Talk:Jump anywhere

Draft task

My new task description might not match my intent. I have a short Ruby example, but I want to add more Ruby code, and some C code. --Kernigh 02:41, 19 May 2011 (UTC)

OK, I'll monitor and ask about changes needed for the Python as things develop :-)
--Paddy3118 02:54, 19 May 2011 (UTC)

Tcl (8.6) has pretty much the same capabilities as Python in this area. (When people ask for general jump anywhere capabilities, we tend to be somewhat brusque in our dismissals; it's totally against nice structured programming as well as being evil.) I'll keep an eye on the evolution of the page to see whether I want to omit or implement. Generally, should continuations be a solution of this task? Or is it more about setjmp/longjmp? –Donal Fellows 17:51, 19 May 2011 (UTC)

Just to be clear, this is not to diss having a task on continuations. It's just they're not simple jumps; if we split, we can easily justify another task specifically for them. –Donal Fellows 19:09, 19 May 2011 (UTC)

Demonstrate a local jump and a global jump

It might be better for the task to be demonstrate a local jump and a global jump. (Currently the requirement appears to be for a global jump, requiring a new task to be created for a local jump).

Markhobley 14:19, 19 May 2011 (UTC)

I propose that unwinding of the call stack is moved to a separate task, and rename this task to "Demonstrate local and global jumps". Keeping the tasks separated, enables the solutions to remain simple, and separates aspects. --Markhobley 16:40, 6 June 2011 (UTC)

What is a "global jump"? Continuation? Longjump (which involves stack)? --Ledrug 03:04, 11 June 2011 (UTC)

I assumed that a local jump is to a label within the same procedure (or block of procedures), whereas a global jump is to anwhere within the program. --Markhobley 08:12, 11 June 2011 (UTC)

That would normally involves unwinding the stack, then, because if you are going out of a proc or function and not coming back, you have to clean up what's already allocated for your current scope, or everyone gets confused -- unless your language doesn't have a stack concept, which is rare. --Ledrug 08:39, 11 June 2011 (UTC)

Yeah. It depends on the language. A local jump within the same procedure would probably not need an unwind. A global jumb could need an unwind, and the unwind may be done at source, prior to the jump or at destination following the jump. Jumping out of a procedure (following a call) is not usually a good idea, although it may be for a critical reason, such as an abort. I would just stick a note saying "the stack will need to be unwound, before this jump is performed" and cross reference to the stack unwinding task. Some languages may have a garbage collector, which removes the need to unwind the stack. Maybe it would be tidier to split the task into Jumps/Local, Jumps/Global and Jumps/With stack unwinding. --Markhobley 10:20, 11 June 2011 (UTC)

Signal use in Rexx

__Warning__: Not only jumping into Do loops (as noted under REXX) is usually an error. Jumping into or within Do groups is equally wrong.

Jumping (in REXX, signaling) within a DO loop is legal and is illustrated within the Regina PDF documentation. [What is wrong is merely a judgement call.] It's the execution of the DO or the END statement [after a signal instruction is executed] that is illegal and thus causes a syntax error. I don't want to endorse nor propose its use. I was merely showing what is legal, not what is considered bad practice (or wrongness) by some. I don't pontificate what should be considered/practiced/used when it comes to what is legal vs. what is bad programming. -- Gerard Schildberger 21:02, 9 December 2012 (UTC)

Your text:

"Note: some REXXes don't allow jumping into a DO loop, although the language specifications appear to allow it, as long as the END or the DO loop isn't executed. The following used PC/REXX to illustrate this example." is, in my opinion too liberal. It's not only Do loops where Signal is bad! --Walterpachl 08:22, 10 December 2012 (UTC)

What exactly is too liberal about my statement? I don't understand your concerns about what is allowed as far as the REXX language specification (as per the rules of using the SIGNAL instruction). I have no opinion about the validity of the REXX specifications, only that what was shown is a legal construct --- especially if the code is actually shown in the Regina REXX language specification ("The Regina Rexx Interpreter" PDF document). As for your opinion about the "badness" of the signal instruction, that's an opinion, but signal is still part of the REXX language, warts and all. -- Gerard Schildberger 09:00, 10 December 2012 (UTC)

Too liberal: That you talk only about DO LOOPS when do groups are affected as well. --Walterpachl 11:44, 10 December 2012 (UTC)

Then you should say what you mean. Restricting my comments about the legal syntax of the REXX signal instruction as mentioned in a language definition shouldn't be considered liberal. If I mention that the (Regina REXX) documentation mentions only DO loops (and not DO groups), then that shouldn't be labeled liberal; I'm just parroting what the documentation says, not interpreting it. The signal instruction shouldn't effect do groups, just active do loops (and a few others such as select groups and if statements). I say shouldn't because the Regina REXX documentation says one thing but implements another. Please take care about labeling the statements/opinions of others with judgemental names. Also note that the Regina documentation says the construct that I quoted as being legal, but the Regina REXX documentation appears to waffle (with statements like: Although TRL1 and TRL2 allow this construct, it will probably be disallowed in ANSI). I feel very uncomfortable and uneasy when language (specification) documentation uses such words like will probably. -- Gerard Schildberger 19:22, 10 December 2012 (UTC)

Is this the right place to discuss the Regina documentation which you seemingly do here? I haven't touched your text 'over there!'. Just tried to warn others (and maybe even you) about use of Signal. Please leave it at that and/or rephrase your text in the task if I convinced you to some extent.

I am talking here about classic Rexx and ooRexx, by the way. --Walterpachl 20:04, 11 December 2012 (UTC)

Example: <lang rexx> Do

  Signal label
  label: Say 'label reached'
End</lang>

Output:

     4 *-* End
Error 10 running H:\sig.rex line 4:  Unexpected or unmatched END
Error 10.1:  END has no corresponding DO, LOOP, or SELECT 

Please note that that your (illegal) code is "executing" the END instruction which has been rendered invalid by the execution of the signal instruction, which resulted in the syntax error (as cautioned in the comments in the REXX section header). -- Gerard Schildberger 21:02, 9 December 2012 (UTC)

This explain what I tried to show.

What you showed was an illegal REXX construct which resulted in a SYNTAX error. The code snipette isn't a valid REXX program (as mentioned earlier). I was showing a use for the SIGNAL instruction, not how to not use the SIGNAL instruction. -- Gerard Schildberger 09:00, 10 December 2012 (UTC)

I use the signal instruction for other than error condition handling or very global jumps for E-O-J and such. There are other uses for the signal instruction than that. -- Gerard Schildberger 21:02, 9 December 2012 (UTC)

Thats what I said:

"Please consider Signal only for condition handling or very global jumps such as Signal end_of_job."

What I was objecting mildly to was your suggestion that the signal instruction should only be considered for condition handling or very global jumps such as signal end-of-job. It has other uses. This isn't the forum for what REXX statements should or shouldn't be used. If it solves a problem, use it. If you have a better solution, include an example of its use. Rosetta Code is about showing solutions (or flavors of solutions) for the tasks specified. There's always room for more examples and methods. There are always more ways to skin a cat. -- Gerard Schildberger 09:00, 10 December 2012 (UTC)