Revision as of 18:32, 19 October 2011

Analizing OpenStack code

Tools

• cvsanaly
• MySQL
• GNU R
• Pentaho Report

List of repositories tracked

• compute-api
• identity-api
• keystone
• nova
• openstack-dashboard
• quantum
• swift
• glance
• image-api
• netconn-api
• object-api
• openstack-manuals

How to do it

Installing cvsanaly2 from https://projects.libresoft.es/projects/cvsanaly/wiki Clone all the repositories listed on https://github.com/openstack/:

$ git clone git://.... [glance|swift|nova...]

Then create the databases for CVSanaly2 to store the logs in

$ mysqladmin -u mlstats -p create openstackrepos

And run cvsanaly2 on each of the repositories:

~/projects/swift$ cvsanaly2 --extensions Metrics,FileTypes,CommitsLOC -u mlstats -p xxxxxxx -d openstackrepos
Parsing log for swift/
Executing extensions
Executing extension FileTypes
Executing extension Metrics

Relevant queries

Exclude from results system authors Jenkins and Gerrit. Database schema

Total number of commits per repository in past 30 days

• View: who commited to any repo, how many times, in the 30 days

SELECT
     Count(scmlog.date),
     `people`.`name`,
     `repositories`.`name` AS RepoName,
     `people`.`email` AS email
FROM
     `scmlog` INNER JOIN `people` ON `scmlog`.`author_id` = `people`.`id`
     INNER JOIN `repositories` ON `scmlog`.`repository_id` = `repositories`.`id`
WHERE
     date_sub(curdate(),interval 30 day) <= `scmlog`.`date`
AND
     (`people`.`name` NOT LIKE 'jenkins%' AND `people`.`name` NOT LIKE 'gerrit%')
GROUP BY
     people.name
ORDER BY
     Count(scmlog.date) DESC

See the demo report

Total number of commits across all repos aggregated per month

Total number of commits per repository aggregated per month

todo

Total number of commits per author per repository

todo

Total number of commits per author in total

todo

Total number of commits per author per repository in past 30 days

todo

Average number of Lines of Code changed per commit per repository

todo

Average number of Lines of Code changed per commit per repository per author

todo

Using GNU R to produce graphs

After starting R, connect to MySQL database:

library(RMySQL)
con <- dbConnect(dbDriver("MySQL"), user="mlstats", password="xxxxx", dbname="openstackrepos")

Number of commits per month

query <- "SELECT date_format(s.date, '%m/%Y') date, count(s.id) commits
          FROM scmlog s group by date_format(s.date,'%Y%m');"
results <- dbGetQuery(con,query)
evol_commits <- ts(results$commits, start=c(2010,7), freq=12)
  png("nova_commits_month.png")
plot(evol_commits, type="l", xlab="Date", ylab="Commits",
     main="Number of commits per month - NOVA")
query_avg <- "SELECT AVG(g.numcommits)
    FROM
    ( SELECT date_format(s.date, '%Y') myyear,
        date_format(s.date, '%m') mymonth, count(s.id) numcommits
        FROM scmlog s
        GROUP BY date_format(s.date,'%Y%m') ) g;"
result_avg <- dbGetQuery(con,query_avg)
qqline(result_avg, col="blue", lty=2)
query_max_min <- "SELECT MAX(g.numcommits) as max, MIN(g.numcommits) as min
    FROM
    ( SELECT date_format(s.date, '%Y') myyear,
        date_format(s.date, '%m') mymonth,
        count(s.id) numcommits
        FROM scmlog s
        GROUP BY date_format(s.date,'%Y%m') ) g;"
result_max_min <- dbGetQuery(con,query_max_min)
qqline(result_max_min$max, col="red", lty=2)
qqline(result_max_min$min, col="green", lty=2)
legend("topright", inset=.05, c("average","maximun","minimun"),
       fill=c("blue","red","green"))
dev.off()

Number of Commits per author

query <- "SELECT p.name author, count(s.id) commits
    FROM scmlog s LEFT JOIN people p ON s.committer_id=p.id
    GROUP BY committer_id ORDER BY commits;"
results <- dbGetQuery(con,query)
plot(results$commits, xlab="Author", ylab="Commits",
     main="Number of commits by author")

Number of commits by author per year

query <- "
SELECT year, name, num
FROM
(SELECT date_format(s.date, '%Y') AS year, p.name AS name, count(s.id) AS num
FROM scmlog s LEFT JOIN people p ON s.committer_id=p.id
GROUP BY year, name
ORDER BY year, num) g
WHERE g.num > 10;
"
results <- dbGetQuery(con,query)
results$year <- factor(results$year)
query_names <- "
SELECT DISTINCT(name)
FROM
(SELECT date_format(s.date, '%Y') AS year, p.name AS name, count(s.id) AS num
FROM scmlog s LEFT JOIN people p ON s.committer_id=p.id
GROUP BY year, name
ORDER BY year, num) g
WHERE g.num > 10;
"
names <- dbGetQuery(con,query_names)
col <- 1
for (i in names$name) {
    results$color[results$name==i] <- col
    col <- col + 1
}
dotchart(results$num, groups=results$year, labels=results$name,
         color=results$color, cex=.7, xlab="Number of commits",
         main="Commits by author per year")

Aggregated number of commits up to now

query <- "SELECT g.myyear, g.mymonth, g.numcommits,
        (@sumacu:=@sumacu+g.numcommits) aggregated_numcommits
    FROM
        (SELECT @sumacu:=0) r, (SELECT date_format(s.date, '%Y') myyear,
            date_format(s.date, '%m') mymonth,
            COUNT(s.id) numcommits
        FROM scmlog s
        GROUP BY date_format(s.date,'%Y%m')) g;"
results <- dbGetQuery(con,query)
evol_num_commits <- ts(results$aggregated_numcommits, start=c(2010,10), freq=12)
plot(evol_num_commits, type="h", xlab="Date", ylab="Commits",
     main="Aggregated number of commits", col = "dark blue")

In order to save an image of the plot use

  png("<filename>.png")
  plot <command>
  dev.off()

Using Pentaho Report to produce graphs

TODO