The only pre-processing phase in the execution of a RXS program is a scan through the program to find all user defined variables:
All user-defined variables are made global, that is, all action blocks shares a common definition of the data. A variable is user defined if the variable exists in the RXS coding and if the variable is not a general order or output from a general order.
Output variables word.x and unit.x and variables carrying output from 'sql', 'namespace' and 'prompt' input, plus variables created in imbedded coding, are visible 'downwards' in the RXS program: The variables are visible in the action block where they are created plus all action blocks contained in - or imbedded in - this action block. But do notice that word.x and unit.x are given new content whenever a new action block using default func is entered during execution of the RXS program.
Variables created dynamically using the REXX interpret command are local. Using the RXS instruction make_global 'varname' to make such a variable visible in contained or imbedded action blocks.
RXS queues are always global.
General orders out, outfunc, outfile are local for the action block or text block at which they are stated, and are visible in blocks contained in - or imbedded in - this action block.
Remaining general orders in, func, prompt, imbed, caps etc., are local for the action block.
If a general order is to receive an assignment prior to the execution of the action block, this can be accomplished using a user defined variable. All user defined variables are global.
This example uses variable w_outfunc to transport an assignment into the inner action block:
w_outfunc = word('browse view', random(1, 2))
"What's up doc?"
About half the times this program is executed, the user will end in browse on output, about half the time in view.
The reason for these rather uneven principles is the possible use of RXS for code generation. A COBOL program using RXS code generation normally consists of separate islands of RXS code, separated by sequences of normal COBOL code ('dead code' as seen from the RXS program). This is because you do not generate the whole COBOL coding; but only the parts of the program that is to reflect some specification file. These separate islands of RXS code must be able to communicate, therefore the use of global variables. A schema of clean inheritance of variables will not do.
A variable in RXS is not to be referenced before it is assigned a value. Violating this rule will cause the program to end in error.
This strict rule helps finding typing errors in the program. It also helps finding errors caused by referencing the variable in an action block outside the variable's scope.
if x = 'x' then x = 0
x = x + 1
the RXS program will end in error, giving an error message in line 3: "x has no value".
To see if a variable has a value or not, use the following logic:
if symbol('x') = 'LIT' then do
say 'x does not contain a value'
Notice: the variable name x used in symbol is quoted - elsewhere we would find ourselves back in the tarpit again, with the program making an immediate end saying "x has no value".