50000 Revisions Committed to R

Oct 102009

Featured, R Graphics, R Language Add comments

Today Romain Francois posted an interesting topic in the R-help list, and you can read his blog post for more details: celebrating R commit #50000. 50000 is certainly not a small number; we do owe R core members a big “thank you” for their great efforts in this fantastic statistical language in the 13 years. When I saw Romain’s data, I suddenly remembered a question I asked to one of Prof Ripley’s student a couple of years ago: does Prof Ripley ever sleep? And he answered “No!”. No wonder we can see Prof Ripley so frequently in the R-help/devel mailing list. If you have stayed on R-help list for enough long time, you’ll surely know several facts, e.g. Martin Maechler will arrive in less than 3 minutes if you dare call an R package “library”, and you will get “Ripleyed” if you are not careful enough in posting your R code.

> library(fortunes)
> fortune("Ripleyed")

And the fear of getting Ripleyed on the mailing list also makes me think, read,
and improve before submitting half baked questions to the list.
 -- Eric Kort
 R-help (January 2006)

While these facts are revealing their great efforts in helping R users, we can see their work hours in committing revisions to R. For example, the answer to my question is clear in the graph below:

Prof Ripley Never Sleeps

## R code borrowed from Romain Francios
process_chunk <- function(txt) {
    if (length(txt) == 1L)
        return(NULL)
    header_line <- strsplit(txt[2L], " | ", fixed = TRUE)[[1]][c(1L,
        2L, 3L)]
    revision <- substring(header_line[1], 2)
    author <- header_line[2]
    if (author %in% c("apache", "root"))
        return(NULL)
    date <- substring(header_line[3], 1, 25)
    nlines <- length(date)
    matrix(c(rep.int(revision, nlines), rep.int(author, nlines),
        rep.int(date, nlines)), nrow = nlines)
}
data <- local({
    lines <- readLines("rsvn.log")
    index <- cumsum(grepl("^-+$", lines))
    commits <- split(lines, index)
    do.call(rbind, lapply(commits, process_chunk))
})
colnames(data) <- c("revision", "author", "date")
simple <- data[!duplicated(data[, "revision"]), ]
hour.data = data.frame(author = simple[, "author"],
      hour = as.integer(substr(simple[, "date"], 12, 13)), year = as.integer(substr(simple[,
          "date"], 1, 4)))
hour.data = subset(hour.data, year >= 1997 & (author %in%
    c("hornik", "maechler", "pd", "ripley")))
library(ggplot2)
# png("ripley-work-hour.png")
qplot(hour, data = subset(hour.data, author == "ripley"),
    main = "Does Prof Ripley Ever Sleep?") + stat_bin(binwidth = 1)
# dev.off()

Here I only selected four authors who have largest number of commits during 1997~2009. We can see the changes of working hours along these years:

Working hours of four R core members

hour.max = max(with(hour.data, table(author, year,
    hour)))
library(animation)
# you need ImageMagick to create the GIF animation!
saveMovie({
for (i in sort(unique(hour.data$year))) {
    print(qplot(hour, data = subset(hour.data, year == i), xlim = c(0,
        23), ylim = c(0, hour.max), main = i) + facet_wrap(~author) +
        stat_bin(binwidth = 1))
}
}, interval = 1.5, moviename = "r-core-work-hour", outdir = getwd())

The patterns are clear: Kurt does not like burning night oil; Martin tends to work very early in the morning (esp during 2000~2004); Peter always work at mid-night (highly centered around 12pm); and for Prof Ripley, he works round the clock but most in the morning (probably that’s when he begins to “Ripley” users? after that time, less people dare to report bugs so his work decays exponentially?)

6 Responses to “50000 Revisions Committed to R”

Comments (5) Pingbacks (1)

fan says:

2009/10/10 at 11:06 pm

Wow, amazing~

Reply
Kevin Wright says:

2009/10/26 at 4:12 pm

I don’t use ggplot2, so maybe I’m wrong, but it does not appear that ggplot2 is respecting the bin width of 1 hour. Here is an equal-bin-width version using lattice:
```
library(lattice)
histogram(~hour, data = subset(hour.data, author == "ripley"),
          type="count",
          breaks=0:24-.1,
          scales=list(x=list(at=c(0,6,12,18,24)))
          )
```
Your graph does a nice job of answering the question. Ripley usually quits working by 12:00 and almost always by 1:00 AM. He almost never starts before 5:00 AM.

Shifting the ‘window’ from 0-24 hours to 4-28 hours provides a different (more natural?) perspective on Ripley’s ‘day’:
```
hour2 <- subset(hour.data, author == "ripley")
hour2$hour <- ifelse(hour2$hour < 4, hour2$hour + 24, hour2$hour)
histogram(~hour, data = hour2,
          type="count",
          breaks=4:30-.1,
          scales=list(x=list(at=c(0,6,12,18,24)))
          )
```
Kevin Wright

Reply
- Yihui Xie says:
  
  2009/10/27 at 5:26 pm
  
  Thanks for your reply, Kevin. Yes, you second suggestion is a more natural representation of the working hours — a day starts from the morning instead of the mid-night (for most people). For the histogram, actually I’m not a regular user of ggplot2, and I didn’t check stat_bin() carefully, but after your reminding, I found a weird behavior of it which confused me: qplot(1:10)+stat_bin(binwidth=1) will produce a histogram which is not flat…
  
  Reply
Thomas Levine says:

2010/06/16 at 3:20 pm

I think Ripley’s high commit count is misleading when you’re comparing whether he sleeps to whether other R core members sleep. You should change the y axes from raw count to percentage of total count. Once you do this, I think the other authors’ commits will look more spread-out.

Reply
- Yihui Xie says:
  
  2010/07/02 at 2:07 am
  
  Thanks! That will be a reasonable enhancement
  
  Reply

Statistics, R, Graphics and Fun