Fix automap for infinite maps #1692

ketanhwr · 2017-08-20T22:30:01Z

This patch aims to fix the automap issue in infinite-maps.

bjorn

See inline comments.

Does this also make infinite maps work for automapping rules?

bjorn · 2017-08-21T08:28:20Z

src/tiled/automapper.cpp

+    int endY = dstY + height;
+
+    if (!mMapWork->infinite()) {
+        startX = qMax(dstX, startX);


dstX and startX are the same at this point. You meant 0 instead of dstX.

bjorn · 2017-08-21T08:49:48Z

src/libtiled/tilelayer.cpp

+        r &= QRect(dx, dy, other->width(), other->height());
+    } else {
+        r = rect;
+    }


Why do you need the rect parameter? As far as I can see, r just needs to cover both layers and it should be in local coordinates, so this should do it:

const QRect r = bounds().united(other->bounds()).translated(-position());

ketanhwr · 2017-08-21T10:53:42Z

No, it doesn't work if the automap rules are infinite as well, right now.

bjorn · 2017-08-21T11:29:13Z

src/tiled/automapper.cpp

@@ -691,7 +691,7 @@ static bool compareLayerTo(const TileLayer *setLayer,
                bool matchListNo  = false;


-                if (!setLayer->contains(x + offset.x(), y + offset.y())) {
+                if (!(setLayer->contains(x + offset.x(), y + offset.y()) || setLayer->map()->infinite())) {


Intuitively, this check and the special handling inside should actually not be necessary at all. If the setLayer does not contain the given location, then c1 will just be an empty cell, which I think should be treated the same as a cell outside of the layer boundaries. Can you check this?

No no, it needs to have that check. I found it strange too, but it doesn't work if this check is removed.

Can you give an example of where it goes wrong?

This is what happens when I remove the check:

This is what happens when I keep the check (this obviously is the expected behaviour):

What part did you try removing exactly? I suggested removing this entire bit of code:

if (!(setLayer->contains(x + offset.x(), y + offset.y()) || setLayer->map()->infinite())) { foreach (const TileLayer *comparedTileLayer, listYes) { if (!comparedTileLayer->contains(x, y)) return false; const Cell &c2 = comparedTileLayer->cellAt(x, y); if (!c2.isEmpty()) return false; } continue; }

It's not clear to me what is happening from those screenshots.

Aah, I thought you meant setLayer->map()->infinite()

bjorn · 2017-08-24T15:33:09Z

Reminder to self, or when you have time: need to check this change against issue #1224 (since that is where this special code we have now removed was introduced).

bjorn · 2017-08-24T20:58:53Z

Alright, I checked this against issue #1224 and unfortunately it re-introduces that issue. However, I think the consistency is more important than getting that particular example to work. The example in issue #1224 could be adjusted to have separate input and output regions in order to make the rule work at the border.

However, I found an issue with the automapping in that even after this patch it doesn't work with infinite maps outside of the already used bounds. Please open this example.zip and try to place a rock somewhere. You should see that the first time it works, but then when you paint rocks in many places, it only places the flowers within the chunks that were allocated the first time the automapping was activated.

ketanhwr · 2017-08-25T04:37:10Z

Oh 😕

I'll try to fix this.

ketanhwr · 2017-08-26T12:03:14Z

What I've observed is when we apply automap and then paint outside the bounds and try automap again it doesn't work. Then if we undo the first automap and then paint outside the bounds and then try automap again, it works. What exactly is stored after calling 'Automap' for the first time ?

ketanhwr · 2017-08-26T14:52:14Z

Okay, so after spending the past 3 hours into this, I can say that I've wasted the whole time 🙁

So actually, automapping works even if we paint out-of-bounds. If you test after my previous commit, you cannot visually see the automapped tile layer. If you save it and then restart Tiled, the automapped tile layer is clearly visible 🤦‍♂️

I have no idea why this is happening. Here's a gif of the same.

😢

bjorn

Okay, so after spending the past 3 hours into this, I can say that I've wasted the whole time

Actually it's not wasted, because you made me realize that it's probably to do with the bounds changed signal not being emitted. You know the thing you added to the TileLayerChangeWatcher. Can you verify this and look into a fix if that's the reason?

Your change to TileLayer::copy looks like a sensible code cleanup, so please keep that part.

bjorn · 2017-08-26T20:50:08Z

src/libtiled/tilelayer.cpp

@@ -236,7 +233,10 @@ void TileLayer::setCells(int x, int y, TileLayer *layer,
                         const QRegion &mask)
 {
    // Determine the overlapping area
-    QRegion area = QRect(x, y, layer->width(), layer->height());
+    int width = qMax(layer->width(), layer->bounds().width());
+    int height = qMax(layer->height(), layer->bounds().height());


I think this makes little sense, please revert it back to what it was.

