logo separator

[mkgmap-dev] Problem in splitter (Africa)

From GerdP gpetermann_muenchen at hotmail.com on Wed Jan 7 16:04:39 GMT 2015

Hi all,

I've committed r417 to fix a few rare problems in splitter:
1) The initial problem reported by Stephen:
splitter printed "Warning: No solution found for partition ..." but did not
stop when at least one partition was ok.
2) It tries a bit harder to find a solution when it turns out
that no good solution is found.
3) It prints a warning when the user choses a rather bad combination
of --max-nodes and --resolution.

I've also removed the code that optimized parts of a solution,
it did not really work in most cases. I'll try to find a solution for that
again
later.

Gerd


GerdP wrote
> Hi Steve,
> 
> I suggest to look at the rules in the default style and the style 
> manual:
> http://www.mkgmap.org.uk/doc/pdf/style-manual.pdf
> 
> I am not that familiar with style rules, and I prefer java coding.
> 
> Gerd
> 
> Date: Wed, 7 Jan 2015 17:15:59 +1000
> From: 

> steve.sgalowski@

> To: 

> mkgmap-dev at .org

> Subject: Re: [mkgmap-dev] Problem in splitter (Africa)
> 
> gerd p would love more on the routing mate in my map but when i tried
> before to upgrade the script file , it did not work may be i send the
> script file i use ,  so you can help me convert it over to the new
> scripting you do mate 
> or provide me with plenty of examples and i can try gerd 
> stephen 
> 
> On Wed, Jan 7, 2015 at 5:11 PM, Gerd Petermann <

> gpetermann_muenchen@

> > wrote:
> 
> 
> 
> Hi Stephen,
> 
> okay, I forgot to ask you about other details:
> 1) You use options like --process-exits and other used for routing, but
> your
> style doesn't set any of the access attributes like mkgmap:car,
> mkgmap:foot etc
> which are needed to get proper routing info in the map.
> I guess you don't care about routing?
> 
> 2) Your cmd file contains the option --pois-to-areas-placement=tagelist 
> I think this is a very old typo which overrides a good default:
> pois-to-areas-placement=entrance=main;entrance=yes;building=entrance
> 
> Please check the docu about the meaning:
> http://www.mkgmap.org.uk/doc/options
> 
> Gerd
> 
> Date: Wed, 7 Jan 2015 16:57:41 +1000
> From: 

> steve.sgalowski@

> To: 

> mkgmap-dev at .org

> Subject: Re: [mkgmap-dev] Problem in splitter (Africa)
> 
> gerd , have tried some of your ideas no not all worked for me however have
> worked out and are combining my poi strings am re checking now with the
> new poi test file you have just uploaded to mkgmap server 
> stephen 
> 
> On Tue, Jan 6, 2015 at 7:54 PM, Gerd Petermann <

> gpetermann_muenchen@

> > wrote:
> 
> 
> 
> Hi Stephen,
> 
> yes, you should have received an answer a now.
> 
> Gerd
> 
> Date: Tue, 6 Jan 2015 19:15:51 +1000
> From: 

> steve.sgalowski@

> To: 

> mkgmap-dev at .org

> Subject: Re: [mkgmap-dev] Problem in splitter (Africa)
> 
> gerd P 
> did you get my direct e-mail to you sir 
> stephen 
> 
> On Tue, Jan 6, 2015 at 10:18 AM, GerdP <

