[MITgcm-support] Bug in rdmds.m when loading multiple record files

[Yuan Lian]

Hi Jean-Michel,

The default MITgcm outputs work fine with the current rdmds, with or 
without multiple records. The problem will occur only when I define some 
variables that may confuse rdmds, e.g., the example I provided.

The reason why I suggest to make change to the current rdmds is that 
sometimes people may opt to use multiple record files to reduce the number 
of outputs. For instance, there is a hard limit of number of files on NASA 
Pleiades. To avoid constantly reaching the limit I had to use 
multiple-record format for all outputs, including the time-averaged 
quantities such as Ttave, though I need to hack the code in timeave to do so.

Best,
Yuan


Sent with AquaMail for Android
http://www.aqua-mail.com


On May 14, 2013 9:55:01 AM Jean-Michel Campin <jmc at ocean.mit.edu> wrote:
> Hi Yuan,
>
> Thanks for the report. Will take a look at this.
>
> I also think that as soon as a user specify one or a list of (or all, if NaN)
> iteration number to read-in, then this problem does now show up
> (since it become T.[10-digits-iteration-number]*.meta ),
> so that it makes it less likely to happen. We have few output file with
> multiple records corresponding to different fields (e.g., "thSIce_tave"),
> but they still have iteration number in their file-name; or may-be I missed 
> something. Did you turn-on special option to generate these files ?
>
> Cheers,
> Jean-Michel
>
> On Mon, May 13, 2013 at 04:03:41PM -0700, Yuan Lian wrote:
> > Hi all,
> > I noticed a bug in rdmds.m under MITgcm/utils/matlab, which only
> > affects the multiple-record files.
> > The problem comes from line 149:
> > allfiles=dir( sprintf('%s*.meta',fname) );
> > For example, when one wants to load A=rdmds('T', 'rec', 1), it
> > actually will look up all variables starting with
> > 'T', including 'Ttave' etc. This is particularly problematic if
> > there are user defined output variables, e.g.,
> > if one wants to use variable name 'TSK' (surface temperature in this
> > example). Because the function
> > "localrdmds" in rdmds is called within the j=1, size(allfiles,1)
> > loop (line 157), the desired variable "T"
> > will be partially or entirely overwritten by "TSK" depending on the
> > size of TSK.
> > Moving "*" right before "meta" can easily address the issue without
> > affecting other functionality:
> > allfiles=dir( sprintf('%s.*meta',fname) );
> > Hope this helps.
> > Best,
> > Yuan
> >
>
> > _______________________________________________
> > MITgcm-support mailing list
> > MITgcm-support at mitgcm.org
> > http://mitgcm.org/mailman/listinfo/mitgcm-support
>
>
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support