But why? Suppose a layer was created with size (1, 1) and then some cells were set out-of-bounds. If this layer is then used in setCells for some other layer, then technically, these 'out-of-bounds' cells should also be set in this layer. Don't you think?

Well, the part that does not make sense, is to take the width/height of the bounds but still copy starting from 0,0.

So I guess we need to unite QRect(0, 0, layer->width(), layer->height() with layer->bounds().

And adjust the offset accordingly.

No, if you use the bounds then the rect() of the layer is completely irrelevant.

I got confused by something else 😕

ketanhwr · 2017-08-27T09:02:28Z

Okay, automap is fixed now. But still some problem persists. When we undo, we need to zoom in/out or move around to see the effect of undo. This is because when we undo, the bounds of Tile Layer are not changed. But I don't think it's that big of a problem.

bjorn · 2017-08-27T10:25:22Z

Right, so this signal should actually be made part of the undo command. Would that be hard to do?

ketanhwr · 2017-08-27T10:49:54Z

This is because when we undo, the bounds of Tile Layer are not changed.

This is primarily the reason, the signal cannot emit. Since the bounds do not change after removing cells, we cannot actually check if it is necessary to emit this signal.

bjorn · 2017-08-27T10:56:52Z

This is primarily the reason, the signal cannot emit. Since the bounds do not change after removing cells, we cannot actually check if it is necessary to emit this signal.

If the bounds do not change, it is not necessary to emit the signal. Then there must be some other thing missing to trigger an update.

ketanhwr · 2017-08-27T11:15:12Z

Done! 🙂

bjorn · 2017-08-27T20:01:26Z

Nice, thanks!

Fix automap

350a81e

bjorn requested changes Aug 21, 2017

View reviewed changes

Review Changes

8eba7ba

bjorn reviewed Aug 21, 2017

View reviewed changes

Removes unnecessary code

7790f36

Trying to fix

27f8a62

bjorn requested changes Aug 26, 2017

View reviewed changes

Automap fixed!

44252f0

ketanhwr added 3 commits August 27, 2017 16:32

Revert change in setCells

c6f1878

Bug fix in Automap Undo

18d2380

Better fix

0112c91

bjorn merged commit e493426 into mapeditor:master Aug 27, 2017

bjorn changed the title ~~Fix automap~~ Fix automap for infinite maps Dec 31, 2017

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Fix automap for infinite maps #1692

Fix automap for infinite maps #1692

ketanhwr commented Aug 20, 2017

bjorn left a comment

bjorn Aug 21, 2017

bjorn Aug 21, 2017

ketanhwr commented Aug 21, 2017

bjorn Aug 21, 2017

ketanhwr Aug 21, 2017

bjorn Aug 21, 2017

ketanhwr Aug 21, 2017

bjorn Aug 21, 2017

ketanhwr Aug 21, 2017

bjorn commented Aug 24, 2017 •

edited

Loading

bjorn commented Aug 24, 2017

ketanhwr commented Aug 25, 2017

ketanhwr commented Aug 26, 2017

ketanhwr commented Aug 26, 2017

bjorn left a comment •

edited

Loading

bjorn Aug 26, 2017

ketanhwr Aug 27, 2017

bjorn Aug 27, 2017

ketanhwr Aug 27, 2017

ketanhwr Aug 27, 2017

bjorn Aug 27, 2017

ketanhwr Aug 27, 2017

ketanhwr commented Aug 27, 2017

bjorn commented Aug 27, 2017

ketanhwr commented Aug 27, 2017

bjorn commented Aug 27, 2017

ketanhwr commented Aug 27, 2017

bjorn commented Aug 27, 2017

Fix automap for infinite maps #1692

Fix automap for infinite maps #1692

Conversation

ketanhwr commented Aug 20, 2017

bjorn left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ketanhwr commented Aug 21, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

bjorn commented Aug 24, 2017 • edited Loading

bjorn commented Aug 24, 2017

ketanhwr commented Aug 25, 2017

ketanhwr commented Aug 26, 2017

ketanhwr commented Aug 26, 2017

bjorn left a comment • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ketanhwr commented Aug 27, 2017

bjorn commented Aug 27, 2017

ketanhwr commented Aug 27, 2017

bjorn commented Aug 27, 2017

ketanhwr commented Aug 27, 2017

bjorn commented Aug 27, 2017

bjorn commented Aug 24, 2017 •

edited

Loading

bjorn left a comment •

edited

Loading