From: Wolfgang Mauerer <wolfgang.mauerer@xxxxxxxxxxx>
Initiate-response diagrams are informative for a
given release cycle. Means of summarising the information
over the lifetime of a project are missing.
Try to fill this gap by providing a single-scalar summary
value for each release range. The methods contains several
parameters that will need to be adjusted to sensible
values.
Signed-off-by: Wolfgang Mauerer <wolfgang.mauerer@xxxxxxxxxxxxxxxxx>
---
codeface/R/ml/init.response.r | 39 +++++++++++++++++++++++++++++++++++++++
1 file changed, 39 insertions(+)
diff --git a/codeface/R/ml/init.response.r b/codeface/R/ml/init.response.r
index 742cdf7..eb1702a 100644
--- a/codeface/R/ml/init.response.r
+++ b/codeface/R/ml/init.response.r
@@ -52,3 +52,42 @@ plot.init.response <- function(ir, title=NULL) {
return(list(g1, g2))
}
+
+summarise.init.response <- function(con, ml.id, pid) {
+ range.ids.list <- query.range.ids.con(con, pid)
+ cycles <- get.cycles.con(con, pid)
+
+ ## Comute a very heuristically motivated summary of
+ ## the mailing list structure for each release range.
+ ## For each range, compute the fraction of
+ ## responses received/messages initiated. A high ratio
+ ## indicates that topics of interest have been raised.
+ ## Compute the 0.75 quantile of the distribution to
+ ## obtain the most influential contributors
+ ## Then, compute the median degree of these contributors
+
+ dat <- lapply(range.ids.list, function(range.id) {
+ dat <- query.initiate.response(con, ml.id, range.id)
+
+ dat$frac <- dat$responses.received/dat$initiations
+ qnt <- quantile(dat$responses.received/dat$initiations,
+ na.rm=TRUE, probs=0.75)
+ median(dat$deg[dat$frac > qnt], na.rm=T)
+
+ return(data.frame(cycle=cycles$cycle[cycles$range.id==range.id],
+ med.deg=median(dat$deg[dat$frac > qnt], na.rm=T),
+ num.msg=nrow(dat)))
+
+ })
+
+ return (do.call(rbind, dat))
+}
+
+## Visualise a given initiate-response data structure
+plot.ir.summary <- function(dat.ir.summary) {
+ g <- ggplot(dat.ir.summary, aes(x=cycle, y=med.deg)) +
+ geom_point(aes(size=num.msg)) +
+ xlab("Release cycle") + ylab("Med. important contributor degree") +
+ scale_size_continuous("Number of\nmessages")
+ return(g)
+}
--
2.1.4
