Goto in Sequence Diagram?

Question

How can I show goto statements in Sequence Diagrams.

For example in the diagram below, once "sleep till hold period" is expired I want to take control back to the "is_item_inventory_onhold_state(item_id)" statement. How can I show this is a diagram?

I am using https://sequencediagram.org/ to create these diagrams.

Code written to generate the diagram above:

title Item Executor

loop  for each item in a list 
Client->ItemExecutor: execute(item)

ItemExecutor -> ItemStateService:is_item_inventory_onhold_state(item_id)

alt True - Item state is on hold
ItemStateService -->ItemExecutor: True
ItemExecutor ->ItemExecutor: sleep till hold period 

goto  ItemExecutor -> ItemStateService:is_item_inventory_onhold_state(item_id)

else False - Item is not in Held State
ItemStateService -->ItemExecutor:False
ItemExecutor ->ItemExecutor: do_something()

end

ItemExecutor ->Client : Acknowledge
end 

Answer 1

Goto is not supported in sequence diagrams (for good reasons). Use a combination of a loop and a break operator instead. See this diagram: sequencediagram.org/Item Executor

sequencediagram.org/Item Executor (with Execution Specifications)

Some remarks on this diagram

• the break-fragment leaves the immediately enclosing fragment. Therefore it must be contained directly in the loop-fragment. If it is within an alt-fragment, only this fragment is left.
• Both alt and opt fragments don't use a guard. The fragment that happens is choosen by the occurrence of the reply message with a certain return-value. If you want to use a guard, you have to assign the return-value to a local variable. This would then happen above the alt-fragment (see diagram below).
• return values are shown with a preceeding colon. The message name would preceed this, but when it is obvious it can be omitted (as it is here).
• execution specifications (sometimes referred to as "activations") are only shown, where they help the readability. Contrary to popular belief they are not mandatory.
• UML doesn't know for each loops. Therefore I added iterator operations. The term "for each item in a list" is not a guard condition. If you want to avoid spelling out the iterator, you can use a - semantic free - comment attached to the loop. Misusing the boolean guard for this makes no sense. If you want a formal definition, you have to add you own stereotype «for each loop»
• I assume ItemExecutor and ItemStateService are class-names. They need a preceeding colon to distinguish them from role-names. Of course if the role- and class-names are identical, your diagram can be correct.
• The "Acknowledge" message is just the reply message for the execute message. As such it would carry the same name (which is omitted here).
• In the version with Execution Specifications, the drawing tool didn't allow the end of the execution specifications to coincide with the send events of the reply-messages, which would have been correct.

Example with guards for the alt and break fragments (excerpt): sequencediagram.org/Item Executor (with Execution Specifications and guards)

Answer 2

You do not need a 'goto', like we don't use them in language since long time

Add a second loop and go out with a combined fragment 'break'

Answer 3

A goto itself is not shown. You just show what operations are sent. You might add a note at the place where the goto happens. However, I think that the usage of gotos should not be encouraged in any way. Instead it should probably some exception handling.

According to your comments you can use a break fragment like this:

It will break the outer loop, so the is_item... is repeated after sleep....

Note As per comment from @AxelScheithauer the break will just leave the enclosing fragment. However, I'd regard this a specification flaw. A break is "usually" used with loop control flows (plus a case control flow; but UML does not have a fragment for that). It's probably the best to name the break fragment so it's clear it will affect the outer loop fragment (as shown in my edited picture).