drww / newhall Goto Github PK

Occurs on datasets 169013 and 20080, these will be added to the unit tests 
after this has been completed.

When model produces a results object, the input ntd[] array is full of nulls 
for some reason.  This has a negative impact on everything else as a result.

It would be helpful to bring a year column over from the input file to the csv 
output table.  Similarly, the total PET column could be added to the csv 
output, next to the SWB column.

What steps will reproduce the problem?
1. BIO5 calc from northeastern weather stations--Central Park, Boston, and 
Philadelphia
2. Values exceed 360 days; randomly occur 1 or 2 per 100 yrs record
3. Occurs in the original BASIC code as well; same data run in the basic gives 
the same result; inherent in the origianl model

What is the expected output? What do you see instead?
Values >360; normally 512 to 600, which isn't feasible


What version of the product are you using? On what operating system?
1.5.1 on Windows 7

Please provide any additional information below.

Looks like it is doubling the possible value

Might need to manually escape the character into it's hexadecimal or decimal 
ASCII value.

The <srcunitsys> tag should not be altered in any way.  Whatever it is coming 
out of an XML file, save it in the resulting XML file.  Do preserve current 
behavior for working with legacy data.

Check these stations:

28468 AZ
41614 CA
262780 NV

Soil temperature calculations seem strange.

Change conversion behavior with dataset, do not convert source, preserve unit 
system.  Alter XML import to match this behavior.

The <srcunitsys> tag needs to be a child of <metadata>.

In any report locations in the application, replace the word "degree(s)" with 
the degree symbol.

Verify that parsing and exporting works after the <nettype> and <srcunitsys> 
tags have changed the structure of the XML file.

Psudocode has been generated from Newhall source code.  A Java implementation 
of this psudocode should be generated.  For the time being, being able to 
generate the exact same results as the DOS version is paramount.

JUnit tests should be implemented for the main simulation body, using the 
results of several DOS Newhall runs.  Lower resolution floating point fields 
may prove slightly problematic, so be aware.

The end result should generate everything that the FLX output file provides 
from Newhall, including the report following the simulation.

Newhall needs a GUI to make it user-friendly.  Goals should be as follows:

1. User can open a dataset.
2. User can repeatedly run model over dataset.
3. All data presented by BASIC model should be present in results.
4. Unit conversion should be available and immediately applied.
5. User should be able to trigger an export of the results.
6. Application should terminate through "X" button and through File>Exit.
7. Some sort of About dialog should be present with minimal information about 
version.

The results tab needs to have precipitation placed under PET calcs, and the 
deltas, along with their sum for the year, also listed in that table.

Add the ability to run hundreds of files at once using the same model 
assumptions for each.  Should take in a folder of datasets, and export to a 
different folder of results.

Summarizer tool will be separate from the main Newhall application, but may be 
integrated back in at a later date, potentially as part of an added batch mode.

Input: Directory of Newhall XML files, only ones with <output> tags.
Output: CSV file matching the attached file format.

Need to add <nettype> tag inside <stninfo>, at the very top.  This tag contains 
data of type String.

Once this has been generated, code generation can begin.  A secondary effort 
later is going to be required to make what has been translated into something 
more meaningful.  The BASIC source is full of tiny variable names (for speed, 
it is claimed) which are highly ambiguous.

Write test classes for the main simulation class of the Newhall model.  
Writing tests for other regions is secondary but also a good idea.  A minimum 
of three test cases based on Newhall simulation runs should prove sufficient.