A new function is introduced into the IOInterface: writeMeteoData. It expects one vector<vector<MeteoData>> and one vector<vector<StationData>> as well as an optional string (for database info, file name, etc). The relevant io.ini tag is:
METEODEST = GEOTOP

Only GeotopIO has this method fully implemented thus enabling to convert all meteo data captured from any other source to a GEOtop compliant meteo files. The relevant tags are:
METEODESTPATH = /tmp
METEODESTSEQ	= WindS WindDir RelHum AirT SWglobal Iprec

METEODESTPATH determines where the files will be written to and METEODESTSEQ determines the order in which the meteo parameters will be written into the CSV file

Added Filter to average wind speeds and directions called WindAvgFilter, accessible through the io.ini tags: wind_avg;
example:
VW::filter1 = wind_avg
VW::arg1 = soft center 1 60

The averaging is done vector based by computing the east-west and north-south components of the wind (using cos and sin). The vectors are weighted with the corresponding wind speed.

Updated the gsnclient app to return the stations available before any other output. Fixed spelling issue in GEOtopIO.

The java bindings now use the latest 2D interpolations interface. This means that all data is immediately put into a vector of MeteoData and then routed as usual through the Meteo2DInterpolator. The includes have also been cleaned up (only using MeteoIO.h), a few const keywords have been used as well as passing arguments by "&" instead of explicitely by reference.

A better example of DEM usage has been prepared: it uses a 1000m resolution DEM of the whole of Switzerland (that is, with holes in the dem) and valid coordinates are now much easier to guess...

A new slope algorithm has been added (HORN), even if it computes the same slopes as CORRIPIO (except for border cells). This is the algorithm that is used in ArcGIS. Some cleanup has been done in the references (it seems some algorithms have been published several time by different people). The embryo of a real hydrological slope algorithm is coming: it is not nammed CARD any more, but D8 and gives NODATA to the aspect of points that don't have any slope (slope=0). It will have to be completed by adding a sink resolution algorithm.

This is an updated version of the java interface to take into account the recent changes in MeteoIO's API. The Makefile has been improved, it now compiles on Linux.

A problem was found during the intialization of the plugins (as well as A3DIO): the parameters that were passed to initialize the local copies were not always initialized with the raw parameters passed to the constructor but usually with the copy of the previous element. This lead to some copies not to be up to date when calling their constructor (see A3DIO.cc for a clearer view!!).

The write2DGrid members of the plugins now use the features of the Coords object to output the grid in the input coordinate system (as specified in the io.ini file). A copy of the Coords object is made, so that it does not change the given grid. This is fixing issue 3

The plugin's interface (IOInterface) has been slightly modified: readSpecialPoints now returns a vector of Coords. This means that a plugin can provide coordinates as (lat,long) or (easting, northing) or (grid_i, grid_j). All the plugins have been modified accordingly and POINTS and POINTSArray have been removed from Array.h

A few issues have been fixed in Coords: the copyProj method had a bug, the distance method now uses cartesian calculation if both projections are equal, a new getProj() method is now available (but it should be rarely needed). The GridxDObject::gridify method now correctly handles "NULL" projection and the Meteo2DInterpolator makes sure that all projections agree before doing the interpolations.

Finally, the implementation of issue 35 is here... A few bugs left with the Coords class have been fixed (namely, the methods take 1 optional parameter and not more, because for this usage scenario it was way too dangerous and the user (myself in this case!) could call the method with the wrong parameter left empty). New mebers have been added: grid_i and grid_j as well as methods to deal with them. These contain the grid indices matching the position. Since these indices depend on the chosen DEM/grid, they are NOT updated by Coords but by Grid2DObject or Grid3DObject instead. If the user calls Coords::setLatLon() or Coords::setXY(), then these grid parameters are reset to nodata.

A bug in a constructor of StationData has also been fixed (assigning a temporary object to a member of the class). Some bugs/unlucky deletes in the examples have been fixed (this is the price to pay for work too late at night!)

New constants "lat_epsilon" and "lon_epsilon" have been created in IOUtils. They represent the smallest latitude/longitude resolution (ie: two points within lat_epsilon are considered to have the same latitude)

The constructor Coords(ConfigReader) was a bad idea... Convenient, but way too specific. It has been removed. The documentation in the Coords class has been updated.

A saner handling of nodata values by the plugins has been implemented. Now, each plugin should properly get its own internal definition of nodata and return data using IOUtils::nodata to the caller. A new method has been implemented in IOUtils to help cleaning a parameter for nodata.

