This is a kind of ad-hocish response to an interesting problem. In
previous Syndicates, assertion changes were gathered into *patches*
which *netted out* intra-turn changes. In this implementation, each
change is relayed unchanged, so there is *no netting* going on. The
closest we get is a convention that when things are being replaced,
incoming assertions should be made before outgoing ones are retracted,
so that the kind of glitch that's seen is a double-up of records,
rather than a brief window where no records are present.
So here, by moving the stop continuation into an onStop for the
stopping facet (that nonetheless *executes* in that facet's parent's
context), we allow the following pattern to run without problem
glitching:
react {
assert Something(1);
stop on asserted Condition() => react {
assert Something(2);
}
}
Observers like this...
during Something(_) => { // **A**
...
during Something($specific) => { // **B**
}
}
will see
+ Something(1)
+ Something(2)
- Something(1)
rather than what they saw before this change,
+ Something(1)
- Something(1)
+ Something(2)
The consequence is that the line marked **A** above will *remain
active* after this change, with the facet at **B** being replaced,
rather than briefly tearing down the **A** facet when `- Something(1)`
and then creating a fresh one when `+ Something(2)`.
This came up in flappy bird, where I had
react {
at mainDs {
assert ui.html('#board-area', template`<h1 class="score">${score.value}</h1>`);
}
at gameDs {
stop on asserted GameOver() => react {
at mainDs {
assert ui.html(
'#board-area',
template`<h1 class="score">${score.value}<br/>GAME OVER</h1>`);
}
}
}
}
which, in combination with the particular implementation of index.ts
in the html package, caused the UIFragment responder to get confused
and to not show the GAME OVER message - the previous score message was
removed, but the new one wasn't manifested in the DOM. Changing the
last `ui.html` above to `ui.context('foo').html` caused the problem to
go away (by making it effectively two unrelated fragments, rather than
a replacement of the content of a single fragment).
Tony Garnock-Jones