Hi Al,
What version are you running? There were a few updates to USE/ORDER in the last several years. The last item in the release notes for V7.6 lists this bug fix:
- USE/ORDER= did not reorder the grids when the dimensionality of the grids for variables isn't the same in the file, e.g. when there are 1-D variables and 2-D variables. Now, the grids of the variables with the largest number of dimensions are permuted first, determining the directions of the axes of other variables.
-Ansley
On 3/15/2021 10:20 AM, Albert Hermann - NOAA Affiliate wrote:
(just re-posting this with a more appropriate subject line)Hi ferreters,
I have been experiencing some ferret problems lately, regarding the use of "use/order" on netcdf files. Here's what I get for the file with ncdump:
-------------------------------------------------yes? sp ncdump -h ensall/tfunc_aggregate_flipped_roms.nc
netcdf tfunc_aggregate_flipped_roms {
dimensions:
XAX1_10 = 10 ;
ZMO = 12 ;
TLONG = 510 ;
ENSEMBLE = 13 ;
EAX = 13 ;
STRING1_15 = 15 ;
variables:
double XAX1_10(XAX1_10) ;
XAX1_10:point_spacing = "even" ;
XAX1_10:axis = "X" ;
double ZMO(ZMO) ;
ZMO:point_spacing = "even" ;
ZMO:axis = "Z" ;
ZMO:standard_name = "altitude" ;
double TLONG(TLONG) ;
TLONG:point_spacing = "even" ;
TLONG:axis = "T" ;
TLONG:standard_name = "time" ;
double ENSEMBLE(ENSEMBLE) ;
ENSEMBLE:units = "realization" ;
ENSEMBLE:point_spacing = "even" ;
ENSEMBLE:axis = "E" ;
double tfunc(ENSEMBLE, TLONG, ZMO, XAX1_10) ;
tfunc:missing_value = -1.e+34 ;
tfunc:_FillValue = -1.e+34 ;
tfunc:long_name = "IF (FLIPT[D=FLIP] EQ 1) THEN TFUNC_RAW[D=TFUNC_AGGREGATE] ELSE TFUNC_RAW[D=TFUNC_AGGREGATE]*(-1)" ;
double tfunc_scaled(ENSEMBLE, TLONG, ZMO, XAX1_10) ;
tfunc_scaled:missing_value = -1.e+34 ;
tfunc_scaled:_FillValue = -1.e+34 ;
tfunc_scaled:long_name = "IF (FLIPT[D=FLIP] EQ 1) THEN TFUNC_SCALED_RAW[D=TFUNC_AGGREGATE] ELSE TFUNC_SCALED_RAW[D=TFUNC_AGGREGATE]*(-1)" ;
double EAX(EAX) ;
EAX:point_spacing = "even" ;
EAX:axis = "E" ;
char member_tfunc(EAX, STRING1_15) ;
member_tfunc:long_name = "MEMBER_TFUNC_RAW[D=TFUNC_AGGREGATE]" ;
// global attributes:
:history = "PyFerret V7.5 (optimized) 9-Mar-21" ;
:Conventions = "CF-1.6" ;
}
----------------------------------------------
As shown, the time axis is of length 510. A simple read without the "order" option yields this:
-----------------------------------------------yes? can dat/all;use ensall/tfunc_aggregate_flipped_forcing.nc;sh dat
currently SET data sets:
1> ensall/tfunc_aggregate_flipped_forcing.nc (default)
name title I J K L M N
TFUNC IF (FLIPT[D=FLIP] EQ 1) THEN TF 1:10 ... 1:12 1:510 1:8 ...
TFUNC_SCALED
IF (FLIPT[D=FLIP] EQ 1) THEN TF 1:10 ... 1:12 1:510 1:8 ...
MEMBER_TFUNC
MEMBER_TFUNC_RAW[D=TFUNC_AGGREG ... ... ... ... 1:8 ...