A check for xmlParseFile (used for compiling the Boschung plugin) was working poorly and has been commented out in configure.ac.

Finally, more cleanup in the ph files has been performed.

The new Coords object is now used by all grids (Grid2DObject, Grid3DObject, DEMObject) as well as by all plugins (but checking proper operation was not possible for GSNIO because of missing libraries on my system...). The documentation has been updated as well as the code examples. A few extra methods have been added when usage has shown the need/convenience of having them (specially for standardazing some processing called by the plugins). A few small bugs have been fixed (possibility of not updating or improperly updating the coordinates in some rare cases).

The proper marshalling for Grid3DObject has been implemented (it was more or less a dummy method until now). One bug has been fixed for Proj4 support. The code of most of the plugins has been cleaned up in order to get rid of the few remaining "using namespace" in header files. 

The class Stationdata now properly uses the new Coords class. All usual dependencies have been fixed (A3DIO.cc) but not (yet) the optional plugins. The Coords class has been modified so that it computes the matching coordinates as soon as possible instead of the previous implementation that was delaying it as much as possible (thus making "const Coords..." impossible).

The includes have been cleaned up (compilation was failing because of multiple "using namespace" in header files) and all "using namespace" have been removed from header files (please don't put any of these back!!). The doxygen comments has been updated.

The MapProj class has been renamed in Coords. It now contains the coordinates (lat,long) and (xcoord, ycoord) and should therefore be used as a storage and conversion class. The necessary changes have been made to the rest of the code so it uses this class, but it still does NOT use it as a storage of coordinates (which will have to be done in the future). This addressed partially issue 35.

The doxygen comments for MapProj have been reorganized. An invalid type has been fixed in DEMObject and a structure updated in LegacyIO.ph.

Bugfix for A3DIO: Now it is no problem anymore if some stations exist in one Meteo2D file (e.g. wspd2005.txt) but not in another (e.g. wspd2006.txt). Unfortunately BufferedIOHandler will try to rebuffer the missing data and fail, which might be time consuming.

New methods have been added/updated in the Arrays (1D, 2D, 3D): getMin, getMax and getMean. They accept an optional argument (either RAW_NODATA or PARSE_NODATA) that tells how to interpret nodata values (as normal values or really as nodata).

The dem usage example code has been improved in order to show more capabilities.

A few static members have been added to MapProj to handle latitude/longitude: a method for converting a decimal latitude or longitude to degrees, minutes, seconds (formatted as d°m's"),  a method for converting a string containing degrees, minutes, seconds to decimal latitude or longitude and a method for parsing a position expressed as latitude/longitude (given in a string) and returning decimal latitude and longitude. This is the implementation of issue 34.

In the MapProj class, a new conversion has been added: UTM to/from WGS84. Simply give "UTM" as the coordinate system and specify in the optional arguments the zone number (for example, 31T).

A (tricky) error has been fixed in the VincentyDistance. The bearing that was chosen was the average between the initial and reverse bearing. That seemd a good idea at the time, but this made reprojecting a local grid computed using Vincenty impossible: the required bearing (starting from the reference point) is the initial bearing, the average one was sending us hundreds of kilometers away from the correct coordinate (when doing (lat,lon) -> local -> (lat,lon)). 

Finally, a few more powN() methods have been added to IOUtils for convenience (yes, the projection business loves these!!)

The Special_PTS and CSpecialPointsArray structures have been renamed in "POINT" and "POINTSArray" (there was no reason to call them "special points"). 

A few new methods have been added to DEMObject: getPointsBetween that returns a vector containing the indices of all points located on the line between two points (given by their real world (x,y) coordinates), horizontalDistance that returns the horizontal distance between two points and terrainDistance that returns the distance between two point accros the terrain as described by the DEM (ie: going up and down the terrain).

A bug was preventing RH interpolations from working (the TA grid was initialized to size 0, and therefore the safety checks were later on stopping the interpolation of RH because the TA grid was of the wrong size). 

A "sanitize" method has been added to DEMObject so that all points that have an elevation but no slope/azimuth/normals/curvature see their elevation set to nodata (so that testing the elevation of a point is enough to guarantee that all information is available). The caller has therefore to decide if he prefers potentially losing a few points to guarantee a more consistent DEM or not...

