Input/Output
Input/Output
Input/output (i/o) can be more a lot more flexible than
just reading typed input from the terminal window
and printing it back out to a screen.
Fortran allows for multiple file streams.
Fortran allows multiple representations of the data
for i/o,
Fortran allows multiple approaches to the sequencing
of i/o.
Multiple File Streams
A keyword nearly universal to all Fortran i/o
statements is the Logical Unit
write(11,*)u
write(12,*)v
integer :: lun=3
read(lun,*) n
! u written to file associated with logical unit 11
! v written to file associated with logical unit 12
! logicial unit can be a variable
Default filename associated with logical unit lun is
fort.lun (fort.11, fort.12 fort.3). Compilers may vary!
Some More Definitions
File - a collection of data
Data is organized into records, which may
be formatted (character representation),
unformatted (machine binary
representation), or denote an end of file.
The Open Statement
The Open statement associates a logical unit with a
specific file:
OPEN([UNIT=]<integer >,&
FILE=<filename >, IOSTAT=<status_tag >, &
STATUS=<status >, ACCESS=<method >,&
FORM=<format >,&
ACTION=<mode >, RECL=<int-expr >)
Examples:
OPEN(unit=10,file=input.u,form=formatted)
OPEN(21,file=output.dat,form=unformatted,status=OLD)
The Open Statement (cont)
More on open keywords:
IOSTAT: a returned integer variable, zero for successful execution;
other values for various errors.
STATUS: character string - UNKNOWN (default); OLD; NEW;
REPLACE; SCRATCH
ACCESS: SEQUENTIAL (default) or DIRECT
FORM: FORMATTED or UNFORMATTED
POSITION: ASIS (default) or REWIND or APPEND
RECL: record length for direct access i/o
ACTION: actions one can take with the file - READWRITE (default);
READ; or WRITE
One can open an already connect file to change its
properties
The Close Statement
The Close statement terminates the connection of a
file to a logical unit.
Close([UNIT=]<integer >,&
IOSTAT=<status_tag >, &
STATUS=<status >)
IOSTAT: integer returned containing theerror status of the call,
zero if no errors
STATUS: what to do with the closed file - KEEP (default) or
DELETE
The Write Statement
The Write statement does output generally:
WRITE([UNIT=]<integer >,&
[FMT=]<format >, IOSTAT=<status_tag >, &
END=<label >, ERR=<label >, &
ADVANCE=<advance_mode >, REC=<record_num >)
END and ERR are obsolete, avoid!
Examples:
WRITE(11,(5e15.8))t(:)
WRITE(IOSTAT=status_int,UNIT=lun, ADVANCE=NO)t
WRITE(ERR=909,UNIT=11)t
.
.
909 CONTINUE
The Read Statement
The Read statement does input:
OPEN([UNIT=]<integer >,&
[FMT=]<format >, IOSTAT=<status_tag >, &
END=<label >, ERR=<label >, &
ADVANCE=<advance_mode >, REC=<record_num >)
Examples:
READ(11,(5e15.8))t(:)
READ(IOSTAT=status_int,UNIT=lun, ADVANCE=NO)t
READ(ERR=909,END=910,UNIT=11)t
.
.
909 CONTINUE
.
.
910 CONTINUE
Formatting
The Format specifier is used in read, write and print
statements.
* - default, or list-directed formatting (space or comma delimited)
f (floating point) for printing of reals
syntax: fw.d
w = total number of positions
d = number of places after the decimal point
the decimal point occupies a position, as does a minus sign
e (exponential) for large or small real numbers - ew.d
d = number of digits in mantissa
a (alphanumeric) for character strings
i (integer) for integer - can use iw.d format, where the d will pad in front of
the value with zeroes
Examples:
exp_format1.f90, exp_format2.f90, exp_format3.f90
Unformatted I/O
When the file is opened with
form=unformatted the binary will be
read/written in the machine
representation. Use no format specifier!
Warning! Different machines may have
different representations - big_endian vs.
little_endian; the latter is generally found
on PC chips.
