disposing of generated rt variables

Message Topic Search Topic Options Post Reply Create New Topic Printable Version Translate Topic

   Hi,
Following the example in (https://www.metacase.com/support/51/manuals/mwb/Mw-5_2_8.html), i generated associative runtime variables -- to indicate which elements i have already visited

variable 'eoid_' oid write 'visited' close

I defined the variable as global since this information is generated in a subreport and is needed upon return to a main report. 

Once the main report, which itself is a subreport, completed its job, i want to dispose of all variables, so, if the main report is called again, a new sleight of eoid_'s are generated. Is there a way to do that

thank you,

Daniel

Author	Message Topic Search Topic Options Post Reply Create New Topic Printable Version Translate Topic
grossd18 Members Profile Send Private Message Find Members Posts Add to Buddy List Major Contributor Joined: 30.Sep.2016 Points: 22	Post Options Post Reply Quote grossd18 Report Post Thanks(0) Quote Reply Topic: disposing of generated rt variables Posted: 12.Oct.2016 at 21:21
	Hi, Following the example in (https://www.metacase.com/support/51/manuals/mwb/Mw-5_2_8.html), i generated associative runtime variables -- to indicate which elements i have already visited variable 'eoid_' oid write 'visited' close I defined the variable as global since this information is generated in a subreport and is needed upon return to a main report. Once the main report, which itself is a subreport, completed its job, i want to dispose of all variables, so, if the main report is called again, a new sleight of eoid_'s are generated. Is there a way to do that thank you, Daniel

grossd18 Members Profile Send Private Message Find Members Posts Add to Buddy List Major Contributor Joined: 30.Sep.2016 Points: 22	Post Options Post Reply Quote grossd18 Report Post Thanks(0) Quote Reply Posted: 12.Oct.2016 at 21:27
	in the meantime i solved the problem by making every generated variable a conceptual entity that is generated (e.g. if i generate a subclass relation then the variable name is the child and parent and an indicator for subclass) -- like this i am building up a variable based representation of the model that can then be associatively queried. i dont know how expensive this is, and i would prefer keeping variable generation small sized, so better if i can only temporarily generate small chunks of the model and then dispose of it, when not needed anymore.

stevek Members Profile Send Private Message Find Members Posts Add to Buddy List MetaCase Joined: 11.Mar.2008 Points: 643	Post Options Post Reply Quote stevek Report Post Thanks(0) Quote Reply Posted: 12.Oct.2016 at 22:43
	There's no operation to dispose of a global variable - but also virtually no cost associated with them. Premature optimization is the root of all kinds of evil :). It's best to aim for a generator that is understandable by someone else without too much work. For code generation, there's virtually never a problem of speed. To collect a 'visited' list, the easiest way is the __Unique() generator. Using it is as simple as: if __Unique() then /* handle the object */ endif By default, it checks for uniqueness of id per type, i.e. the first time you meet an object Foo of type Bar, it will be handled, but if you meet that same object (or another of type Bar also called Foo) again, it will be ignored. Internally it simply concatenates the id's to a newline-separated stream in a variable named after the type. It takes two optional parameters, which default to __Unique(type, id). If you want names to be unique across all types, you can make the first parameter a fixed string, e.g. __Unique('MyList'). If you want to check for uniqueness based on oid rather than string comparison of names, supply that as a second parameter: __Unique(type, oid). The first parameter thus determines the name of the variable where this object is collected, and the second its string representation to add to that variable, and that you want to be unique there. There's not generally a need to duplicate the model in variables. You already have the model there, so you can simply use that. Of course I don't know exactly what you're doing, but in most cases where this is seen, people are trying to turn the model into string variables, and then use MERL as a 3GL to navigate in it. MERL is much better at querying models than string variables: not only do you save a whole phase of converting the model into string variables, you also get to write "do ~Parent.() {:Name}" rather than: "variable 'name' variable 'class' variable 'parent' oid read read read".

grossd18 Members Profile Send Private Message Find Members Posts Add to Buddy List Major Contributor Joined: 30.Sep.2016 Points: 22	Post Options Post Reply Quote grossd18 Report Post Thanks(0) Quote Reply Posted: 12.Oct.2016 at 23:09
	thank you. Since there is more than one way in a model to describe a relation across objects, i need to check which were already visited/generated. These are usually a combination of at least two objects and their relationship between them, so checking for individual object uniqueness is not sufficient. Daniel

stevek Members Profile Send Private Message Find Members Posts Add to Buddy List MetaCase Joined: 11.Mar.2008 Points: 643	Post Options Post Reply Quote stevek Report Post Thanks(0) Quote Reply Posted: 12.Oct.2016 at 23:24
	OK, I understand. I remember the frustration of having to check for attribute, aggregation or association all the time in the Web App Example :). Sounds like you might also be facing the downside of a UML-like approach, where information is spread over several graphs, with no way to say that "this is where this Class is defined" as opposed to "here I'm just referencing this Class". There are a variety of ways of improving that, but they're pretty standard conceptual modeling. You can use __Unique fine with 'compound keys' by building them up like this: __Unique('subclass', $parentClassOid '_' $subClassOid)