New behavior for QUIT

This document describes a new behavior for making QUIT work better with loaded scripts.

This is just a summary of the method. See Defining a better QUIT for a full discussion of the theory.

Table of Contents Summary Example QUIT/return QUIT/now QUIT in Console How it Works

Summary

The behavior of QUIT has changed to depend on how it is used. QUIT can either quit the entire Rebol system, or just exit the current script being evaluated.

Example

It's easiest to show an example.

Take the Rebol script quit.r below:

Rebol []
print "hello"
quit
print "end"

If you start Rebol with that script (on the command line or from an icon), the script will print hello, exit, and Rebol will terminate.

However, if you start Rebol and type from the console, the script will return to the console when the QUIT is evaluated:

>> do %quit.r
hello
>>

Notice that "end" is not printed.

QUIT/return

If you want the script above to return a value:

Rebol []
print "hello"
quit/return "done"

Now you can see that result:

>> do %quit.r
hello
== "done"

QUIT/now

There may be a few places within the code where QUIT must terminate the Rebol system immediately. For example, a looping error situation may want to force such a QUIT after the second recursion.

This is done with:

QUIT/now

The Rebol system will terminate immediately.

QUIT in Console

If you use QUIT within the console, it will quit normally, just as before.

How it Works

If you look at the source for system/intrinsic/do, you will see that the script is evaluated using CATCH with a /quit refinement. This is a special variation of CATCH that is capable of catching the QUIT when it is thrown. (The QUIT throw is identical to the method used by HALT.)