The parallelization was lagging behind, so this has been fixed: One internal variable was not serialized in DEMObject, there was no serialization at all in Meteo2DInterpolator and some marshalling methods had to be written (in marshall_meteoio). The serialization of Meteo2DInterpolator should be tested however!

The marshalling code has been cleaned up and some missing "std::" added in FilterAlgorithms (doxygen was complaining).

A new method "setDefaultAlgorithm" has been added to the DEMObject so that the caller can specify which slope computation algorithm to use. The default algorithm argument has also been added to the constructors (as optional argument). This means that each constructor MUST call setDefaultAlgorithm.

A new slope computation algorithm has been added: FLEM for (Fleming and Hoffer, 1979). This algorithm is faster, simpler and usually better then the other ones (according to (Hodgson, >1997)).

Finally, some benchmarks have been done comparing the current slope computation speed with the old implementation (which could not handle holes in DEM): the current method is only 5% slower and it can now handle holes!

The DEMObject has been modified so that it can now correctly handle grids with nodata values when computing slope/azimuth/curvature (this is in effect the fix for issue 15). This is "best effort" approach: several strategies are applied if some points are missing in order to still be able to compute as much as possible the parameters for their neighbours (if the latter contain an elevation!). The resulting grids are consistent for slopes (ie: if the slope can be computed, then the azimuth too and the normals too) and the curvature is handled separately (ie: it is possible to find points where ONLY the curvature is nodata).

Several inappropriate behaviors have been fixed for cells next to a nodata point (or next to an edge). This should have fixed issue 7.

The necessary check have been added to libinterpol2D:SimpleDEMWindInterpolate

Added the MeanAvgFilter and updated the MedianAvgFilter. Now the position of the filter window can be defined. The keyword soft has the meaning that the configuration of the window position is not strict (it will adapt if there is not enough data available for the user defined window position). If both soft and the position for the window (left, right, center) are missing "center" is assumed. Usage:

TA::filter1 = mean_avg
TA::arg1   = soft right 10 100
RH::filter1 = median_avg
RH::arg1  = 10 200

Meaning: use mean_avg filter on TA, the window shall be right leaning spanning at least 10 points and 100 minutes. Use median_avg filter on RH, the window shall be centered (strictly).

Oups... My previous commit was wrong, this is the commit that fixes the compilation bugs in libinterpol2D.

A mistake has been found in BoschungIO: the key for precipitation was wrong (we were reading the precipitation type as a precipitation amount, which was quite bad for the simulation!).

A few missing cases of nodata have been identified (thanks Cyril!) and fixed in libinterpol2D

dynamic use of meteoio (ie. dynamically loaded from an application -- for exemple from Java application with the JNI mechanism).
This leads to save the dem file loading time in further interpolation (more than half the time is saved in some cases).

* Tries to make the "dynamic" interpolation call thread-safe in a multithreaded context (ie. concurrent interpolation are called by 
different threads) : 
	-> identifies the Critical Section and the critical resources
	-> use of a 'synchronized' java-like macro to isolate these sections and resources

void interpolate(const MeteoData::Parameters& meteoparam, Grid2DObject& result);

The new Meteo2DInterpolator class parses the ConfigReader object passed to it and stores all possible algorithms for each MeteoData variable (user configured). In the next step the user configured algorithms shall be invoked and a quality estimator chooses the most appropiate of the algorithms to interpolate the 2D grid for a given meteo parameter.

Adjustments needed to be made to the ConfigReader class (function findKeys) and a few comments have been added to MeteoFilter for more clarity of the ConfgReader parsing process.

Add JNI and JNA compatible interfaces to dynamically call
meteoio from JAVA.
At the moment only two simple functions are exposed with some restrictions (only one interpolation by call, not all IO plugin supported [ARCIO, Geotop, GrassIO] ).

The successfull call of MeteoIO from Java has been implemented by Cyril Perot but needed a few adjustments in MeteoIO. This commit makes the said adjustments: a few more "#ifdef _POPC_", a missing ConfgReader.addKey(key, section, value) and some previously private members of Meteo2DInterpolator now exposed. 

The plugin unloading sections have been protected by ifndef _METEOIO_JNI because the current implementation of such Java integration uses JNI and does its own plugin loading/unloading.

In an effort to make ConfigReader more flexible, its interface has been expanded: it is now possible to read several files into one ConfigReader, or to manually add key/value pairs and it is planned that a string containing command line arguments could be given to populate the internal map of ConfigReader.