source: inundation-numpy-branch/pyvolution/wiki/issues.txt @ 3330

OPEN ISSUES:
------------


Try to use properties with getters and setters
E.g. as in
class Contact(object):

    def __init__(self, first_name=None, last_name=None, 
                 display_name=None, email=None):
        self.first_name = first_name
        self.last_name = last_name
        self.display_name = display_name
        self.email = email

    def print_info(self):
        print self.display_name, "<" + self.email + ">"            

    def set_email(self, value):
        if '@' not in value:
            raise Exception("This doesn't look like an email address.")
        self._email = value

    def get_email(self):
        return self._email

    email = property(get_email, set_email)
----------------------------------------------------

Issue:  File function (e.g. boundary) should give an error message (or set a defoult) if underlying netcdf has missing values. Make a unit test.
Maybe the underlying code should detect NaNs.


Issue: finaltime should be renamed 'duration'

Issue (interpolate_sww (new functionality): Give a warning if points
 are outside the mesh.
Importance: Mid
Comment (Ole): If points and triangles are listed as discontinuous e.g. vertex coordinates replicated for each triangle, then the bounding polygon gets mangled and some points appear to be outside.
Who: DSG

Issue (interpolate_sww (new functionality)): Be able to specify at
what times data is needed e.g. input a start, finish and period.
Importance: low

Issue (least_squares): Do a bug report to sci py
Importance: low to don't do
Who: Stephen

Issue (least_squares): Runs out of memory with e.g 6 million points and 
300,000 basis functions.
Importance: Mid 
Suggested Action: Look at blocking strategies

Issue (least_squares): Code to not product D matrix (if it isn't
needed) isn't working.
Importance: Mid - low
Suggested Action: Fix

Issue: Parameters specified in config.py must be tranferred to C code 
and it should be possible to override them when setting up a domain.
Background: N/A
Importance: Mid
Suggested Action: 
          1: Let domain read the config file to provide default values.
          2: Write getters and setters to override them.
          3: Pass pertinent values from domain into the c-extensions.
Status: Largely resolved. action 2 ongoing


Issue: Checkpointing - needs to respect changes in script.
Importance: Low


Issue: Adding unknown quantities cause a KeyError.  Known quantities
are given in the list other_quantities, in shallow_water.  Do we want
unknown quantities attached to domain in the future?  If so, it
shouldn't cause a key error.
Importance: low
Suggested Action: 
Who: OLE


Issue: The water level can not be less than the bed elevation, as an
assert.  
Importance: Medium
Suggested Action: How about having the functionality that water levels are set
to == the bed elevation, if level < bed elevation.  This will make
setting things up easier.  A following suggestion is don't initialise
quantities without values, eg in quantity __init__. 
Status: resolved.  Needs unit test though

Issue: Transmissive boundary condition can cause unphysical momentums if
material is not moving out of the domain.
Importance: Med
Suggested Action: Comment in transmissive boundary code.  look into
sponge boundary.



        
CLOSED ISSUES:
--------------

Issue (least_squares): The current is-a-point-in-a-triangle algorithm
is slow if a point is outside of the mesh.  It will check each
triangle 3 times to see if the point is in that triangle.
Importance: low
Suggested Action: Form a polygon of the mesh boundary (by constructing
the neigbour structure) and throw out points not in the polygon (this
algorithm is there).
Who: DSG
Comment (OLE): I have done this (I think)


Issue: Get the reservoir example working for the 'new' pyvolution
Issue: Mid
Who: DSG
Status: Done  

Issue (least_squares): datapoints outside the mesh are not used in
least squares.
Importance: Mid - low
Suggested Action: Talk to Steve about including them
Status: Talked to Steve, it was dicided not to use them

Issue (least_squares): Do a test that the number of attribute titles and number
of attributes match
Importance: Mid
Who: DSG
Status: closed, this is checked when loading points files now

Issue (Malpasset example):
Pmesh no longer takes care of attributes but still accepts xya with attributes.

Background:
Chris imported xya file into pmesh. It contained x, y, bed elevation
The tsh thus generated did add bed elevation, but didn't produce the proper 
tsh file.

Importance: mid
Suggested actions: Either spit out an error message or generate a proper 
tsh-header.
Status:  The story of this is wrong.  Pmesh does produce a proper tsh file

Issue: Second order limiters may cause numerical problems 
manifesting themselves as QNAN (Quiet NANs_ or INF)
Background: The second order gradient limiter is not robust enough - 
especially with shallow water depths. Heighs and momentum are 
limited independently and high speeds may result.
Importance: Mid
Suggested Action: 
          1: Try to incorporate a weighted averaging process between 
          shallow depths and large depths as in the first order scheme.
          2: Look at limiting height and momentum together, by keeping 
          an eye on the speeds 
          (e.g make speed constant across a volume and recompute momentum)
          3: Catch errors resulting from these instabilities as early 
          as possibly and produce and throw an appropriate exception
          4: Make small example that reveals this bug.
Status: 1 succesfully implemented by Ole and Stephen 12/5/4.
        2 Didn't work but using the weighted average for momentums seemed 
        to work well.
        3 and 4 not done.
        June 2004: With the new limiter (item 1 and 2) 3 and 4 are probably 
        not necessary now.



Issue: Tags for specifying boundary conditions should be arbitrary
Background: Domain has the functionality for using arbitrary keys
when associating boundary objects to boundary values.
Triangle, which pmesh is based on, uses integers only.
It would be nice to have mnemonic names for boundaries throughout.
Importance: Mid
Status: Duncan resolved this in early June.





Issue: Checkpointing - needs to always back one timestep up in 
       case output wasn't written
Background:
       Data at latest checkpoint may be incomplete or even corrupted 
       depending on when and how program was stopped.
Importance: Mid
Suggested Action: Always try second latest checkpoint first.
Status: Done (Ole 14/5/4)
              

                
Issue: interpolated_conserved_quantities in shallow_water.py
Background: Was computed by flux functions and used by bed_slope.
It turns out to be unneccessary as conserved quantities at centroid works just as well.
Importance: Mid
Suggested Action: Get rid of it everywhere
Status: Was done when whole things was refactored in April 2004 by OMN


Issue: Exceptions thrown by Python functions called from C don't get caughts
Background: Have a look at gravity in shallow_water. It calls a C function, gradient, which may throw an exception if arguments are scalars.
However, it doesn't get caught becuase gravity itself is called from within C
Importance: Low
Suggested Action: Have a look at the calling convention from C (David Beasley)
Status: Unresolved - but no longer relevant as we don't call Python from C