SecretPlans.org

I thought it would be cool to be able to define some subsets of a data frame without having to specify the query each time, but still have them update dynamically every time the data frame changes. Like views in a database.

Here’s the solution I’m using. The example application is parsing an Apache server log file.

require(plyr)
log <- read.table(file='httpd.combine.20120509')
# in the file I used, there was a space between the time and the time zone, creating two fields.
names(log) <- c('host', 'identity', 'user', 'time' ,'V5','request', 'status', 'bytes','referer','agent')
# Paste the two fields together
log$time <- paste(log$time, log$V5, sep=' ')
# remove the extra field
log$V5 <- NULL

# extract the URIs from the request field
log$uri <- gsub('GET |PROPFIND |HEAD |OPTIONS | HTTP/*.*','',log$request)

# convert the timestamp to something R can work with
log$rtime <- strptime(log$time, '[%d/%B/%Y:%H:%M:%S')

# Identify the obvious bots from the agent field
log$isbot <- grepl (".*bot.*", log$agent)

views <- function(x, subset.code) {
            switch(subset.code,
            'nobots' = subset(x, isbot==TRUE),
            'pdfs' = subset(x, grepl('.*.pdf$', x$uri))
            )
            }

views(log, 'pdfs')

That views function is where the action is. Each subset is specified and assigned a nickname. For example, nobots shows only records that don’t contain the string “bot” in the user-agent, and pdfs shows all the pdfs that were requested.

Rather than copy/pasting subset(log, grepl('.*.pdf$', log$uri)) whenever you want to work with the pdfs, you just write views(log, 'pdfs'). Not a huge difference, but for more complex subset queries it comes in handy.

UPDATE: Yes, I know I’ve included the require(plyr) line even though the demonstration code doesn’t use it. You’re free to run R without plyr loaded. You’re also free to not put on pants.