Whenever a program error or interrupt occurs, IML automatically issues a pause, which places the module in a paused state. At this time, any statements pushed to the input command queue get executed. Any subsequent RESUME statement (including pushed RESUME statements) resume executing the module from the point where the error or interrupt occurred.

If you have a long application such as reading a large data set and you want to be able to find out where the data processing is just by entering a break-interrupt (sometimes called an attention signal), you push the interrupt text. The pushed text can, in turn, push its own text on each interrupt, followed by a RESUME statement to continue execution.

For example, suppose you have a data set called TESTDATA that has 4096 observations. You want to print the current observation number if an attention signal is given. The following code does this:

   start obsnum;
      use testdata;
      brkcode={"print 'now on observation number',i;"
                "if (i<4096) then do;"
                "call push(brkcode);"
                "resume;"
                "end;"
                };
      call push(brkcode);
      do i=1 to 4096;
         read point i;
      end;
   finish;
   run obsnum;

After the module has been run, enter the interrupt control sequence for your operating system. Type S to suspend execution. The IML procedure prints a message telling which observation is being processed. Because the pushed code is executed at the completion of the module, the message is also printed when OBSNUM ends.

Each time the attention signal is given, OBSNUM executes the code contained in the variable BRKCODE. This code prints the current iteration number and pushes commands for the next interrupt. Note that the PUSH and RESUME commands are inside a DO group, making them conditional and ensuring that they are parsed before the effect of the PUSH command is realized.