26 June 2012
I showed how perceptual diffs are core to our team's continuous deployment during my Ignite talk at Velocity.
The gist:
Example 1: Fixing a typo. I can't see the difference. Computers can.
Example 2: Histogram axis regression. One of the first where we caught a bug before pushing it to users.
Perceptual diffs have closed the loop on continuous deployment, giving us the confidence to push new code more often.
Hopefully they put the Ignite video up somewhere~
The gist:
- On every commit we build and push to staging, which takes ~8 minutes
- We screenshot each staging version and compare it visually to the live version on known URLs
- When staging and live versions differ, we get an email
Example 1: Fixing a typo. I can't see the difference. Computers can.
Example 2: Histogram axis regression. One of the first where we caught a bug before pushing it to users.
Perceptual diffs have closed the loop on continuous deployment, giving us the confidence to push new code more often.
Hopefully they put the Ignite video up somewhere~
1 comments:
very cool.
it seems like this is similar to unit and integration tests, applied to parts of a product where they aren't (traditionally) appropriate. does that mean this isn't so meaningful for e.g. non user visible infrastructure?
also, i'm curious how you generate and maintain the set of test requests and data. seems like that could take some work on an ongoing basis.
looking forward to the video!
Post a Comment