Wednesday, January 22, 2014

Cucumber Sidebar Conversations

Last nights Geeky Book Club continued on with the Cucumber book. These chapters didn't elicit earth-shaking insights. However, the sidebar conversations were wonderful.

One of the special pleasures of having such a great group is that you get to find out about all kinds of things you've never heard of. It's almost an unofficial Technology Radar.

The big one I discovered last night was Liquibase. This is almost like a Git repo, but for your database. And these discussions led me on to mention to the group an idea that had been brewing in my mind for a while.

Relational databases are based around a few simple concepts:
  • The concept of a table: rows & columns, almost like a spreadsheet.
  • A few relational operations, with an underlying mathematical model.
  • Set processing of all data.
  • Metadata (the catalog) for discovery/documentation and enforced at creation automatically.
But an enormous complexity arises from these.  Maybe not with the canonical example from Date's book of the Employee and Department tables. But when you get up to ten tables with foreign key constraints, and then move up to hundreds or even thousands of tables (as in Peoplesoft), all with complex relationships -- the complexity is very hard to manage. It's also hard just to grasp.

This is fairly common. The idea in Finance of Options is very simple. Two types: calls & puts. Two sides to each type: long and short. But the complexity that results from this resulted in a Nobel Prize for Fisher Black! And I've read many a neuron-busting treatise about valuing such options, back when I consulted for Wall Street Derivatives traders: the gamma-neutral strategies, delta-neutral hedging, etc. etc. etc.

My idea is that just as we have tools like Cucumber to store our scenarios and link them to our code, so also we need such a tool that shows the database transformations that come from each step of our scenario.

This would be both a documentation tool, a tool for systematically collecting and creating test-data, and educational tool for new developers, or those who want to understand the system. In short, such a tool is most earnestly to be desired.

So all of this crystalized in my mind as I put it forth to my honorable colleagues, last night. And they agreed that such a system would be very useful.

If any of my readers want to help me work on it, TheHackerCIO is now set up on Github, and I think I'll begin hacking ...

Now if only I had a catchy name ...

I Remain,


No comments:

Post a Comment