[gdal-dev] S-57 Driver WRONG Geometry

[Bogdan Grama]

I guess is a commodity. The authority has sevencs production software.
So, if it “looks ok” in that software “never mind the others”.
Until Frank implemented the modification IÂ’ve recomposed myself the
geometry. But there is no way other than visual to decide if I should
draw edge by edge or a single linestring since there is no masking
information available. :(

Bogdan GRAMA


On 6/1/2010 4:37 AM, ogr user wrote:
> Sevencs succeed in many other aspects where others fail.
>
> For example, they will actually parse and show older Netherlands chart
> data (I still have a few cells from a year ago), which encoded polygon
> edges using multiple SG2D fields, rather than a single SG2D field with
> multiple values. Not only that, but even once I modified s57 reader to
> support that, those polygons could not be reassembled using the "vertex
> proximity" approach as used by GDAL. Yet Sevencs could read those files
> and show them reasonably well. Still have no idea how they did that - I
> could not come up with a reasonable method, and I tried a few.
>
> You can do a lot by building specific heuristics into a product that
> adjust the way it handles data based on certain guesses. Whether that is
> a good practice I don't know - I certainly don't think so. Ultimately,
> it appears that Netherlands decided to reencode their data - and the
> current data set is properly readable by GDAL.
>
> That may be a moot point if you have a user that absolutely *must use
> that chart* no matter how badly encoded it is, so Sevencs have a good
> practical approach there, I guess. But generating data as closely
> following the standard as possible is better, if you ask me.
>
> --Mike
>
> Bogdan Grama wrote:
>   
>> Dear Mike,
>>
>> I could not agree more. But, if u parse the sample ENC, you will see
>> that for the ROADWY layer:
>>
>>     ---->NAME_RCNM = (10:130,130,130,130,130,130,130,130,130,130)
>>     ---->NAME_RCID = (10:114,103,89,97,99,112,111,108,110,124)
>>     ---->ORNT = (10:1,1,1,1,1,1,1,1,1,1)
>>     ---->USAG = (10:255,255,255,255,255,255,255,255,255,255)
>>     ---->MASK = (10:2,2,2,2,2,2,2,2,2,2)
>>
>> So, no masking :). Now the Question is this: How Sevencs succeeds
>> showing the correct representation, where all others fail?
>>
>>
>>   
>>     
>
>