Do not render small water areas on low zoom #2952

matthijsmelissen · 2017-11-20T18:34:33Z

On zoom levels 0-7, don't render small water areas (smaller than 32px).

On zoom levels 0-8, don't render small water areas (smaller than 32px).

matthijsmelissen · 2017-11-20T18:34:53Z

@rrzefox would it be possible to test this on your server?

kocio-pl · 2017-11-20T18:39:31Z

Could you also publish some before/after renderings? Probably the places that you gave as example of problems with small water areas would be good.

matthijsmelissen · 2017-11-20T18:39:46Z

@kocio-pl asked in #2925:

what filter do we want to use and up to which zoom level?

It currently bothers me both on zoom levels 6 and 7, and perhaps 8 and 9. So i think we definitely need measures for the zoom levels up to 7 (on levels 0-5 the small water areas are not very useful to me either).

I eyeballed 32px as a cut-off, let's see how it looks like when we have a rendering demo.

matthijsmelissen · 2017-11-20T18:41:21Z

Could you also publish some before/after renderings?

I don't have enough data in my database at the moment for realistic example renderings at these zoom levels.

rrzefox · 2017-11-20T22:27:53Z

@rrzefox would it be possible to test this on your server?

Done, and Z0-8 have already been rerendered.

matthijsmelissen · 2017-11-20T22:29:30Z

Cool! Demo at http://bl.ocks.org/matthijsmelissen/raw/af7a602c222dbf1ff1a2c0d84ed755b7/.

kocio-pl · 2017-11-20T22:34:19Z

Unfortunately it reintroduces #2580 on z0-z2.

matthijsmelissen · 2017-11-20T22:38:02Z

This looks like an improvement to me.

kocio-pl · 2017-11-20T23:02:39Z

On z0-z1 we don't see any of the biggest lakes (except Caspian, of course). On z2 Ontario, Erie and Onega are still not visible.

kocio-pl · 2017-11-21T03:05:08Z

I also think that filtering produces nice effects, so it's worth applying. On the other hand La Plata and big lakes are special, so it would be good to preserve them. The simplest solution would be to start filtering from z3 instead of z0. What do you think about it?

matthijsmelissen · 2017-11-21T07:24:39Z

Yes, I agree.

kocio-pl · 2017-11-21T14:14:42Z

water.mss

-      polygon-gamma: 0.6;
+  [landuse = 'basin']::landuse {
+    [zoom >= 7][way_pixels >= 32],
+    [zoom >= 8] {


On zoom levels 0-8, don't render small water areas (smaller than 32px).

So it should be:

[zoom >= 9] {

kocio-pl · 2017-11-21T14:58:02Z

water.mss

    }
  }

  [natural = 'water']::natural,
  [landuse = 'reservoir']::landuse,
  [waterway = 'riverbank']::waterway {
-    [zoom >= 0] {
+    [zoom >= 0][way_pixels >= 32],
+    [zoom >= 8] {


The same as above: [zoom >= 9] {

rrzefox · 2017-11-24T14:39:42Z

So some screenshot for later reference:
Z1 old:

Z1 new:

The big lakes between .us and .ca disappeared.
Z3 old:

Z3 new:

On Z3, all the mini-waters in Canada are gone. The same happens in northern europe (no screenshot).
In .nl, the Ijsselmeer disappears, it appears on Z4.

I have now deployed a slightly different version of this PR on the "demo server", it uses this less aggressive filtering:

  [natural = 'water']::natural,
  [landuse = 'reservoir']::landuse,
  [waterway = 'riverbank']::waterway {
    [zoom >= 0][way_pixels >= 4],
    [zoom >= 1][way_pixels >= 16],
    [zoom >= 2][way_pixels >= 32],
    [zoom >= 8] {

kocio-pl · 2017-11-24T14:49:48Z

Thanks! I was also thinking about progressive filtering and with these values it seems to be the best of both worlds, indeed. I would merge this version.

kocio-pl · 2017-12-03T00:15:40Z

@matthijsmelissen Do you plan to embrace the @rrzefox proposition? If it's just lack of time, I can fine tune the code and merge it soon.

matthijsmelissen · 2017-12-03T00:32:31Z

Generally I prefer steps of factor 4, as for every zoom level, lines get twice as large and areas four times as large. So for example 4-16-64 or 2-8-32. However, 4-16-32 might work as well (it means on z2 some additional water areas pop up, which might make sense).

