monitoring of unfinished activities

[bart6114]

I've ran into some results that are unintuitive for me. Consider the following

library(simmer)

a <- create_trajectory() %>%
  timeout(11)

b <- create_trajectory() %>%
  timeout(1) %>%
  rollback(1, times=Inf)

c <- create_trajectory() %>%
  timeout(5)

simmer() %>%
  add_generator("a", a, at(0)) %>%
  add_generator("b", b, at(0)) %>%
  add_generator("c", c, at(0)) %>%
  run(until=10) %>%
  get_mon_arrivals()

Which results in:

  name start_time end_time activity_time finished replication
1   c0          0        5             5     TRUE           1

I would expect the a and b arrival to also show up with an end_time equal to until and a finished status of FALSE.

Not looked into it too deep, but I think we can relatively easily implement this by making a terminate call on all the active arrivals after the Simulator::run function runs out.

What do you think @Enchufa2?

[Enchufa2]

For me, the nice thing is that you can continue the simulation. If you call terminate when the until condition is met, you cannot do that.

[bart6114]

Hmz, you're completely right.
What do you think about a terminate(env) command?
This could be used stand alone, or could optionally be called by e.g. get_mon_arrivals().

Either way, this behaviour is something we should describe in more detail in #29

[Enchufa2]

I see no benefit from aborting trajectories. There is no need: we can simply report these ongoing processes without terminating them. It would be an extra option to get_mon_arrivals: ongoing=TRUE (FALSE by default), for instance. Then, these ongoing arrivals should be reported with the finished column, as well as the end_time and activity_time columns, equal to NA, to avoid messing around with the other data.

So following your example above,

library(simmer)

a <- create_trajectory() %>%
  timeout(11)

b <- create_trajectory() %>%
  timeout(1) %>%
  rollback(1, times=Inf)

c <- create_trajectory() %>%
  timeout(5)

env <- simmer() %>%
  add_generator("a", a, at(0)) %>%
  add_generator("b", b, at(0)) %>%
  add_generator("c", c, at(0)) %>%
  run(until=10)

get_mon_arrivals(env, ongoing = TRUE)
#>   name start_time end_time activity_time finished replication
#> 1   c0          0        5             5     TRUE           1
#> 2   a0          0       NA            NA       NA           1
#> 3   b0          0       NA            NA       NA           1

The advantage of not having terminated those two is that you can now do the following:

env %>% run(until=12)

get_mon_arrivals(env, ongoing = TRUE)
#>   name start_time end_time activity_time finished replication
#> 1   c0          0        5             5     TRUE           1
#> 2   a0          0       11            11     TRUE           1
#> 3   b0          0       NA            NA       NA           1

[Enchufa2]

The activity_time column will be NA for those ongoing arrivals (edited above).

[bart6114]

great work 👍