stop-when w/last-picture handler about code.pyret.org

[jpolitz]

Got it. That's really nice, actually. It maintains the invariant that to-draw is only called with worlds for which stop-when returns true. This is straightforward to add.

[schanzer]

Poking since this will likely become a higher priority as spyret progresses...

[jpolitz]

Can you say a sentence about spec for this, like where this last-picture gets rendered? Does it just stay up in the big-bang window, or does it get shown at the REPL or somewhere else?

@sorawee added close-when-stop, which controls whether or not the big-bang window stays open on a stop-when, and it stays open on the first frame for which stop-when returned false. Is that accomplishing the same thing?

For example, if I do:

big-bang(0, [list: on-tick(lam(w): w + 1 end), to-draw(lam(w): text(tostring(w), 24, "black") end), stop-when(lam(w): w > 15 end)])

The final frame shows 16, not 15. Is this the goal?

[schanzer]

See http://docs.racket-lang.org/teachpack/2htdpuniverse.html#%28form._world._%28%28lib._2htdp%2Funiverse..rkt%29._stop-when%29%29

[sorawee]

Just curious: can't normal big-bang do this already, using a combination of on-tick, stop-when, and to-draw? This is basically what I did for Rocket (CS19 assignment) to get the same functionality as this last-picture.

[jpolitz]

@sorawee yes, this doesn't add new expressiveness. However, it's useful that last-picture doesn't need to be the same as what to-draw would generate, so the to-draw function can be written to only consider the non-stopping range of values.

[shriram]

@sorawee I had this added to Racket because in its absence, you have to engage in a complex encoding that is inaccessible to most beginners. (For instance, sometimes you have to augment your notion of state — your world becomes a conditional of "game in progress" vs "game over". Not only does every single handler then has to have a cases, worst of all, most handlers will never see the "game over" state: so the only thing they can say in that branch of the cases is "should not get here and if I do, I have no idea what to do".)

Since it's very natural for students to want to put up a splash screen when their game finishes (you won! you lost! your score is …!), this is a key feature to make nicer-feeling games accessible.

[blerner]

Dupe of brownplt/pyret-lang#808?

[shriram]

@blerner Not quite. The main thrust of that report is the semantics of when the handlers are called; last-picture is a bit of a side-track from that:
brownplt/pyret-lang#808 (comment)
The main thrust of this is to add a last-picture feature at all.

[schanzer]

There's a PR for this at #218

[schanzer]

big-bang has been deprecated, no? If so, let's close this. If not, what else needs to be done with the PR?

[shriram]

Before you close, confirm that it's not a problem with reactor.

[schanzer]

@shriram I can confirm that this is implemented (and then some!) in reactors. Closing, but feel free to reopen if you disagree