Aside from the exact numbers, I'm happy with @rrzefox's approach. I was planning to get back at this some time, but I'd be glad if you can pick it up earlier.

kocio-pl · 2017-12-03T00:40:11Z

@rrzefox Were there some quirks which would be visible with 4-16-64 or 2-8-32 approach or was it just a free form handcrafting? If the latter is the case, could you test one of the proposed "4x" schemes?

rrzefox · 2017-12-06T19:36:42Z

For reference, this is how it looks with ~~4-16-32~~: Edit: Due to a bug, this actually showed 4-4-4 instead of the intended 4-16-32.

While I also tried to make steps with a factor of four, I found deviating from that in the last step acceptable. I had tried 2-8-32 but not liked the result because it did not filter enough. But see for yourself, I deployed that now... I'll deploy 4-16-64 on Friday.

kocio-pl · 2017-12-06T19:49:22Z

OK, let's see the 4-16-64 then.

BTW - because Matthijs has changed the account name, the link to compare is different now:

http://bl.ocks.org/matthijsmelissen/raw/af7a602c222dbf1ff1a2c0d84ed755b7/#2.00/18.8/14.7

matthijsmelissen · 2017-12-06T21:44:40Z

4-16-32 works in fact quite well, I added it to the PR now.

matthijsmelissen · 2017-12-06T21:47:32Z

On zoom levels 0-8, don't render small water areas (smaller than 32px).

I changed the documentation, it is now 0-7.

rrzefox · 2017-12-08T15:44:37Z

I've now deployed 4-16-64 for testing (Z0-3 rerendered)

kocio-pl · 2017-12-08T15:54:01Z

It looks OK for me.

On z0 La Plata is not visible, but the border too, so it's not a problem. Lack of Ontario, Erie, Ladoga, Onega and Victoria on z0 are also not visible if you're not looking for them. Anything on z1+ looks decent in both 4-16-32 and 4-16-64 for me, so it's up to you which version do you like more, Matthijs.

StyXman · 2017-12-08T19:23:30Z

Me being from Argentina I can say that this is a nice improvement, and I don't think anybody would complain about Río de la Plata not being present in z0.

matthijsmelissen · 2017-12-09T14:03:36Z

    [zoom >= 0][way_pixels >= 4],
    [zoom >= 1][way_pixels >= 16],
    [zoom >= 2][way_pixels >= 32],

@rrzefox This does not work. It will still render water areas of size 8 on z1 because the first line is still true.

matthijsmelissen · 2017-12-09T14:12:07Z

@rrzefox I have fixed this in my branch, would you be able to deploy my branch to your server?

I used 4-16-32, as filtering with 4-16-64 will almost certainly be too aggressive.

kocio-pl · 2017-12-09T15:50:38Z

@rrzefox I have fixed this in my branch, would you be able to deploy my branch to your server?

It was already tested and the screenshots are here: #2952 (comment)

matthijsmelissen · 2017-12-09T16:00:28Z

Nope, that test contained the bug I mentioned in #2952 (comment) .

matthijsmelissen · 2017-12-09T20:28:30Z

Just to be clear, the 4-16-32 deploy actually showed 4-4-4, the 2-8-32 deploy actually showed 2-2-2, and the 4-16-64 deploy actually showed 4-4-4 again.

rrzefox · 2017-12-10T18:31:03Z

The fixed 4-16-32 version has been deployed and Z0-7 rerendered.

kocio-pl · 2017-12-11T18:06:31Z

That's OK for me, we can merge it any time soon.

matthijsmelissen · 2017-12-11T18:23:03Z

@rrzefox Thanks!

@kocio-pl If you Approve it, I can merge it.

kocio-pl · 2017-12-17T23:16:58Z

I'm confused, it looks like z1 and z2 are wrong now - La Plata and smaller Big Lakes are missing now on both proposed and current versions (OSMF re-rendering is happening right now)...

http://bl.ocks.org/matthijsmelissen/raw/af7a602c222dbf1ff1a2c0d84ed755b7/#2.00/27.4/42.3

kocio-pl · 2017-12-24T09:52:21Z

Another thing that we could make better would be to not filter riverbanks, as suggested here:

http://www.openstreetmap.org/user/imagico/diary/42975#comment40531

matthijsmelissen · 2017-12-24T10:40:47Z

My intention was not to render riverbanks at all. The solution is exactly the opposite of what @imagico suggests: these riverbanks are way too big, and should be cut up in smaller parts.

kocio-pl · 2017-12-24T10:50:35Z

Current code still allows showing them, so maybe it'd be better to get rid of them completely on some initial zoom levels to avoid problems with some segments being visible and some others invisible.

matthijsmelissen · 2017-12-24T10:56:11Z

Makes sense.

matthijsmelissen · 2017-12-24T10:56:38Z

I'm confused, it looks like z1 and z2 are wrong now

Do you have any idea what could cause this?

kocio-pl · 2017-12-24T10:59:45Z

I guess we just got lost with testing all the possibilities on @rrzefox server. 😄 It was probably some small, stupid error but nobody checked it on his own machine.

