Hello all
The implementation of JSR-363 (Unit of Measurement API, the JSR-275
successor) is still under way on the JDK8 and JDK7 branches, but will
hopefully be finished soon. A reference implementation exists outside
Apache, but providing our own implementation in Apache SIS allow us to:
* Associate EPSG code with units.
* Support GML (un)marshalling for Sensor Web (not yet implemented).
* Make unit parsing and formatting compatible with geospatial data
formats (e.g. NetCDF has its own syntax).
* Provide pre-defined constants for units that would not be of special
interest outside the geospatial domain (e.g. "US survey foot",
"grad", etc.).
* Design the system in a way that facilitate integration with the
coordinate operations implemented by sis-referencing module.
The implementation takes in account the fact that most unit conversion
factors are defined in base 10 (e.g. one inch is exactly 2.54 cm by
definition) or as ratios (e.g. one US survey foot is defined as exactly
1200/3937 metres). Those numbers have no exact representation in IEEE
754 'double' type.
The tests are temporarily disabled until JSR-363 implementation is
completed. After completion and verification that all tests pass, the
proposal is to deploy a GeoAPI 3.0.1 release candidate on a temporary
Maven repository and use it for trunk. After we verified that all tests
pass on trunk with that release candidate, I would submit GeoAPI 3.0.1
for a formal release at OGC.
Is there any comment about this approach or things that should be done
differently?
Martin
