Re: [MapServer-users] Clipped labels in mapcache/mapserver map

2022-11-03 Thread Thomas Ellett von Brasch 
Thanks for taking a look Steve!

I wondered about the angle follow as well, but I’ve been hoping its not that as 
it’d be a real shame to have to lose those. And I’ve tried with both negative 
and positive values for map edge buffer, but it didn’t seem to make a 
difference, and I read somewhere (mapserver dcos I guess ) that the absolute 
value is now used.

I’ll try recreating a single metatile and let you know what happens.

/Tom

Fra: Steve Lime 
Sendt: torsdag 3. november 2022 14:57
Til: Thomas Ellett von Brasch 
Kopi: mapserver-users@lists.osgeo.org
Emne: Re: [MapServer-users] Clipped labels in mapcache/mapserver map

VARSEL: Avsenderen er fra utenfra organisasjonen. Vær trygg på innholdet før du 
trykker på linker eller åpner vedlegg. Er du i tvil tar du kontakt med IT 
Servicedesk.

So I checked a map here and found similar settings:

  *   label-level: PARTIALS FALSE,
  *   map-level metadata: "labelcache_map_edge_buffer" "-20"
  *   mapcache: 20
I don't use the LABEL_NO_CLIP processing directive. The one obvious difference 
is the negative value for the map edge buffer - I know negative values were 
originally how that was specified but I thought we generalized that value some 
time ago.

I wonder if the "ANGLE follow" might be the cause - perhaps there's an error in 
the size computation of the rotated text - perhaps at different position 
values? Just guessing though.

The only way to easily debug/test is going to be examining a metatile that 
results in a clipped label. You could delete an affected tile and then let 
mapcache generate a replacement. You should be able to pick up the metatile WMS 
call from the logs.

--Steve

On Tue, Nov 1, 2022 at 8:25 PM Steve Lime 
mailto:sdl...@gmail.com>> wrote:
Have you tried a larger edge buffer? I suppose that’s tough to to test given 
the randomness. I’ll check my settings on a similar setup when I’m in the 
office tomorrow and let you know. It would be interesting to see the 
corresponding metatile to see if the label is whole there. I wouldn’t think so 
but having a reproducible test case would help immensely.

Anyone know how to compute a metatile URL for a given tile x/y/z?

—Steve

On Mon, Oct 31, 2022 at 10:16 AM Thomas Ellett von Brasch 
mailto:thomas.ell...@kartverket.no>> wrote:
Hi all,

I'm really struggling to fix a longstanding issue with clipped labels in a 
couple of our wmts services.
I've tried following the mapserver/mapcache docs and numerous variations of 
different attributes, but with no luck.

The main problem bug fixing is that the issue seems to be quite random, just 
popping up here and there.

Examples can be seen in these urls:

https://norgeskart.no/#!?project=norgeskart=1002=11=6879652.31=167744.51

(the 'Stor rar' and 'S' names in the centre of the map)

https://norgeskart.no/#!?project=norgeskart=1002=6=6440416.41=127931.07

('Kristiansan' , which should be 'Kristiansand')

I have "labelcache_map_edge_buffer" "10" in the mapfile and 
10 set in the mapcache file.

In the source WMS, the placenames layers are line layers (with very small line 
geometries used just to give curved lines) and the layers (we have several name 
layers for the different zoom levels, but they're all set up in the same way) 
look like (truncated):

LAYER
NAME "N100Stedsnavn"
PROCESSING "CLOSE_CONNECTION=DEFER"
PROCESSING "LABEL_NO_CLIP=True"
TYPE LINE

CLASS
  NAME "N100 Stedsnavn"
  EXPRESSION ([fmx_rotation] != 0)
  LABEL
  PARTIALS FALSE
ANGLE follow
ANTIALIAS TRUE
FONT "verdana"
SIZE [fontsize]
COLOR [geodb_color]
ENCODING "UTF-8"
  PRIORITY [fontsize]
FORCE FALSE
  MINDISTANCE 50
 OUTLINECOLOR 254 254 254
 POSITION [horizontalalignment]
 TYPE TRUETYPE
  END
  TEXT "[textstring]"
END
END

And the tileset definition in the mapcache config file looks like:

Re: [MapServer-users] Clipped labels in mapcache/mapserver map

2022-11-03 Thread Steve Lime 
So I checked a map here and found similar settings:

   - label-level: PARTIALS FALSE,
   - map-level metadata: "labelcache_map_edge_buffer" "-20"
   - mapcache: 20

I don't use the LABEL_NO_CLIP processing directive. The one obvious
difference is the negative value for the map edge buffer - I know negative
values were originally how that was specified but I thought we generalized
that value some time ago.

I wonder if the "ANGLE follow" might be the cause - perhaps there's an
error in the size computation of the rotated text - perhaps at different
position values? Just guessing though.

The only way to easily debug/test is going to be examining a metatile that
results in a clipped label. You could delete an affected tile and then let
mapcache generate a replacement. You should be able to pick up the metatile
WMS call from the logs.

