3 ms·

At the risk of exposing my ignorance, I thought >Other “common wisdom”, like using locals instead of globals, yields relatively little gain. this advice was t

At the risk of exposing my ignorance, I thought >Other “common wisdom”, like using locals instead of globals, yields relatively little gain. this advice was typically more related to avoiding collisions with variable names, rather than performance?

thristian 15y ago
In Python, it's both - the basic structured-programming advice to avoid global variables is always good, but it's a specific quirk of Python that makes global variables (including built-in functions) slower than local variables. Python has full dynamic scoping, which means that inside a function you can refer to any variables set in outer scopes. Because Python is a dynamic language, every time you refer to a variable, the Python interpreter looks for it in the local scope first, and then each enclosing scope until it hits the containing module. A local variable will always be found in the first iteration of that loop, a global variable will take at least two iterations.
- adgar 15y ago
  > Python has full dynamic scoping, which means that inside a function you can refer to any variables set in outer scopes. You are describing lexical scope, not dynamic scope.
- encukou 15y ago
  Another CPython quirk is that global (module-level) variables are looked up in a dict, but a function's local variables normally get a reserved chunk of memory that's directly indexable.
lloeki 15y ago
A "global" variable is one that is not residing in the local scope. Global here is really non-local, which is a term that changed during the py3k transition. As you see, str and ord are "system" functions, but really are variables containing function objects. The same goes for WHITELIST, which is a variable (conventionally named as a constant) probably defined at the module level. All of those are not local to the function, hence more taxing to call in a loop. Besides, the general advice is not to avoid collisions (you use both namespaces and locality to resolve that) as the non local stuff here really has a reason to be global, but non local stuff has to survive concurrency. It is the case here as the things called really are constants.

thristian 15y ago

In Python, it's both - the basic structured-programming advice to avoid global variables is always good, but it's a specific quirk of Python that makes global variables (including built-in functions) slower than local variables. Python has full dynamic scoping, which means that inside a function you can refer to any variables set in outer scopes. Because Python is a dynamic language, every time you refer to a variable, the Python interpreter looks for it in the local scope first, and then each enclosing scope until it hits the containing module. A local variable will always be found in the first iteration of that loop, a global variable will take at least two iterations.

adgar 15y ago
> Python has full dynamic scoping, which means that inside a function you can refer to any variables set in outer scopes. You are describing lexical scope, not dynamic scope.
encukou 15y ago
Another CPython quirk is that global (module-level) variables are looked up in a dict, but a function's local variables normally get a reserved chunk of memory that's directly indexable.

lloeki 15y ago

A "global" variable is one that is not residing in the local scope. Global here is really non-local, which is a term that changed during the py3k transition. As you see, str and ord are "system" functions, but really are variables containing function objects. The same goes for WHITELIST, which is a variable (conventionally named as a constant) probably defined at the module level. All of those are not local to the function, hence more taxing to call in a loop. Besides, the general advice is not to avoid collisions (you use both namespaces and locality to resolve that) as the non local stuff here really has a reason to be global, but non local stuff has to survive concurrency. It is the case here as the things called really are constants.