shelve module and
ZODB package of the prior sections are powerful tools. Both allow
scripts to throw nearly arbitrary Python objects on a keyed-access
file and load them back later—in a single step for shelves and with a
small amount of administrative code for ZODB. Especially for
applications that record highly structured data, object databases can
be convenient and efficient—there is no need to split and later join
together the parts of large objects, and stored data is processed with
normal Python syntax because it is normal Python objects.
Shelves and ZODB aren't relational database systems, though; objects (records) are accessed with a single key, and there is no notion of SQL queries. Shelves, for instance, are essentially databases with a single index and no other query-processing support. Although it's possible to build a multiple-index interface to store data with multiple shelves, it's not a trivial task and requires manually coded extensions.
ZODB supports some types of searching beyond shelve (e.g., its cataloging feature), and persistent objects may be traversed with all the power of the Python language. However, neither shelves nor ZODB object-oriented databases provide the full generality of SQL queries. Moreover, especially for data that has a naturally tabular structure, relational databases may sometimes be a better fit.
For programs that can benefit from the power of SQL, Python also supports relational database systems. Relational databases ...