[Peter Lewerin] 2014-02-19:

The term ''stack frame'' is based on the low-level implementation of the procedure of allocating memory for a `[proc]` to store its ''data objects'' (arguments, local variables, etc) in. Briefly, the ''execution stack'' is a linear structure holding data objects which is managed by setting a pointer to the latest object added to the stack: objects are discarded by moving that pointer back to the previous object, and so on. When a `proc` is called, each data object that it will use is added to the stack, together with the old value of the stack pointer, and the `proc` is then allowed to access each data object in the segment, or ''frame'', between the places pointed to by the current value of the stack pointer (which might change: new data objects can be added to the stack after the `proc` is called, moving the stack pointer further) and the old value of the stack pointer. When the `proc` is done and returns, the stack pointer is reset to the old value, automatically discarding each object in the `proc`'s stack frame. All of this is managed by the Tcl implementation and none of it requires the attention of the Tcl programmer.

There are a number of commands designed to cheat with this arrangement, such as `[global]`, which adds a global variable to the set of data objects the `proc` is allowed to access; `[uplevel]`, which allows a script to be evaluated in another stack frame; and `[upvar]`, which links a variable inside the `proc`'s stack frame to a variable in another stack frame. (More properly, `uplevel` and `upvar` access other stack ''levels'', which is a similar but not entirely equivalent thing, though somewhat conflated in Tcl's documentation.) The commands `[info frame]` and `[info level]` provide some information on the state of the stack.

<<categories>>Concept