version 2

[esler]

Current version of library is suboptimal and it has a lot of smaller issues. Idea is to take gain experience and write completely new version. Main features.

• cache close as possible to the storage
  • ideally storage itself is the cache (DB dump)
• each tests has its own DB
  • no need for (problematic) cleanup
  • allow parallel test execution
• minimal testing set
  • only tables from fixtures (some kind of master DB with complete schema would be necessary)
• fill DB at once
  • prepare SQL in advance (from cache) and import it by one exec
• lazy loading of fixture connections
• better merging strategy
  • replace missing values with DEFAULT