DiagnosticTimeseries Class Reference

A class for storing and writing diagnostic time-series. More...

#include <Timeseries.hh>

Inheritance diagram for DiagnosticTimeseries:

Public Member Functions
	DiagnosticTimeseries (IceGrid *g, string name, string dimension_name)
	DiagnosticTimeseries (MPI_Comm com, PetscMPIInt rank, string name, string dimension_name)
	~DiagnosticTimeseries ()
	Destructor; makes sure that everything is written to a file.
PetscErrorCode	append (double T, double V)
	Adds the (t,v) pair to the interpolation buffer.
PetscErrorCode	interp (double time)
	Use linear interpolation to find the value of a scalar diagnostic quantity at time `T` and store the obtained pair (T, value).
PetscErrorCode	flush ()
	Writes data to a file.
Public Attributes
size_t	buffer_size
string	output_filename
Protected Attributes
size_t	start
deque< double >	t
deque< double >	v

Detailed Description

A class for storing and writing diagnostic time-series.

This version of Timeseries only holds buffer_size entries in memory and writes to a file every time this limit is exceeded.

Here is a usage example:

First, prepare a file for writing:

  char seriesname[] = "ser_delta_T.nc";
  NCTool nc(grid.com, grid.rank);
  nc.open_for_writing(seriesname, true, false);
  nc.close();

Next, create the DiagnosticTimeseries object and set metadata. This will prepare the offsets object to write delta_T(t) time-series to file ser_delta_T.nc, converting from degrees Celsius (internal units) to degrees Kelvin ("glaciological" units). Time will be written in years (i.e. there is no unit conversion there).

  offsets = new DiagnosticTimeseries(g, "delta_T", "t");
  offsets->set_units("Kelvin", "Celsius");
  offsets->set_dimension_units("years", "");
  offsets->buffer_size = 100; // only store 100 entries; default is 10000
  offsets->output_filename = seriesname;
  offsets->set_attr("long_name", "temperature offsets from some value");

Once this is set up, one can add calls like

  offsets->append(t_years, TsOffset);
  offsets->interp(time);

to the code. The first call will store the (t_years, TsOffset). The second call will use linear interpolation to find the value at time years. Note that the first call adds to a buffer but does not yield any output without the second call. Therefore, even if interpolation is not really needed because time==t_years, the call to interp() should still occur.

Finally, the destructor of DiagnosticTimeseries will flush(), which writes out the buffered values:

  delete offsets;

Note that every time you exceed the buffer_size limit, all the entries are written to a file by flush() and removed from memory. One may also explicitly call flush().

Definition at line 122 of file Timeseries.hh.