Lifetime / Persistence of called programs [RTC 36149]

When migrating programs from legacy languages like Cobol or RPG to EGL we often encounter the problem that in all these languages the programmer has control about how long a program stays in memory.

- In Cobol the developer just can CANCEL another program in memory.
- In RPG the program itself determines according to the status of *INLR if it should stay in memory or not after leaving the program.

In EGL programs and all their variable content is lost after leaving a called program. There we only have the possibility to create a library which is a whole other thing compared to a called program.

This major architectural difference between legacy code and EGL makes it very difficult to do highly automated migrations.

So what we would need is an enhancement that is similar to the COBOL CANCEL instruction.

- PGM A (main) calls B
- B returns but stays in memory
- A calls B again
- B now works with all the values from the first call but the parametes
- B returns but stays in memory
- A says ... ok, don't need B anymore, I'll CANCEL it. B is now removed from memory.

Especially in Java generation this would also have a major impact on performance because we reuse all the objects already available from the first call.

This feature is needed for Java generation as well as for Cobol generation. From a performance perspective the implementation for Java Generation is more critical for us. From a functional perspective this enhancement is critical for Java as well as for Cobol generation.