Do not render small water areas on low zoom

5d750f0

On zoom levels 0-8, don't render small water areas (smaller than 32px).

matthijsmelissen mentioned this pull request Nov 20, 2017

Low-zoom: decide which features to render #2925

Closed

kocio-pl reviewed Nov 21, 2017

View reviewed changes

Less aggressive filtering of water on z0, z1 and z2

a7cfb23

Move cutoff for water on z2-7 to 64 pixels

bc1e777

Fix low-zoom water filtering

ba368d3

matthijsmelissen mentioned this pull request Dec 9, 2017

Release v4.6.0 #2973

Closed

kocio-pl approved these changes Dec 11, 2017

View reviewed changes

matthijsmelissen merged commit 63c05ac into gravitystorm:master Dec 11, 2017

kocio-pl mentioned this pull request Dec 25, 2017

Low zoom waterbody rendering and move to water polygons #2507

Open

kocio-pl mentioned this pull request Jan 19, 2018

Release v4.7.0 #3028

Closed

imagico mentioned this pull request Mar 5, 2018

[WIP] Render maritime borders differently #3102

Closed

jeisenbe mentioned this pull request Mar 7, 2020

Remove way_area filtering for low zoom water and simplify code #4060

Merged

imagico mentioned this pull request Oct 18, 2023

Ziplines should not be rendered at low zoom levels #4343

Open

Do not render small water areas on low zoom #2952

Do not render small water areas on low zoom #2952

Conversation

matthijsmelissen commented Nov 20, 2017 • edited Loading

matthijsmelissen commented Nov 20, 2017

kocio-pl commented Nov 20, 2017

matthijsmelissen commented Nov 20, 2017

matthijsmelissen commented Nov 20, 2017

rrzefox commented Nov 20, 2017

matthijsmelissen commented Nov 20, 2017 • edited Loading

kocio-pl commented Nov 20, 2017

matthijsmelissen commented Nov 20, 2017

kocio-pl commented Nov 20, 2017

kocio-pl commented Nov 21, 2017

matthijsmelissen commented Nov 21, 2017

kocio-pl Nov 21, 2017

Choose a reason for hiding this comment

kocio-pl Nov 21, 2017

Choose a reason for hiding this comment

rrzefox commented Nov 24, 2017

kocio-pl commented Nov 24, 2017

kocio-pl commented Dec 3, 2017

matthijsmelissen commented Dec 3, 2017 • edited Loading

kocio-pl commented Dec 3, 2017 • edited Loading

rrzefox commented Dec 6, 2017 • edited Loading

kocio-pl commented Dec 6, 2017

matthijsmelissen commented Dec 6, 2017

matthijsmelissen commented Dec 6, 2017

rrzefox commented Dec 8, 2017

kocio-pl commented Dec 8, 2017 • edited Loading

StyXman commented Dec 8, 2017

matthijsmelissen commented Dec 9, 2017

matthijsmelissen commented Dec 9, 2017

kocio-pl commented Dec 9, 2017

matthijsmelissen commented Dec 9, 2017

matthijsmelissen commented Dec 9, 2017

rrzefox commented Dec 10, 2017 • edited Loading

kocio-pl commented Dec 11, 2017

matthijsmelissen commented Dec 11, 2017

kocio-pl commented Dec 17, 2017

kocio-pl commented Dec 24, 2017

matthijsmelissen commented Dec 24, 2017

kocio-pl commented Dec 24, 2017

matthijsmelissen commented Dec 24, 2017

matthijsmelissen commented Dec 24, 2017

kocio-pl commented Dec 24, 2017 • edited Loading

matthijsmelissen commented Nov 20, 2017 •

edited

Loading

matthijsmelissen commented Nov 20, 2017 •

edited

Loading

matthijsmelissen commented Dec 3, 2017 •

edited

Loading

kocio-pl commented Dec 3, 2017 •

edited

Loading

rrzefox commented Dec 6, 2017 •

edited

Loading

kocio-pl commented Dec 8, 2017 •

edited

Loading

rrzefox commented Dec 10, 2017 •

edited

Loading

kocio-pl commented Dec 24, 2017 •

edited

Loading