[OSM-newbies] blue square bues

[David Groom]

----- Original Message ----- 
From: "Arunmozhi" <aruntheguy at gmail.com>
To: <newbies at openstreetmap.org>
Sent: Tuesday, June 08, 2010 5:57 PM
Subject: Re: [OSM-newbies] blue square bues


>
> I face a similar problem in one of my edits. I edited a river bed
> region in india (sorry i don't have any lat long data at hand). And
> there is a blue block extending over few tiles. It doesn't cover full
> tiles but a rectangular stretch. One thing that i have noticed is that
> it is only in osmrender.
>
> Can someone tell why it all happens.
>
> Arunmozhi
>
> On 6/8/10, doug brown <dougcb68 at hotmail.com> wrote:
>>
>> For the past few weeks I have been editing coastlines and roads South of 
>> the
>> Rio Grande de Santiago in western Mexico (permalink =
>> http://www.openstreetmap.org/?lat=21.637&lon=-105.226&zoom=11&layers=0B00TTF)
>> in JOSM.
>>
>> A couple of weeks ago a blue square appeared when viewing the area in the
>> Osmarender.  I cannot find anything in JOSM that would cause this.
>>
>> I waited for couple of weeks for the data base to digest my edits of this
>> area, expecting it to go away. But it hasn't.
>>
>> It does not appear in mapnik.  The issue persists in all browsers 
>> (firefox,
>> IE, chrome).
>>
>> Any thoughts or suggestions?
>>
>>

Hopefully my explanation of a possible cause is still a valid explanation, 
and that the process used to produce the Osmarender layer has not changed.

Basically in order to determine whether a tile is ocean, land , or part land 
and ocean, the Tiles at home (T at H) clients which produce the Osmarender tiles 
look at  this data file:

http://trac.openstreetmap.org/browser/applications/rendering/png2tileinfo/oceantiles_12.dat

the process is explained a bit more in

http://trac.openstreetmap.org/browser/applications/rendering/png2tileinfo/README

Essentially :

i) if a tile is expected to be all ocean (by reference to the 
oceantiles_12.dat file) then T at H simply delivers a blue square;
ii) if a tile is expected to be all land (by reference to the 
oceantiles_12.dat file)then it is rendered in the normal manner
ii) if a tile is expected to be part land and part ocean (by reference to 
the oceantiles_12.dat file) then a "special process" kicks in to produce the 
correct rendering. this "special process" looks for ways tagged as 
"natural=coastline" and from them creates a closed area to render as blue, 
leaving roads and other objects rendered in the non-blue area.  In general 
this works OK.

Problems arise however when a tile which did contain a way marked as 
"natural=coastline" gets edited and the coastline way is removed, most 
usually by replacing with a tag such as "waterway=riverbank".  The T at H 
clients look in the oceantiles_12.dat file and expect to see some coastline 
ways, but these have been removed from the main OSM database, and this can 
cause the whole tile to be marked as blue.

Ideally when an area which previously had some ways tagged as 
natural=coastline has these tags removed ,then the oceantiles_12.dat file 
needs to be updated to mark that tile as now "all land".

According to the last line of 
http://trac.openstreetmap.org/browser/applications/rendering/png2tileinfo/README 
the easiest way to do this is to grab the tile number and email it to T at h 
mailing list at tilesathome at openstreetmap.org.


Apologies to the T at H experts for my over simplification, and to the non TAH 
experts for probably still confusing them

Regards

David