As well as two bug fixes (now properly deallocating memory and properly reading
the time zone).

The SYNTH plugin has gained much improved configurability and abilities. A new method has been added to Config

It is now possible to define how to generate meteo parameters in the
SYNTH plugin, per station and per parameter. Currently two methods are
implemented (constant and step functions), more will come.

A new getValuesRegex() method has been implemented in Config. It might
replace the getValues() method as it is more powerful but this will need
to be carefully validated...

Thanks Simon!

If no splitting model is provided, it will default to the new "NONE" model that
does nothing. This makes it much more logical when using this generator to
convert PSUM_S / PSUM_L into PSUM / PSUM_PH for example.

This is not the case anymore: now non-standard parameters will be created by the
data generators if they don't exist and if possible given a value. Other than
that, some migration to c++11.

It is better to create a projection context instead of using the default one, to allow the functions to be safely used in parallel applications. Thus far, this solved random seg faults in these functions with Alpine3D.

A new option has been added to the format specifiers, YY.

In such a case, the transformation defined by CSV_UNITS would be applied
first and afterwards, the CSV_UNITS_OFFSET / MULTIPLIER. This allows for
more complex transformations for some edge cases (like 1/8th of a kg/m2
kind of units).

See https://github.com/wmo-im/CF-NetCDF/issues/31 for the discussion about
station IDs in the context of the WMO

This comes handy for non-conforming files although a data creator must be
configured in the [dataEditing] section otherwise I expect something to blow up
later on (the coordinates are assumed to be available).

Small trick in libncpp::readVariableMetadata() in order to look for time units
in the "description" field if nothing has been found in the "units" field.

In some cases, the sampling rate extracted by smet_range was too large for the chosen format specifier

The link to DATA_QA_LOGS is more visible and both options are now in the same
section. Please beware that you will need to update your ini files!!

Keys such as "STATION2", "STATION15" are considered indexed keys. A previous
commit attempted to properly sort such keys so key 15 comes *after* key 2. This
was done by extracting the index and sorting the keys based on the numerical
value of the index. But depending on the matching pattern that has been given,
we might attempt to sort keys such as "TA::GENERATOR1", "RH::GENERATOR1"... In
this case, both keys were considered to have the same index "1" and thus the
last one would overwrite the previous one. Now, this is properly handled by also
storing the "key root", here "TA::GENERATOR" and "RH::GENERATOR" so different
roots with the same index will not overwrite each other any more (while still
sorting keys that share the same root in numerical order).

Since with Thomas' commit we should have major changes, it makes sense to have a
less ambitious release in between.

Implementing new PROJ API

See merge request !16

Updating FindPROJ.cmake to stick to the old PROJ4 interface for as long as it is available, and switch to PROJ otherwise (version 8.0.0 or higher of libproj). This because we use a function that was only introduced in version 8.0.0.

Now the user only has to enable PROJ and cmake will find out if either
proj4 or proj is installed on the system, then compile and link with the
proper version of the library.

Using c++11 regex, if all the matching keys are build such as {some
prefix}{some integral number} then they will be sorted by natural
numerical order (ie 10 comes after 9 and before 11). If not, purely
alphabetical order is used (so STATION10_a comes before STATION2_a).

The code between getValues() and getKeys() will not be unified.

More c++11, more consts, improved readability and fixed a linking error
on some platforms

Now it is possible to set Meteo1dInterpolation::WINDOW_SIZE to zero, this
internally sets ENABLE_RESAMPLING to false.

Fixed the writing of varying number of parameters in SMET

See merge request !15

Handling varying number of parameters over a large time range has always
been tricky, as it was necessary to figure out the exact list of
parameters before allocating the memory as well as before writing the
data. The memory allocation part was done in a previous commit, but the
SMET plugin did not manage to reliably write such data.

This is now fixed, knowing that the parameters are now writen in
alphabetic order.

ERA5 request: code usage guide

See merge request !12

Feature/ci without manual zipping

See merge request !13