Conversation

I was curious about this statistic so I did some digging. The blog post cites ieeexplore.ieee.org/document/62929/ which says something *much* different.

Quote Tweet

Cory House

@housecor

· Dec 8, 2016

Woah. "Each hour spent on code review saves 33 hrs of maintenance." - Glen Russell More via @_ericelliott: buff.ly/2hixryP

Jamison Dance

@jamison_dance

Dec 8, 2016

1. The study is about "code inspection" which is v different from code review. Teams of people literally paraphrasing every single line.

Jamison Dance

@jamison_dance

Dec 8, 2016

See ssltest.cs.umd.edu/class/fall2006 for a detailed overview. Typically 2 hours to review 300 lines with several people at once. Way more involved.

Jamison Dance

@jamison_dance

Dec 8, 2016

2. 1 hr of code inspection didn't save 33 hrs maintenance. It found 1 bug per hour. Their bugs (on telcom switches) took 4.5 days to repair.

Jamison Dance

@jamison_dance

Dec 8, 2016

Paper assumes 7.5 hr workday, thus 4.5 * 7.5 = 33. Fixing a bug includes time to physically drive to the switch, put new software on it.

Jamison Dance

@jamison_dance

Dec 8, 2016

This same thing has happened almost every time I've followed up on original sources from a sensational citation in some blog post.

Jamison Dance

@jamison_dance

Dec 8, 2016

I believe code reviews are helpful. I also believe we love to add legitimacy to our beliefs by mis-citing things that we think confirm them.

Jamison Dance

@jamison_dance

Dec 8, 2016

But also sometimes you just gotta get those sweet sweet clicks and no one clicks b/c of context and nuance so ¯\_(ツ)_/¯

Jamison Dance

@jamison_dance

Dec 8, 2016

Final point: the OP was about tradeoffs in static type systems which is a fascinating topic. Pendulum is swinging back to static.

Jamison Dance

@jamison_dance

Dec 8, 2016

A common argument for static typing is reducing error rates. Evidence that static type systems reduce error rates is basically inconclusive.

Replying to

and

I've run up against this a few times, and it frankly baffles me based on my own experience. Suggestive, though, as you say!

5:15 PM · Dec 8, 2016Twitter for iPhone

Replying to

Yeah it does feel very weird to me as well, but I'd rather acknowledge it than just cherry-pick studies I like.