zodb / zodbshootout / 105

86%

Pull #32

travis-ci

Don't include calls to `random` in the add tests.

Instead pre-cache all the random data we'll need.

Turns out that this was a major performance drag for larger random
sizes, as shown by the blob tests. It was in the top part of the
profile, now its not.

This also ensures that each test-rep within a run uses consistent
random data so compression/hashing/whatever effects should be more
consistent.

108 of 108 new or added lines in 4 files covered. (100.0%)

636 of 701 relevant lines covered (90.73%)

3.51 hits per line