# PyClaw output¶

PyClaw supports options to output more than just the solution $$q$$. It can provide:

• Output of derived quantities computed from $$q$$; for instance, pressure (not a conserved quantity) could be computed from density and energy.

• Output of scalar functionals, such as the total mass summed over the whole grid.

• Output of gauge values, which are time traces of the solution at a single point.

Derived quantities and functionals are written out at the same times that the solution $$q$$ is written. While these could be computed in postprocessing, it is more efficient to compute them at run-time for large parallel runs.

Gauge output is written at every timestep. In order to get this data without a gauge, one would otherwise have to write the full solution out at every timestep, which might be very slow.

## Outputting derived quantities¶

It is sometimes desirable to output quantities other than those in the vector q. To do so, just add a function p_function to the controller that accepts the state and sets the derived quantities in state.p

>>> def stress(state):
...    state.p[0,:,:] = np.exp(state.q[0,:,:]*state.aux[1,:,:]) - 1.

>>> state.mp = 1
>>> claw.p_function = stress


## Outputting functionals¶

In PyClaw a functional is a scalar quantity computed from $$q$$ that is written to file at each output time. For now, only functionals of the form

$$$F(q) = \int |f(q)| dx dy$$$

are supported. In other words, the functional must be the absolute integral of some function of $$q$$. To enable writing functionals, simply set state.mf to the number of functionals and point the controller to a function that computes $$f(q)$$

>>> def compute_f(state):
...    state.F[0,:,:] = state.q[0,:,:]*state.q[1,:,:]

>>> state.mf = 1


## Using gauges¶

A gauge in PyClaw is a single grid location for which output is written at every time step. This can be very useful in some applications, like comparing with data from tidal gauges (from whence the name is derived) in tsunami modeling. The gauges are managed by the grid object, and a grid at location $$(x,y)$$ may be added simply by calling grid.add_gauges((x,y)). Multiple gauges can be set at once by providing a list of coordinate tuples

>>> state.grid.add_gauges([(x1,y1),(x2,y2),(x3,y3)])


By default, the solution values are written out at each gauge location. To write some other quantity, simply provide a function $$f(q,aux)$$ and point the solver to it

>>> def f(q,aux):
...    return q[1,:,:]/q[0,:,:]

>>> solver.compute_gauge_values = f


## Logging¶

By default, PyClaw prints a message to the screen each time it writes an output file. This message is also writing to the file pyclaw.log in the working directory. There are additional warning or error messages that may be sent to the screen or to file. You can adjust the logger levels in order to turn these messages off or to get more detailed debugging information.

The controller provides one means to managing the logging with the verbosity parameter and is provided as an easy interace to control the console output (that which is shown on screen). Valid values for verbosity are:

 Verbosity Message Level 0 Critical - This effectively silences the logger, since there are no logging messages in PyClaw that correspond to this level. May be useful in an IPython notebook for instance if you want the plots to appear immediately below your code. 1 Error - These are logged by the IO system to indicate that something has gone wrong with either reading or writing a file. 2 Warning - There are no warning level logger messages. 3 Info - Additional IO messages are printed and some minor messages dealing with hitting the end time requested. 4 Debug - If this level is set all logger output is displayed. This includes the above and detailed time step information for every time step (includes CFL, current dt and whether a time step is rejected).

When running on a supercomputer, logging to file can be problematic because the associated I/O can slow down the entire computation (this is true on Shaheen). To turn off all logging (both to screen and to file), you need to change the level of the root logger:

import logging
logger = logging.getLogger('pyclaw')
logger.setLevel(logging.CRITICAL)


Again since we don’t use CRITICAL logger messages in PyClaw, this has the effect of turning the loggers off.