Helping users improve build performance

In general, it would be nice if the service would help users improve build performance by suggesting workflow improvements. This hit me when checking PyPi statistics: there are millions of downloads a month of common packages such as requests or numpy.

Could travis identify frequent pip install traffic from a build script and suggest to users that they use available caching mechanisms to improve build performance?

Or could travis notice that a build starts from one configuration and then proceeds to install packages that exist in another, similar configuation, and suggest to the user they use it instead?

Since caching is not done on the build server hosts, downloading from the cache instead of PyPi will not make a difference in time or traffic.
You’ll just move the load to the Travis cache hosts.
Besides, caching of common traffic is the domain of the network operators, in the grand scheme your unique cache is actually worse.

Just use the cache for things that can not be downloaded.

1 Like
Imprint