[mkgmap-dev] Options overhaul

Tue Mar 29 16:48:08 BST 2011

On 29.03.2011 17:25, Carlos Dávila wrote:
> El 29/03/11 16:50, Steve Ratcliffe escribió:
>> The aim is not just to remove options, but is much more about
>> improving how it works in the default case without options, or to
>> remove the need for various options altogether that only ever existed
>> to work around some bug that is fixed or could be fixed very easily.
>>
>> The problem with --remove-short-arcs is that it is not optional.
>> If it is not given then routing will break and you get streams of
>> 'error' messages that are of no help to anyone, since the map data is
>> perfectly fine. To create a correct routing network mkgmap has to deal
>> with short arcs in the network.
>>
>> The --ignore-osm-bounds option was only ever a work around for mkgmap
>> not dealing with files containing several bounds as produced by josm.
>> This option is not harmless; if used with several tiles produced by
>> splitter it will destroy inter-tile routing.  And yet there are several
>> places on the web that demonstrate exactly that problem [1]
>>
>> Another example: --no-sorted-roads, this is documented as making
>> mkgmap run faster, but as far as I know it never did (there was some
>> completely different reason for why mkgmap was running slow at the
>> time this option was introduced). The downside of the option is that
>> it produces a broken map where any kind of searching for roads is
>> likely to fail. Again it is not optional; the roads have to be
>> sorted, and even if it did take longer, that is just the price that
>> would have to be paid to create a working map.
>>
>> A quick search shows that people are actually using this option
>> and so producing broken maps.
>>
>> ..Steve
>>
>> [1] Eg, just to pick the first one I found (there are many more)
>> http://acrosscanadatrails.posterous.com/making-routable-mapsource-maps
> --road-name-pois was also introduced as a workaround before we have
> address search working and so candidate to be deprecated, but I still
> find it useful as it speeds up searches. For example, if I want to find
> a street called Avenida de la Libertad via address search I have to type
> the complete name (after selecting country and city) whereas via spell
> menu typing Libertad is enough to show a list where I can select the
> desired one. Also, if I'm not sure whether it is called "Avenida de la
> Libertad", "Avenida la Libertad" or "Avenida Libertad", using spell menu
> is much faster. Would it be possible to get partial searches working in
> address search?
> _______________________________________________
>
And is another candidate that really should be sent into the desert. I 
get at least 5 people a week complaining search for POI with my maps is 
broken, because they use another map produced with --road-name-pois 
which completly havocs the POI search on older generation GPS. I'm 
really pissed of about the road-name-pois command because it produces 
maps that may even crash GPS and gives the image of mkgmap producing 
faulty maps -- or said differently -- because of this option in many 
forums I read comments as take care when using OSM maps on your GPS, as 
they may break it (no other mkgmap bug currently is able to crash your 
GPS, other than for example not using --remove-short-arcs when --route 
is given, which may also crash your GPS)

Try out a city navigator map to sea what is possible with address 
search. In principle you only need to enter/exit the housenumber field, 
then entering a streetname (with autocompletion for name which does not 
exist on POI search) and results will be listed. There is no need to 
enter country/city except to filter down results (valid for 
Mapsource/Basecamp/Vista HCx/Nuvi 255W)