Those days I'm working on a prototype to supply some weird requisites from our client. For that prototype I'm using Python 2.7, bsddb and cx_Oracle 5.0.3.
Well the work basically consists of converting a C++ batch processor to Python. The original processor have about 4kloc and the Python code will reach about 1,5kloc.
In the first 6 hours of work I've re-coded 50% of the base proprietary API that we use to develop our products (40% of it already was in Python API). In the next 8 hours I coded all the main business rules, and after that I coded some buffering for database queries and debugged the application a little bit.
Using 25 hours of work, the productivity of the Python language allowed me to re-code what took about 200 hours of work in C++.
The next activity is to measure the performance of the execution of the whole process and observe if it will be reasonable to make 20 million of transactions daily. If this goal is achieved, then I'll try to use python on every project for now on.
Even if the prototype goes to a total failure for this project, it will denote that the development costs of smaller solutions can be diminished at least by 50%.