Delay before the next loop starts

[efaulhaber]

Is there a way to set a delay before the next loop starts? I would love to have a few seconds to look at the final output before it disappears when the next loop iteration is started. Currently, it starts immediately, so the last lines are impossible to read.

I checked the Asciinema docs, and didn't find anything useful to make a PR.
I also thought about adding a sleep(4) at the end, but I would like this not to be visible. Is it possible to hide only one input while the other inputs are shown?

[ericphanson]

Can you try suffixing the line with #hide? That works in documenter and maybe we get it for free since we use documenter for code execution. Or maybe we need to implement something here

[ericphanson]

ah, I think #hide works to remove things from the cast (at least with the @cast_str macro, haven't tested the documenter @cast blocks), but if there was a sleep at the end, it wouldn't affect the timing it would just delete it out of the cast, so it would be like there's no sleep at all.

[efaulhaber]

Yeah, I'm just experimenting with it. I get some empty lines with a 1s delay, no matter how long the sleep is.

[efaulhaber]

Okay, so a sleep will only sleep for about a second, even if it is longer.
Short is possible, though, so we want to use 1-second sleeps.
Multiple sleeps in a loop also don't work, unless something is printed in between.
A print() is not enough, we need to print something, like print(" ").
Printing a white space makes the cursor move, so I print("\r") instead.

TL;DR This here works nicely:

```@cast; delay=0.1, loop=true
something()

for _ in 1:5 # See https://github.com/ericphanson/Asciicast.jl/issues/38    # hide
    print("\r")                                                             # hide
    sleep(1)                                                                # hide
end                                                                         # hide
```

[ericphanson]

Ah, that's clever!

so a sleep will only sleep for about a second, even if it is longer

I believe this is idle_time_limit - I had hardcoded it to 1.0 apparently. I made it configurable in #39 and added a loop_delay option which should pause at the end of the cast. One nice thing about that version is that it doesn't actually slowdown docs generation since the pause is by faking a timestamp.