categories <- "Local|National|Regional"
mydf %>%
mutate(Campaign = tolower(str_extract((Campaign), ignore.case(categories)))) %>%
group_by(Campaign) %>%
summarise(sum(Impressions))
I needed to add the tolower, after extracting the strings, to make sure the group_by groups "local" together with "Local".