Reads unformatted sequential access Fortran binary files.
function fbinrecread ( path  : string, rec_num  : integer, rec_dims [*] : integer or long, rec_type  : string ) return_val [rec_dims] : rec_type
Full or relative pathname to binary file.rec_num
Record number to read from the file beginning at 0. Please see caveat below.rec_dims
A singly dimensioned array of integer values that describe how to shape the data, or -1 if the size of the record is unknown.
As of version 6.0.0, rec_dims can be of type int or long.rec_type
String name of the data type of the record.
As specified by the rec_dims, rec_type arguments.
fbinrecread reads the rec_num record of the file path and shapes it according to the dimension sizes in parameter rec_dims. The data type of the record is specified by the parameter rec_type. If the size and dimensionality of the record are unknown the value -1 can be used for parameter rec_dims. In this case fbinrecread will read in the entire record as a singly dimensioned array.
Note: As with any binary file, the "endian-ness" of the data on the file and that of the current system must agree. The "ReadByteOrder" option in the setfileoption procedure can be used to force the "endian" type in a file. This allows big-endian files to be read on a little-endian machine and vice versa. The isbigendian function can be used to determine the endian-ness of your current system.
This function was updated in version 6.3.0 to improve the efficiency of reading Fortran binary records in sequential order, via a new "KeepOpen" option that you set via setfileoption. When set to True, the Fortran file is kept open and the file pointer is left at the beginning of the record following the last one read, allowing the user to read subsequent records faster. When finished reading the file, the user must call fbinrecread once with this option set to False in order to close the file.
This function was updated in version 4.3.0 in order to allow reads into files bigger than 2 GB. This change allows you to access records that occur past the 2 GB point in the file, but not to read into variables that would be larger than 2 GB.
Caveat: we've run across binary files that were written with 8-byte record markers. This function may expect them to be 4-bytes; it just depends on what system NCL was compiled on.
This happens if the Fortran program that created the binary file was compiled with a version of "g77" or "gfortran" that sets the marker size to whatever "off_t" is on that system.
In V6.1.1 and later, you can change this to 8 bytes using the "RecordMarkerSize" option in the setfileoption procedure.
As a work-around for older versions of NCL, you can recompile the code that created the binary file with the option "-frecord-marker=4" to force a 4-byte record, and run the code again to regenerate the file.
It is on our list to create a new option in setfileoption that will allow you to specify the record marker size that the file was written with.
The below is a Fortran (f77) code fragment:
integer a(5) real x(100), y(399), z(128,64) open (11,file="example01",form="unformatted") write (11) a ! 1st record [rec_num=0] write (11) x ! 2nd record [rec_num=1] write (11) y ! 3rd record [rec_num=2] write (11) z ! 4th record [rec_num=3]
Below is the NCL code to read the above. Note that the shape appears to be different for the z variable. In Fortran, the leftmost dimension varies fastest while in NCL [C], the rightmost dimension varies fastest. Thus, the data record will match the fastest varying dimensions correctly.
fili = "example01" ; input file a = fbinrecread (fili, 0, 5, "integer") ; 1st record is zero x = fbinrecread (fili, 1, 100, "float") ; 2nd one y = fbinrecread (fili, 2, 399, "float") ; 3rd record is two z = fbinrecread (fili, 3, (/ 64,128 /), "float") ; 4th record is three ; sample of reading *all* the float variable in one record. ; data will be a 1D array of length (100+399+128*64) data = fbinrecread (fili, 1, -1, "float") ; read x, y, z as 1D