examples of interesting comparisons to show

[tdhock]

Here is the list of interesting comparisons that we can hopefully plot one day using your package (same as I sent you via email).

For example you could try to compare this early commit https://github.com/tdhock/PeakSegJoint/blob/8acd6e24fe6ff321b23b4d7fce7373b0fa1c120d/tests/testthat/test-likelihood.R with the current master.

Or maybe try the plotly package. They use the PR development model, so you should be able to try your code to compare master against any of the PR branches: https://github.com/ropensci/plotly/pulls

There is also a mystery test that I found to be very slow in the current master https://github.com/ropensci/plotly/blob/master/tests/testthat/test-cookbook-axes.R#L96 It would be interesting to take this test from the current master, and run it on past versions of the R code, until we find when the test started getting slow.

[analyticalmonk]

Regarding the PeakSegJoint commit, turns out the earlier commit was a bit faster than the current master.

I wasn't able to test the plotly tests though. I get the "couldn't connect to host" error every time I try to run them.

[tdhock]

Thanks for the analysis Akash. A couple of suggestions:

(1) if you only have 1 time measurement per commit, it is impossible to tell if there is any real/significant difference between these two commits. In the future please run at least 3 separate timings on each commit.

(2) the date_time axis is not readable. Could you rotate the axis labels? and instead of labeling a grid of time points I think it would be better to label the times of each of the commits.

## from ?scale_x_datetime
     start <- ISOdate(2001, 1, 1, tz = "")
     df <- data.frame(
       day30  = start + round(runif(100, max = 30 * 86400)),
       day7  = start + round(runif(100, max = 7 * 86400)),
       day   = start + round(runif(100, max = 86400)),
       hour10 = start + round(runif(100, max = 10 * 3600)),
       hour5 = start + round(runif(100, max = 5 * 3600)),
       hour  = start + round(runif(100, max = 3600)),
       min10 = start + round(runif(100, max = 10 * 60)),
       min5  = start + round(runif(100, max = 5 * 60)),
       min   = start + round(runif(100, max = 60)),
       sec10 = start + round(runif(100, max = 10)),
       y = runif(100)
     )
qplot(day30, y, data = df)+
  scale_x_datetime(breaks=ISOdate(2001, 1, 1:31))+
  theme(axis.text.x=element_text(angle=-90))

[tdhock]

Not sure about the "couldn't connect to host" error. If you want me to help you debug that please paste the R output that you saw.

[tdhock]

Also when you compare a branch with master, it would probably be better to plot each commit as a factor, since maybe the master commit occured later than the branch commit (but you still want to plot it on the left).

commit.list <- head(commits(repo))
commit.when <- sapply(commit.list, when)
commit.name <- c("master", commit.when) # name for plotting, you can change this!
commit.fac <- factor(commit.name, levels=commit.name)
qplot(commit.fac, commit.seconds)