10 Python Performance Tips
DRAFT#
Readability counts
Use discretion. Oftentimes code being obvious to the reader is a greater gain than nanoseconds of performance. Not always.
IO bound performance bottlenecks are usually several orders of magnitude greater than CPU bound bottlenecks.
- Use tuples over lists where the data does not change after creation. Tuples store less data in memory than lists and use resource caching - so can be instantiated quicker. Where the number of elements is 1 to 20 - the garbage collector does not give the data back to the system. So the addresses can be reused without a system call.
- Allocate lists with a list comprehension over use of append. Using
append
overallocates memory. - Use sets when unique data is needed