Re: [Geotools-devel] CRSFactory::createProjectedCRS() usage

[Martin D. <mar...@no...>]

Bryce L Nordgren a =E9crit :
> However, I have more.  Geotools has implemented more MathTransforms tha=
n
> OGC has specified parameters for.  In fact, OGC has specified more
> classification types than they have supplied parameter names for.  Wher=
e do
> the standards for "Mercator_1SP" parameters names come from, for instan=
ce?

I assumes that you are refering to figure 2 and sections 10.1 to 10.6 in=20
OGC 01-009 document.


--- Note for peoples wondering what is OGC 01-009 document ------------

Many classification and parameter names in Geotools come from the legacy=20
OGC 01-009 document. OGC 01-009 was an OGC specification before OGC=20
abandonned their own CRS framework and adopted ISO 19111 instead. So OGC=20
01-009 is a legacy specification, not to be uses anymore for new CRS=20
developments. However, ISO 19111 do not overlaps fully OGC 01-009. There=20
is many areas covered by OGC 01-009 and not covered at all by ISO 19111.=20
So our approach was:

   - Use ISO 19111 fully.

   - Only for areas found in OGC 01-009 with absolutly no equivalent in
     ISO 19111, use the OGC 01-009 stuff. So we used OGC 01-009 for
     filling holes when ISO 19111 said nothing on a particular topic.

One can download OGC 01-009 document there:

     http://www.opengis.org/docs/01-009.pdf

Looking at it, one can see that a lot of GeoAPI and Geotools stuff not=20
in ISO 19111 come from this document, including:

   - Well Know Text (WKT) format
   - MathTransform
   - Factories
   - Many classification names (figure 2)
   - Many parameter names (sections 10.1 to 10.6)


---- end of the note -----------------------------------------------


In addition to OGC 01-009, we used an other source of very valuable=20
references for classification and parameter names: the GeoTIFF=20
projection list:

     http://www.remotesensing.org/geotiff/proj_list/

This list provides both GeoTIFF and OGC classification and parameter=20
names. This web site was actually one of our main sources for=20
classification and parameter names in Geotools. The "Mercator (1SP)"=20
projection in this list provides the "latitude_of_origin" and similar=20
parameters.


> And if someone wants to implement a "Sinusoidal" projection, or anythin=
g
> else off of http://mathworld.wolfram.com, how do they come up
> with...well...even a classification name, much less a parameter name?

If we can't find a standard name, we use ours own. This is the case for=20
example of the "Logarithmic" transform.

However, each name is always associated to an authority (or a "scope" in=20
the context of scoped names). Thats said, the full name of "semi_major"=20
is actually "OGC:semi_major", where "semi_major" is the LocalName (from=20
org.opengis.util package) and "OGC" is the scope (LocalName.getScope()).=20
The mechanism is similar to fully qualified names in Java packages.

So, it should always be possible to know if a particular name is an=20
OGC's name or a GeoTIFF one: just look at the scope. For our own=20
transformation, we use "Geotools:" scope. If we find an official name=20
from OGC later, we will add that "OGC:" scoped name.

This lead me to an other point: each classification/parameter can have=20
an arbitrary number of names. Actually they do have. For example (using=20
again the Mercator_1SP classification), try the following from the=20
command line:

java org.geotools.referencing.operation.DefaultMathTransformFactory=20
Mercator_1SP

Geotools will tells you that it know all of the above names for this=20
transformation:

     OGC:Mercator_1SP
     EPSG:Mercator (1SP)
     EPSG:9804
     GeoTIFF:CT_Mercator
     Geotools:Projection Mercator cylindrique

(where the Geotools name is localized, so its appears in French on my=20
machine).

You can use any of those names. Geotools will recognize "CT_Mercator" as=20
well as "Mercator_1SP".

If you want to be sure that you are looking for a GeoTIFF name and not=20
an OGC name, use the fully qualified name. For example Geotools will=20
accepts this name:

     GeoTIFF:CT_Mercator

but will not recognize this name:

     OGC:CT_Mercator

Of course, it will recognize this name however:

     OGC:Mercator_1SP

You can use DefaultMathTransformFactory in order to experiment the names=20
from the command line.



> Does Geotools implement the OGC standard literally, or does it correct =
for
> "presumed typos" in the OGC standard (e.g., Sect 10.1 "Semi_major" is
> capitalized here and only here).  If Geotools were to implement a
> Geocentric to ellipsoidal transform, would it capitalize or not?

We try to correct presumed typos. If a typo is widely used, we can=20
provided both names (the typo and the correct name) as alias.



> Also: MathTransforms don't maintain authority information.  They aren't
> created with authorityfactories.  The onus for creating a translation
> between (for example) GeoTIFF and the OGC authorities lies on the plugi=
n
> writer (me).  So I'm essentially trying to feel out the waters to see i=
f
> the same physical parameter is always going map to the same name by all
> projection implementations.  If so, how strict are those guarantees?  O=
r
> should I just  expect to need to be alert for nonconformists?  :)

This is true that there is no authority factory for MathTransforms. I=20
don't know yet if we should provide one or not. But anyway, in current=20
Geotools implementation, you shouldn't need to translate parameters from=20
GeoTIFF names to OGC names. Just use the GeoTIFF names and it should=20
work. If you want to be really sure that a GeoTIFF name is not confused=20
with an OGC parameter using the same name for a different purpose, just=20
prepend "GeoTIFF:" in front of every names.

Could it work?

	Martin.