Programming Python, Second Edition

PyErrata Database Interfaces

Now that we’ve seen the user interfaces and top-level implementations of browse and submit operations, this section proceeds down one level of abstraction to the third and last major functional area in the PyErrata system.

Compared to other systems in this part of the book, one of the most unique technical features of PyErrata is that it must manage persistent data. Information posted by readers needs to be logged in a database for later review. PyErrata stores reports as dictionaries, and includes logic to support two database storage mediums -- flat pickle files and shelves -- as well as tools for synchronizing data access.

The Specter of Concurrent Updates

There is a variety of ways for Python scripts to store data persistently: files, object pickling, object shelves, real databases, and so on. In fact, Chapter 16 is devoted exclusively to this topic and provides more in-depth coverage than we require here.^[116] Those storage mediums all work in the context of server-side CGI scripts too, but the CGI environment almost automatically introduces a new challenge: concurrent updates. Because the CGI model is inherently parallel, scripts must take care to ensure that database writes and reads are properly synchronized to avoid data corruption and incomplete records.

Here’s why. With PyErrata, a given reader may visit the site and post a report or view prior posts. But in the context of a web application, there is no way to know how many readers may ...