--Steve

On Tue, Nov 1, 2022 at 8:25 PM Steve Lime  wrote:

> Have you tried a larger edge buffer? I suppose that’s tough to to test
> given the randomness. I’ll check my settings on a similar setup when I’m in
> the office tomorrow and let you know. It would be interesting to see the
> corresponding metatile to see if the label is whole there. I wouldn’t think
> so but having a reproducible test case would help immensely.
>
> Anyone know how to compute a metatile URL for a given tile x/y/z?
>
> —Steve
>
> On Mon, Oct 31, 2022 at 10:16 AM Thomas Ellett von Brasch <
> thomas.ell...@kartverket.no> wrote:
>
>> Hi all,
>>
>>
>>
>> I'm really struggling to fix a longstanding issue with clipped labels in
>> a couple of our wmts services.
>>
>> I've tried following the mapserver/mapcache docs and numerous variations
>> of different attributes, but with no luck.
>>
>>
>>
>> The main problem bug fixing is that the issue seems to be quite random,
>> just popping up here and there.
>>
>>
>>
>> Examples can be seen in these urls:
>>
>>
>>
>>
>> https://norgeskart.no/#!?project=norgeskart=1002=11=6879652.31=167744.51
>>
>>
>>
>> (the 'Stor rar' and 'S' names in the centre of the map)
>>
>>
>>
>>
>> https://norgeskart.no/#!?project=norgeskart=1002=6=6440416.41=127931.07
>>
>>
>>
>> ('Kristiansan' , which should be 'Kristiansand')
>>
>>
>>
>> I have "labelcache_map_edge_buffer" "10" in the mapfile and
>> 10 set in the mapcache file.
>>
>>
>>
>> In the source WMS, the placenames layers are line layers (with very small
>> line geometries used just to give curved lines) and the layers (we have
>> several name layers for the different zoom levels, but they're all set up
>> in the same way) look like (truncated):
>>
>>
>>
>> LAYER
>>
>> NAME "N100Stedsnavn"
>>
>> PROCESSING "CLOSE_CONNECTION=DEFER"
>>
>> PROCESSING "LABEL_NO_CLIP=True"
>>
>> TYPE LINE
>>
>>
>>
>> CLASS
>>
>>   NAME "N100 Stedsnavn"
>>
>>   EXPRESSION ([fmx_rotation] != 0)
>>
>>   LABEL
>>
>>   *PARTIALS FALSE*
>>
>> *ANGLE follow*
>>
>> ANTIALIAS TRUE
>>
>> FONT "verdana"
>>
>> SIZE [fontsize]
>>
>> COLOR [geodb_color]
>>
>> ENCODING "UTF-8"
>>
>>   PRIORITY [fontsize]
>>
>> *FORCE FALSE*
>>
>>   MINDISTANCE 50
>>
>>  OUTLINECOLOR 254 254 254
>>
>>  POSITION [horizontalalignment]
>>
>>  TYPE TRUETYPE
>>
>>   END
>>
>>   TEXT "[textstring]"
>>
>> END
>>
>> END
>>
>>
>>
>> And the tileset definition in the mapcache config file looks like:
>>
>>
>>
>> 
>>
>> norgeskart_bakgrunn
>>
>> norgeskart_bakgrunn
>>
>> 
>>
>> Topografisk Norgeskart og dybdedata og svalbard minus
>> matrikkel
>>
>>Cache for Norgeskart
>>
>> 
>>
>> googlemaps
>>
>> utm33n
>>
>> PNG
>>
>> *8 8*
>>
>> *10*
>>
>> 
>>
>>
>>
>> Is it possible the metatiles being used in the original seed job and
>> later seeding jobs are different somehow?
>>
>>
>>
>> Given that the labelcache_map_edge_buffer is set (along with metabuffer)
>> and we have PARTIALS FALSE in the mapfile, I didn't think that cut labels
>> would be possible?
>>
>>
>>
>> Any thoughts or ideas would be great J
>>
>>
>>
>> Thanks in advance!
>>
>>
>>
>> Tom
>>
>>
>>
>>
>>
>> [image: cid:image001.jpg@01D3A415.20CB4FE0]
>>
>> *Tom Ellett von Brasch*
>> Sjefsingeniør
>> *Geodatatjenester*
>>
>> Direkte: 32 11 84 73
>>
>> Mobil: 45806402
>>
>> E-post: thomas.ell...@kartverket.no
>>
>>
>> Tlf. sentralbord: 32 11 80 00
>> www.kartverket.no
>>
>>
>>
>>
>> ___
>> MapServer-users mailing list
>> MapServer-users@lists.osgeo.org
>> https://lists.osgeo.org/mailman/listinfo/mapserver-users
>>
>
___
MapServer-users mailing list
MapServer-users@lists.osgeo.org
https://lists.osgeo.org/mailman/listinfo/mapserver-users