[OSGeo-Discuss] Parsing well-known text

[Pierre Abbat]

On Sunday, 6 September 2020 18:31:31 EDT Bruce Bannerman wrote:
> Hello Pierre,
> 
> It is dangerous to assume what the coordinates in a data set represent. It
> may not have implications for your particular use, but could have serious
> consequences for other uses downstream.

What would be other uses downstream?

> A good approach is to consult the discovery metadata that describes the
> dataset that you are using. It should help you to interpret the data. Any
> good data provider will also provide discovery metadata, typically in ISO
> 19115 format (or derivative).

What's the relationship between WKT and ISO 19115? The point clouds I have 
have a description of the coordinate system in WKT or no description of the 
coordinate system.

> If such metadata is not available, ask the data provider to better define
> their data, in your case, the spatial reference system used. If they can’t,
> or won’t provide this information then I would question the need to use
> that data set and look for an alternative, as the data quality is suspect.

The data provider may be the surveyor himself, who has flown the site with a 
drone.

> It is important that blind assumptions of what an unqualified coordinate in
> a spatial reference system means is not built into software that will be
> used by others without explicitly stating that this is what is being done.

The formats PerfectTIN exports to do not allow including WKT, except maybe 
LandXML (which I haven't looked for WKT in, but I know that one can specify 
the units in) and Carlson TIN (it's proprietary and undocumented, and I don't 
know what's in the header). The PerfectTIN file format does not have a place to 
put WKT, but I can add it.

What I'm planning to do in Wolkenbase is this:
If there's a WKT record, use it for units, and copy it to the output.
If the units in WKT differ from those in Wolkenbase, convert the measurements 
to the unit set in Wolkenbase and edit the WKT.
If there is no WKT, assume the units are as set in Wolkenbase, and add a 
minimal WKT which specifies only the units.
If the point cloud consists of several files (which is quite common), and the 
variable-length records in different files disagree, I'm not sure what to do.

PerfectTIN currently reads variable-length records but ignores them. I'm 
planning to make it use the units in WKT, if any, instead of the unit set in 
PerfectTIN, when converting coordinates when reading a file.

Pierre
-- 
The Black Garden on the Mountain is not on the Black Mountain.