> gpetermann_muenchen@

> > wrote:
> Hi Stephen,
> 
> 
> 
> the log shows no problems. Why do you think that max-nodes=400000 doesn't
> 
> work?
> 
> Do you see an error message in mkgmap?
> 
> If yes, please provide your style files so that I can reproduce the
> problem.
> 
> Maybe your style still adds one POI for each point of each highway?
> 
> 
> 
> Gerd
> 
> 
> 
> 
> 
> steve sgalowski wrote
> 
>> canada splitter log file
> 
>> as expected ,  looks like i was correct
> 
>> the size of the split has to be smaller
> 
>> stephen
> 
>>
> 
>> On Tue, Jan 6, 2015 at 7:25 AM, Carlos Dávila <
> 
> 
> 
>> cdavilam@
> 
> 
> 
>> >
> 
>> wrote:
> 
>>
> 
>>> Final file size depends on the amount of data in the input, not on the
> 
>>> value of max-nodes. If you need a final img smaller than a given size
>>> you
> 
>>> have to reduce the area covered by the input file or reduce the number
>>> of
> 
>>> osm elements from the input that go into the map playing with your style
> 
>>> files.
> 
>>>
> 
>>> El 05/01/15 a las 22:07, Steve Sgalowski escribió:
> 
>>>
> 
>>>> gerd and carlos
> 
>>>> i am now running the splitter log file setup on my canada map
> 
>>>> and see what it does , the end result on this map = 6.8 gb img file
> 
>>>> wonder why some country can exceed and others not
> 
>>>> stephen
> 
>>>>
> 
>>>>
> 
>>>> On Tue, Jan 6, 2015 at 7:00 AM, Carlos Dávila <
> 
> 
> 
>> cdavilam@
> 
> 
> 
>> >> 
> <mailto:
> 
> 
> 
>>
>  cdavilam@
> 
> 
> 
>> >> wrote:
> 
>>>>
> 
>>>>     Not sure what you mean. If you split a given country in a higher
> 
>>>>     number of tiles (lower max-nodes) final size will be the same or
> 
>>>>     slightly bigger, as there are more duplicated info due to overlap.
> 
>>>>     Or you are loosing some information in the process to reduce final
> 
>>>>     file size.
> 
>>>>
> 
>>>>     El 05/01/15 a las 21:34, Steve Sgalowski escribió:
> 
>>>>
> 
>>>>         in some of the countries i do , if i dont make the node count
> 
>>>>         small , the map size exceedds , size limit of 3 gb
> 
>>>>         then unshure how , canada has done this ok
> 
>>>>
> 
>>>>         stephen
> 
>>>>
> 
>>>>
> 
>>>>         On Tue, Jan 6, 2015 at 5:24 AM, Gerd Petermann
> 
>>>>         <
> 
> 
> 
>> gpetermann_muenchen@
> 
> 
> 
>> >>         
> <mailto:
> 
> 
> 
>>
>  gpetermann_muenchen@
> 
> 
> 
>> >
> 
>>>>         
> <mailto:
> 
> 
> 
>>
>  gpetermann_muenchen@
> 
> 
> 
>> >>         
> <mailto:
> 
> 
> 
>>
>  gpetermann_muenchen@
> 
> 
> 
>> >>> wrote:
> 
>>>>
> 
>>>>             Hi all,
> 
>>>>
> 
>>>>             I wonder what splitter should do in this case:
> 
>>>>
> 
>>>>             Stephen uses paramter --max-nodes=80000
> 
>>>>             and splitter reports
> 
>>>>             "Highest node count in a single grid element is 557,084"
> 
>>>>
> 
>>>>             It is obvious that at least one tile will have much more
> 
>>>>         than the
> 
>>>>             requested 80.000 nodes,
> 
>>>>             on the other hand, the file africa.osm.pbf contains large
> 
>>>>         nearly
> 
>>>>             empty areas,
> 
>>>>             and that makes it very difficult to find a good split.
> 
>>>>             The current version r416 fails because it doesn't accept
> 
>>>> tiles
> 
>>>>             with less than 5% of
> 
>>>>             the max-nodes value, so it searches for a solution where
> 
>>>> every
> 
>>>>             tile has at least 4000 nodes,
> 
>>>>             and that might not exist.
> 
>>>>
> 
>>>>             I see these options:
> 
>>>>             1) splitter can continue trying to split the data,
>>>> accepting
> 
>>>>             almost empty output files
> 
>>>>             (e.g. some with < 5 nodes and very high aspect ratios like
> 
>>>> 32)
> 
>>>>             2) if that fails,  splitter can set the max-nodes value to
> 
>>>>         557,084
> 
>>>>             and try again
> 
>>>>             3) or stop with an error message that tells the user that
> 
>>>>         it is
> 
>>>>             not possible
> 
>>>>             to split with the used resolution
> 
>>>>             4) or restart using a higher resolution  (15 would be
> 
>>>> required
> 
>>>>             here instead of 13),
> 
>>>>
> 
>>>>             @Stephen
> 
>>>>             What reason do you have to use such a small max-nodes
>>>> value?
> 
>>>>             Would it be ok for you to use a higher one?
> 
>>>>
> 
>>>>             Gerd
> 
>>>>
> 
>>>>
> 
>>> _______________________________________________
> 
>>> mkgmap-dev mailing list
> 
>>>
> 
> 
> 
>> mkgmap-dev at .org
> 
> 
> 
>>> http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
> 
>>>
> 
>>
> 
>> _______________________________________________
> 
>> mkgmap-dev mailing list
> 
> 
> 
>> mkgmap-dev at .org
> 
> 
> 
>> http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
> 
>>
> 
>> splitter.log (356K)
> 
>> &lt;http://gis.19327.n5.nabble.com/attachment/5829156/0/splitter.log&gt;
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 
> --
> 
> View this message in context:
> http://gis.19327.n5.nabble.com/Problem-in-splitter-Africa-tp5829130p5829158.html
> 
> Sent from the Mkgmap Development mailing list archive at Nabble.com.
> 
> _______________________________________________
> 
> mkgmap-dev mailing list

> mkgmap-dev at .org

> 
> http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
> 
> 
> _______________________________________________
> mkgmap-dev mailing list

> mkgmap-dev at .org

> http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev 		 	   		  
> 
> _______________________________________________
> 
> mkgmap-dev mailing list

> mkgmap-dev at .org

> 
> http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
> 
> 
> 
> _______________________________________________
> mkgmap-dev mailing list

> mkgmap-dev at .org

> http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev 		 	   		  
> 
> _______________________________________________
> 
> mkgmap-dev mailing list

> mkgmap-dev at .org

> 
> http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
> 
> 
> 
> _______________________________________________
> mkgmap-dev mailing list

> mkgmap-dev at .org

> http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev 		 	   		  
> _______________________________________________
> mkgmap-dev mailing list

> mkgmap-dev at .org

> http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev





--
View this message in context: http://gis.19327.n5.nabble.com/Problem-in-splitter-Africa-tp5829130p5829354.html
Sent from the Mkgmap Development mailing list archive at Nabble.com.


More information about the mkgmap-dev mailing list