Re: [ferret_users] how to loop through many files?

[Ansley Manke <ansley.b.manke@xxxxxxxx>]

Hi all,
A third way, if these are files containing time steps for a variable or set of variables, is to make a multi-file, or descriptor data set. This lists the files and information about their time axes, and once you've got it, you can refer to all your data as one data set. You'll have all the time steps available without having to do any Ferret commands to add time steps when you open each file. Take a look at the documentation - see "descriptor file" in the Users Guide. In particular there are some tools people have contributed for creating a descriptor file; look up "descriptor file, tools for creating" in the Users Guide index.

Ansley

Jaison Kurian wrote:

Hi Zibiao,
            Two methods to loop through input files are give below. Pick
the one, which fits best for your purpose. Few points about the usage of
REPEAT command is given to the last part.

 1. Reading input filenames from an ASCII file.
    -------------------------------------------
      This method is best suited, if your filenames doesn't follow a
    common syntax or have patterns which is difficult to construct
    logically. What all you have to do is, in Linux

       [user@machine]$ ls 200*.nc > filenames.txt

       [user@machine]$ wc -l filenames.txt

    Now, say you have 12 input files (the output from "wc -l"). In
    Ferret, use the following steps:

       define axis/x=1:12:1  xfile
       define grid/x=xfile   gfile
       FILE/grid=gfile/FORMAT=DELIMITED/var="filenames"   filenames.txt

       REPEAT/RANGE=1:12:1/NAME=nf (         ;\
            define symbol ff = `nf`          ;\
            set data `filenames[i=($ff)]`    ;\
            ...............................  ;\
            cancel data `filenames[i=($ff)]` ;\
       )

   Between the FILE command and "cancel data", you can add all the lines
   for doing the required stuff with the datasets. Always remember that
   the loaded NetCDF dataset will be d=2 (d=1 being the filenames.txt). 

2. Construct the input filenames in Ferret
   ---------------------------------------
     This method is best suited, if your filenames follow a systematic
   naming syntax, which can be constructed logically. Say your input
   filenames are like 20050115.nc, 20050215.nc,......20071215.nc : that
   is --> yyyymm15.nc format. Now, in Ferret :

       let year_start = 2005
       let year_end   = 2006
       let mon_start  = 5
       let mon_end    = 10
       
       REPEAT/RANGE=`year_start`:`year_end`:1/NAME=yr  (           ;\
          define symbol yy = `yr`                                  ;\
          let m1 = IF `($yy) EQ year_start` THEN mon_start ELSE 1  ;\
          let m2 = IF `($yy) EQ year_end`   THEN mon_end   ELSE 12 ;\
          REPEAT/RANGE=`m1`:`m2`:1/NAME=mo (                       ;\
             define symbol mm = `mo`                                 ;\
             let m0    = IF `($mm) LT 10` THEN "0($mm)" ELSE "($mm)" ;\
             let fname = "($yy)" + m0 + "15.nc"                      ;\
             set data `fname`                                        ;\
             cancel data `fname`                                     ;\ 
          ) ;\
       )

   after the "set data" and "cancel data", you can add the lines to do
   required stuff.   

   A similar trick can be found in the following mail :    
   http://www.pmel.noaa.gov/maillists/tmap/ferret_users/fu_2004/msg00393.html

With both 1 and 2 above, you can use he "GO" file option effectively. Write
a separate script for doing calculation/making plots with these individual
datasets (let us all it as some_calculation.jnl), and call it from the repeat
loop in 1 or 2 above, as follows :

           set data .......;\
               GO some_calculation arg1, arg2 .... ;\
           cancel data  ....;\ 

NOTE : Don't forget to have a "cancel data" command in the same loop, where 
         you opened it.....otherwise you will run into problems....


Important Points Regarding REPEAT LOOP
--------------------------------------
 -  REPEAT Loops can be on abstract axes I,J,K and L or on an independent 
       option "RANGE" (available only with Ferret versions 5.6 or later). 
       If you are not working on independent coordinate axis, "RANGE" option
       is the best.

 -  All the commands within the REPEAT loop is considered as a single line,
       separated by ";" (and with a "\" if you intend to continue on next line).
       There is a length limitation and hence you cannot go beyond a certain
       length. Hence, if you want to do some detailed stuff, use the GO file
       option mentioned above.

 -  Blank/commented lines are NOT ALLOWED within the REPEAT loop. If you want
       to have them, do it with a preceeding ";\", as follows :
             
             let something= 1*2  ;\ 
             ;\
             ;\  ! here goes my comment within a repeat loop
             ;\
             let others = 1/2  ;\ 
  

If you have any questions, please let me know.

Regards,

Jaison

On Tue, 10 Jul 2007, Zibiao Zhang wrote:

  

Hi Ferret users,

I need to loop through many files, which all have the same data structure, I
know there get to be a way to easily loop through the file names, but I
could not find the answer through user's guild and archives. I want if
anybody here can help. My file name is as following:

20050115.nc
20050215.nc
20050305.nc
.
.
.
20071215.nc

It should be very easy, anybody can help?

Steven



    

-- 

~>   ~>   ~>   ~>   ~>   ~>   ~>   ~>   ~>   ~>   ~> 

Ansley Manke, NOAA/PMEL   ansley.b.manke@xxxxxxxx
7600 Sand Point Way NE, Seattle WA
Phone 206-526-6246,  FAX 206-526-6744