[mkgmap-dev] SEVERE (PolishMapDataSource): error decoding label

Sat Dec 26 13:57:50 GMT 2009

Hi Steve,

Steve Ratcliffe wrote:
> Hi
> 
> On 12/24/2009 06:51 PM, Igor Brejc wrote:
>> I'm trying to generate Garmin maps files from Polish maps generated by
>> GroundTruth. I get the following exception (using the latest version of
>> mkgmap):
>>
>> SEVERE (PolishMapDataSource): error decoding label
>> java.nio.charset.UnmappableCharacterException: Input length = 1
> 
> This is caused by a character in the input that cannot be represented
> in the target character set.
> 
> Instead of throwing an error when this happens I will change the code
> so that it just replaces the character with a '?' character.

I see that you've already committed this change. Thanks, I'll try it and 
let you know.
> 
> There is also the problem with the character set and encoding used to
> write the input .mp file.  We have changed the code in mkgmap a number
> of times to try and match commonly found .mp files.  Currently we read
> .mp files as though they are in utf-8 and then convert into the target
> codepage.  This is common for OSM derived files, since the names are
> in utf-8 to begin with, but I don't know if this is true in general or
> recommened.  Are the labels written by GroundTruth in utf-8, or are
> they already converted into the target codepage (or something else:)?
GroundTruth's MP files are UTF-8 encoded, but "national" non-ASCII 
characters are converted to similar ASCII chars (using an OSM Wiki 
table), so I'm not sure which character would be problematic in this case.

I also set the following polish parameters by default:
CodePage = 1250
LBLcoding = 9

I'll try to experiment with some smaller MP file.

Regards,
Igor