Parameter definition

[mukoki]

I found several problems in how projections are defined, read and written in esri registry (and probably also in epsg). 1 - I don't how where the esri projection file comes from, but if we take a projection like 102676 (NAD 1983 StatePlane Iowa South FIPS 1402 Feet), it has a to_meter=0.3048006096012192, but a x_0=500000.0000000002 (which is in meters according to http://www.expertgps.com/spcs/Iowa-South-FIPS-1402-NAD83.asp). According to ESRI specification (http://support.esri.com/technical-article/000001897), parameter units should be constistant with projection unit. It seems to be the same with OGC specification.

2 - In CRSHelper, x_0 parameter is read as if it was always meter, which is consistent with the current esri file (see above), but not with the specification. It may break interoperability (see how 102676 is defined in http://spatialreference.org/ref/esri/102676/html/).

3 - reading a projection in the proj file and writing it back to wkt does not preserve parameters because WKT writer is rounding all parameters to a precision of 10-11 (org.cts.op.projection.Projection). It should not change a value like 40.66666666666667 to 40.66666666667.

[mukoki]

Just checked in proj4 repository,

• esri registry is very old, so I think the CTS's one is uptodate and the unit mismatch between x_0 and unit is in every file I checked.
• epsg registry is 4 months old only on proj4 repository, but for epsg:3426, it still has x_0=500000.00001016 +units=us-ft which does not seem consistent with epsg definition (http://www.epsg-registry.org/) or maybe I missed something in the proj4 format.

[mukoki]

OK, got it

proj4 specifies for the internal format : "False Easting/Northing? Virtually all coordinate systems allow for the presence of a false easting (+x_0) and northing (+y_0). Note that these values are always expressed in meters even if the coordinate system is some other units. Some coordinate systems (such as UTM) have implicit false easting and northing values.

ESRI documentation on the other side "Projection parameters use the units of measure from both the projected and geographic coordinate systems. Linear parameters, like false easting and false northing, have the same units as the linear unit of measure listed at the end of the projected coordinate system string."

OGC " is an optional attribute for reasons of backward compatibility. It is recommended that it is included explicitly in WKT strings. If omitted then:

Requirements: If is omitted from the

So it seems the only problem is in the WKT writer which writes the projection parameter in meters ignoring that the CS unit is feet. Instead, it should embed in the specific unit (either meter or feet) in the parameter tag when the parameter unit is not the same as the CS unit.

[mukoki]

Here is an prj defining WaukeeStreets.txt

a projection in feet

[ebocher]

Thanks I will look on it asap

[mukoki]

I'll split this issue into 3 parts

• using wrong units in WKT writer for projection parameters when CRS unit is not meter : PR "issue_77" should solve this issue. It is not very clean because it tries solves an ambiguity in WKT1 format (see https://www.geomatys.com/wordpress/blog/introduction-a-la-version-2-du-format-well-known-text-wkt)
• all WKT methods and prj package are relative to WKT 1 format. CTS should be able to handle WKT 2 format, but it would need more work -> new issue
• WKT writer round parameters to 1E-11 (eleventh decimal) -> new issue

[ebocher]

We can closed this issue and try to fix the other ones as @mukoki describes.