Skip to content

Python performance lessons

I ran across this article on profiling python, etc. I took away three things from it:

  1. Numpy's take method could speed up 'fancy' indexing.

  2. Pandas has very fast/efficient join and group by by functionality.

  3. Use named tuples more often when needing 'simple' objects that are not going to change. They are essentially c-structs and thus are contiguous in memory.