[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: GRIB1 quasi/thin grids




Good news, the Grib1 quasi/thin grid work is done for the NCEP and the
ECMWF grids. Now the challenge is how to implement it, since it involved
an API change to the index.  Data offset was changed to the GDS offset in
the grib file. I made the new grib1 data reading code to be able to use
the old index data offset as well as a new index GDS offset. of course,
if the old index data offset is given and it's a Quasi grid, then the grid
doesn't get expanded. But for the motherlode case, in 7 days all the indexes will be the new index type. The older grib1 data reading code will not be able to use the new index offset of the GDS. so the change has to be a code change first, then the indexes can be updated later.
   Another problem occurs in
the nj22 code that checks the index versions before using the index. Maybe
i could change the check to allow index version 3.0 and above. the
motherlode indexes are currently version 3.0 and the new index version is
4.0.

If the nj22 change is made then the new code could be installed on the lead machine without reindexing the whole repository. last time, i reindexed the lead machine, it took over 8 days and some of the new data was lost since the cpu was tied up according to Tom.


the new code is not ci but i plan on doing it later today unless there is
some objections.

RObb...



On Thu, 15 Feb 2007, Robb Kambic wrote:

the fix to make the GRIB1 quasi/thin grids work required an API change in the grib1 index file. instead of seeking to the data section offset, i have to seek to the GDS offset to get the thin grid info when a getData call is made. remember the GDS is before the data section offset and there is no sure way to back up to the GDS section. since i didn't want to reindex all of motherlode again and lead repository, i added code to decide if the offset was a GDS offset or a data offset. the code is not full proof, should work in the high 90's though. after all the indexes have been redone then the code will work 100% and i can delete the extra code. What do you think...

i'm still doing test on motherlode before ci the code. ===============================================================================
Robb Kambic                                Unidata Program Center
Software Engineer III Univ. Corp for Atmospheric Research
address@hidden             WWW: http://www.unidata.ucar.edu/
===============================================================================


===============================================================================
Robb Kambic                                Unidata Program Center
Software Engineer III Univ. Corp for Atmospheric Research
address@hidden             WWW: http://www.unidata.ucar.edu/
===============================================================================


===============================================================================
Robb Kambic                                Unidata Program Center
Software Engineer III                      Univ. Corp for Atmospheric Research
address@hidden             WWW: http://www.unidata.ucar.edu/
===============================================================================