Sequential vs. Direct Access
In sequential access the end of record is
marked in the file.
As name implies, each read/write proceeds
to next record - exception when
ADVANCE=NO used.
Can move file position with Position
statements
Direct Access
Must open file with ACCESS=DIRECT and
specify a record length (recl) (generally in
bytes)
You go directly where you wish in the file by
specifying the record number (rec=n) in the
READ/WRITE
Multiple jobs/processes can access the file
without interference
Other Useful Statements
The Inquire statement can get information about a file.
You may inquire by unit, or by filename:
INQUIRE([UNIT=]<integer >,&
EXIST=<logical >, IOSTAT=<integer >, &
NAME=<character >, OPENED=<logical >)
INQUIRE([FILE=]<filename >,&
EXIST=<logical >, IOSTAT=<integer >, &
NUMBER=<integer >, OPENED=<logical >) ! plus many more available
! arguments
Position statements:
REWIND lun;
REWIND (UNIT=lun, IOSTAT=status_int)
BACKSPACE lun; BACKSPACE (UNIT=lun, IOSTAT=status_int)
ENDFILE lun;
ENDFILE(UNIT=lun, IOSTAT=status_int)
Other Useful Statements
Namelist i/o, a type of formatted i/o (deprecated):
logical :: dopbp
integer :: ijtlen
NAMELIST /pbplist/ dopbp,ijtlen
open(unit=2,file=namel.pbp,form=formatted)
read(2,pbplist)
>cat namel.pbp
&pbplist
dopbp=.true.
IJTLEN=4
&END
Internal files: unit is a program variable rather than a
file, no open statement used.
character (len=4) :: year
write(unit=cyear,fmt=(i4.4)) 1989
I/O Libraries
Typically, with standard fortran i/o statements,
when someone sends you a file, he must also send you a
readme about the contents (which variables,
dimensions, format, etc.) or some code kernel for
reading.
It sure would be nice if the data in the files were s elfdescribing with the use of meta-data.
I/O libraries are publicly available that can do this:
NetCDF, HDF.
NetCDF
Something of a standard for climate/meteorogical
data - http://www.unidata.ucar.edu/soft ware/
netcdf/
Includes command line utilities to inspect the files
(ncdump)
Many graphics packages can read it (IDL)
NCO (http://nco.sourceforge.net/) is a set of utilities
to manipulate netcdf files
Fortran subroutine calls are used to read/write/
inquire about the data.
NetCDF Philosophy
NetCDF (network Common Data Form) is a set of interfaces for arrayoriented data access and a freely-distributed collection of data access
libraries for C, Fortran, C++, Java, and other languages. The netCDF
libraries support a machine-independent format for representing
scientific data. Together, the interfaces, libraries, and format support
the creation, access, and sharing of scientific data.
NetCDF data is:
Self-Describing. A netCDF file includes information about the data
it contains.
Portable. A netCDF file can be accessed by computers with
different ways of storing integers, characters, and floating-point numbers.
Direct-access. A small subset of a large dataset may be accessed
efficiently, without first reading through all the preceding data.
Appendable. Data may be appended to a properly structured
netCDF file without copying the dataset or redefining its structure.
Sharable. One writer and multiple readers may simultaneously
access the same netCDF file.
Archivable. Access to all earlier forms of netCDF data will be
supported by current and future versions of the software.
NetCDF examples
include "netcdf.inc"
status = nf_open(
&
infile.nc", nf_nowrite, ncidin)
status = nf_inq_ndims( ncidin, ndims)
status = nf_inq_nvars( ncidin, nvars)
do n = 1,ndims
status = nf_inq_dim(ncidin, n, dimname, dimlen)
enddo
do n = 1,nvars
status = nf_inq_var(ncidin, n, varname, vartype, vardims, &
vardimids, varnatts )
do k = 1,varnatts
status = nf_inq_attname(ncidin, n, k, attname)
enddo
if(vartype.eq.nf_float)
&
status = nf_get_var_real(ncidin, n, float_1din )
enddo