----------------------------------------------------------------------------
I get the same thing when I explicitly ask for xzte ordering:
---------------yes? can dat/all;use/order=xzte ensall/tfunc_aggregate_flipped_forcing.nc;sh dat
currently SET data sets:
1> ensall/tfunc_aggregate_flipped_forcing.nc (default)
name title I J K L M N
TFUNC IF (FLIPT[D=FLIP] EQ 1) THEN TF 1:10 ... 1:12 1:510 1:8 ...
TFUNC_SCALED
IF (FLIPT[D=FLIP] EQ 1) THEN TF 1:10 ... 1:12 1:510 1:8 ...
MEMBER_TFUNC
MEMBER_TFUNC_RAW[D=TFUNC_AGGREG 1:8 ... ... ... ... ...
-------------------
Now, for my purposes I wish to reorder the axes when I read it in. In one case I wish to reassign the axes from xzte to zxty. If I "use/order=zxty", ferret scrambles things as follows:
------------------------------------------------------------------------------------yes? can dat/all;use/order=zxty ensall/tfunc_aggregate_flipped_forcing.nc;sh dat
currently SET data sets:
1> ensall/tfunc_aggregate_flipped_forcing.nc (default)
name title I J K L
TFUNC IF (FLIPT[D=FLIP] EQ 1) THEN TF 1:510 1:10 1:8 1:12
TFUNC_SCALED
IF (FLIPT[D=FLIP] EQ 1) THEN TF 1:510 1:10 1:8 1:12
MEMBER_TFUNC
MEMBER_TFUNC_RAW[D=TFUNC_AGGREG ... ... 1:8 ...
-------------------------------------------------------------------------------------
Oddly, I do get what I want if I reverse the request to use/order=ytxz:
----------------------------------------------------------------------------------------yes? can dat/all;use/order=ytxz ensall/tfunc_aggregate_flipped_forcing.nc;sh dat
currently SET data sets:
1> ensall/tfunc_aggregate_flipped_forcing.nc (default)
name title I J K L
TFUNC IF (FLIPT[D=FLIP] EQ 1) THEN TF 1:12 1:8 1:10 1:510
TFUNC_SCALED
IF (FLIPT[D=FLIP] EQ 1) THEN TF 1:12 1:8 1:10 1:510
MEMBER_TFUNC
MEMBER_TFUNC_RAW[D=TFUNC_AGGREG ... 1:8 ... ...
----------------------------------------------------------
In a different case I wish to reassign axes from xzte to xzty. Here again things get scrambled:
-----------------------------------------------yes? can dat/all;use/order=xzty ensall/tfunc_aggregate_flipped_forcing.nc;sh dat
currently SET data sets:
1> ensall/tfunc_aggregate_flipped_forcing.nc (default)
name title I J K L
TFUNC IF (FLIPT[D=FLIP] EQ 1) THEN TF 1:10 1:510 1:8 1:12
TFUNC_SCALED
IF (FLIPT[D=FLIP] EQ 1) THEN TF 1:10 1:510 1:8 1:12
MEMBER_TFUNC
MEMBER_TFUNC_RAW[D=TFUNC_AGGREG 1:8 ... ... ...
-------------------------------------
and the only way to get what I truly want is to use this very non-intuitive arrangement of "use/oder=xtyz":
---------------------------------------------------------yes? can dat/all;use/order=xtyz ensall/tfunc_aggregate_flipped_forcing.nc;sh dat
currently SET data sets:
1> ensall/tfunc_aggregate_flipped_forcing.nc (default)
name title I J K L
TFUNC IF (FLIPT[D=FLIP] EQ 1) THEN TF 1:10 1:8 1:12 1:510
TFUNC_SCALED
IF (FLIPT[D=FLIP] EQ 1) THEN TF 1:10 1:8 1:12 1:510
MEMBER_TFUNC
MEMBER_TFUNC_RAW[D=TFUNC_AGGREG 1:8 ... ... ...
-------------------------------------------------------------
Any ideas as to what is going wrong? Does the /order option not work (or get scrambled) when an ensemble axis is present? Any help appreciated.
Best,Al Hermann
--
albert.j.hermann@xxxxxxxx
Pacific Marine Environmental Laboratory
7600 Sand Point Way NE
Seattle WA 98115
206-526-6495