8.5.3 Regina Notes for the Variable Pool Due to the subtleties described at the end of the previous subsection, some notes on how Regina handles RXSHV_NEXTV requests for compound variables are in order. The following rules applies: · Both the stem variable FOO. and the compound variable having FOO. as stem and a nullstring as tail, are returned with the name of FOO.. In this situation, a sequence of RXSHV_NEXTV requests may seem to return values for the same variable twice. This is unfortunate, but it seems to be the only way. In any case, you'll have to perform the RXSHV_SYFET in order to determine which is which. · If a stem variable has not been assigned a value, its compound variables are only returned if they have been assigned an explicit value. i.e. compound variables for that stem that have either never been assigned a value, or have been dropped, will not be reported by RXSHV_NEXTV. There is nothing strange about this. · If a stem variable has been assigned a value, then its compound variables will be reported in two cases: Firstly, the compound variables having explicitly been assigned a value afterwards. Secondly, the compound variables which have been dropped afterwards, which are reported to have their initial value, and the flag RXSHV_NEWV is set in shvret. It may sound a bit stupid that unset variables are listed when the request is to list all variables which have been set, but that is about the best I can do, if I am to stay within the standard definition and return a complete and exact status of the variable pool. If the return code from RexxVariablePool() is less than 128, Regina is guaranteed to have tried to process all requests in the chain. If the return code is above 127, some requests may not have been processed. Actually, the number 127 (or 128) is a bit inconvenient, since it will be an problem for later expansion of the standard. A much better approach would be to have a preprocessor symbol (say, RXSHV_FATAL, and if the return code from the RexxVariablePool() function was larger than that, it would be a direct error code, and not a composite error code built from the shvret fields of the requests. The RXSHV_FATAL would then have to be the addition of all the atomic composite error codes. (Warning: author mounting the soapbox.) The right way to fix this, is to let the function RexxVariablePool() set another flag in shvret (e.g. named RXSHV_STEM) during RXSHV_NEXTV if and only if the value returned is a stem variable. That way, the application programmer would be able to differ between stem variables and compound variable with a null string tail. To handle the other problem with compound variables and RXSHV_NEXTV, I would have liked to return a null string in shvvalue if and only if the variable is a compound variable having its initial value, and the stem of that compound variable has been assigned a value. Then, the value of the compound variable is equal to its name, and is available in the shvname field. I'd also like to see that the shvret value contained other information concerning the variables, e.g. whether the variable was exposed at the current procedure level. Of course, Regina does not contain any of these extra, non- standard features. (Author is dismounting the soapbox.) When Regina is returning variables with RXSHV_NEXTV, the variables are returned in the order in which they occur in the open hashtable in the interpreter. i.e. the order in which variables belonging to different bins are returned is consistent, but the order in which variables hashed to the same bin are returned, is non-deterministic. Note that all compound variables belonging to the same stem are returned in one sequence.