Constant folding, the peephole optimizer, and the AST optimizer

CPython uses various techniques to optimize your code. The first optimization takes place as soon as source code is transformed into the form of the abstract syntax tree, just before it is compiled into byte code. CPython can recognize specific patterns in the abstract syntax tree and make direct modifications to it. The other kind of optimizations are handled by the peephole optimizer. It implements a number of common optimizations directly on Python's byte code. As we mentioned earlier, constant folding is one such feature. It allows the interpreter to convert complex literal expressions (such as "one" + " " + "thing", " " * 79, or 60 * 1000) into a single literal that does not require any additional operations (concatenation or multiplication) at runtime.

Until Python 3.5, all constant folding was done in CPython only by the peephole optimizer. For strings, the resulting constants were limited in length by a hardcoded value. In Python 3.5, this value was equal to 20. In Python 3.7, most of the constant folding optimizations are handled earlier on the abstract syntax tree level. These particular details are a curiosity rather than a thing that can be relied on in your day-to-day programming. Information about other interesting optimizations performed by AST and peephole optimizers can be found in the Python/ast_opt.c and Python/peephole.c files of Python's source code.

Let's take a look at string formatting with f-strings.