Hello all
The following projection methods have been added. They are only small
variants of existing projections:
* Lambert Conic Conformal (2SP Michigan)
* Lambert Conic Conformal (West Orientated)
The list of current projection methods is available at [1]. This list
will continue to expand next week. Note that they are "projection
*methods*", not "projections" (each method can be used by hundred of
projections). The SIS's Well Known Text (WKT) parser can decode a map
projection only if the corresponding method exists in the [1] list.
I created a JIRA task [2] for listing the projection methods that may
not be implemented in the medium term (i.e. this list does not include
the methods that I hope to port in the next month). This list will be
completed in the next weeks. It may be a proposal for a Google summer of
code, or for any other volunteer.
About the implementation of the Mercator projection: section ยง1.3.3 in
EPSG's Geomatics Guidance Note number 7 part 2 (April 2015) uses a
series expansion where Apache SIS uses an iterative method inherited
from older code. I did a comparison of the two methods: for a planet of
relatively small excentricity like Earth, the series expansion provided
by EPSG is twice faster than the iterative method currently used by SIS,
for the same precision. However as we increase the excentricity, the
fast methods lose accuracy while the current SIS method stay accurate.
For the Earth (excentricity of about 0.082) the errors of the fast
method are less than 0.01 millimetres. But the errors become centimetric
(for a planet of the size of the Earth) before excentricity reaches 0.2,
and increase quickly after excentricity 0.3.
So we have a choice:
* Decide that Apache SIS is only for the Earth, and use the fast method.
* Decide that Apache SIS could be used for other planets than the
Earth, and keep the slower method which is more robust to higher
excentricities.
I would add that the fast method is twice faster when those methods are
tested alone, but those methods are only part of a map projection. It is
not obvious what the performance improvement would be when taking the
whole chain of operations in account, so we are not sure that they are
worth the sacrifice of application to other celestial bodies.
Any opinion on the above?
Martin
[1] http://sis.staging.apache.org/content/CoordinateOperationMethods.html
[2] https://issues.apache.org/jira/browse/SIS-212
