Library Reference

If you are programming with Madagascar, please use the utilities described below instead of creating your own equivalents. The RSF library is written in C, and all facilities offered by it are available directly to the C API via inclusion of header files and linking with the library. Other APIs only have access to a limited set of functions from the library. The ultimate guide to what the libraries provide is the file RSFSRC/filt/lib/SConstruct, which contains the instructions for building the RSF library in C and the interfaces to it for the various APIs.

C API

Macros

The following useful macros are accessible through the C and C++ APIs:

Functions
Name	Value	Meaning
SF_ABS(a)	(a) >= 0  ? (a) : (-(a))	Absolute value
SF_MAX(a,b)	(a) < (b) ? (b) : (a)	Maximum of two values
SF_MIN(a,b)	(a) < (b) ? (a) : (b)	Minimum of two values
SF_SIG(a)	(a) >= 0  ? 1  : -1	Sign function

General constants
Name	Value	Meaning
SF_EOL	'\014'	End-of-line character
SF_EOT	'\004'	End-of-tape (?) character
SF_EPS	FLT_EPSILON	The smallest X of type float such that 1.0 + X != 1.0. From float.h
SF_HUGE	FLT_MAX	The largest finite representable value of type float. From float.h
SF_MAX_DIM	9	Maximum number of dimensions in a RSF file
SF_PI	3.141592653589793	The number pi

SEG-Y-related constants
Name	Value	Meaning
SF_SEGY_FORMAT	24	Byte position in SEG-Y binary reel header, where info on encoding of data samples is kept (format parameter in sfsegyread)
SF_SEGY_NS	20	Byte position in SEG-Y binary reel header, where info on nr of samples in one trace is kept (ns parameter in sfsegyread)
SF_SEGY_DT	16	Byte position in SEG-Y binary reel header, where info on time sampling is kept
SF_EBCBYTES	3200	Bytes in the SEG-Y EBCDIC reel header
SF_BNYBYTES	400	Bytes in the SEG-Y binary reel header
SF_HDRBYTES	240	Bytes in each individual SEG-Y trace header
SF_NKEYS	71	Number of mandated header fields
SF_BHKEYS	27	Number of mandated binary fields

These macros were found with: <bash>grep '# define SF_' $RSFROOT/include/rsf.h</bash>

Functions

• sf_input (needs to be followed by a call to sf_fileclose at the end of the program)
• sf_fileclose

Function wrappers

These programs make the respective function callable from the command line, so that it can be used directly with files/other programs through "metaprograms" (shell/Python scripts). This allows the user to get the best of all worlds -- fast development in Python, all of Madagascar's visualization tools for debugging, and the speed of code written in C. This allows fast prototyping, followed by writing a production version in C calling the functions used.

Library function	Program
sf_filedims	filedims
sf_leftsize	leftsize
sf_quantile	quantile

C++ API

Macros

Same as for the C API.

Fortran 90 API

The interface is accessed by including a use rsf statement in your program. The source code for the rsf module can be viewed in the repository. A (G) means that the entity was featured in the Guide to API.

Constants

For use with gettype and settype:

• sf_char=1
• sf_complex=4
• sf_float=3 (G)
• sf_int=2
• sf_short=5
• sf_uchar=0

For use with sf_seek:

• sf_seek_set=0
• sf_seek_cur=1
• sf_seek_end=2

Data types

All procedures in Madagascar's F90 API act upon or take as input one of the following data types:

• axa: Holds some info about a hypercube axis (integer::n; real::o,d)
• file: Simply an interface to the sf_file type described in the C library.

Functions

Data types are indicated in italics and intent in bold italics. In the Wrapper for column, procedure names with no other description belong to the C library. The G? heading means "Demo-ed in the Guide to the F90 API?".

Name	Arguments	Result	What it does	Wrapper for	G?
axisname	integer in i, optional character in mystring	character mystringi	Given a string mystring and a positive number i, this function will return the concatenation of the string and the number converted to string (mystringi). In practice, mystring will probably be "n", "o", "d", "label" or "unit", in preparation for writing to a header file. Default value for mystring is "n".	Verbose string operations in F90	N
dimension	file in tag, integer out dims	integer n	Returns number of dimensions of already-open file handle tag in n, and the dimension values in dims	sf_filedims	N
filesize	file in tag, optional integer in n	integer nr_elements	Computes number of n-dimensional subcubes in file hypercube. When n is absent, default is zero and the number of elements in file.	sf_filesize, sf_leftsize.	Y
gettype	file in tag	integer type	Reads the binary data type from the already-open file handle tag. The type constant should be then compared in the user-written code with one of the six type constants provided by the F90 interface	sf_gettype	Y
rsf_input	optional character in filename	file tag	Opens filename for reading and returns file handle tag. The default is filename="in", which corresponds to the standard input stream	sf_input	Y
rsf_output	optional character in filename	file tag	Opens filename for writing and returns file handle tag. The default is filename="out", which corresponds to the standard output stream	sf_input	Y

Subroutines

Because subroutine names are preceded by a call statement, it is easy to find occurrences of them being used by typing: <bash>grep "call subroutine_name" {filt,user}/*/*.f90 user/*/*/*.f90</bash> in RSFSRC.

Below, data types are indicated in italics and intent in bold italics. The type multi means that multiple data types can be used (subroutine is overloaded). In the Wrapper for column, procedure names with no other description belong to the C library. The G? heading means "Demo-ed in the Guide to the F90 API?".

Name	Arguments	What it does	Wrapper for	G?
from_either	file in tag, character in name, multi out value, optional multi in default	This subroutine looks for the real or integer variable name first in the history file, then in the parameter table. If it it is not found and default is present, it assigns it to value, otherwise it terminates the program.	sf_histint, sf_getint	N
from_par	optional file in tag, character in name, multi out value, optional multi in default	If already-opened file handle tag is supplied, then it reads into value a parameter designated by name from a history file. Else, it reads it from the parameter table. If the parameter is not found and a default is present, then that is used, else ends with failure. It can read from both history files and parameter tables the following types: integer, integer array, real. Strings can be read only from the history file, and not from the parameter table. Real arrays, logicals and logical arrays can be read only from the parameter table.	sf_histint, sf_getint, sf_histints, sf_getints, sf_histfloat, sf_getfloat, sf_histstring, sf_getfloats, sf_getbool, sf_getbools	Y
iaxa	file in tag, axa out axisname, integer in axisnr	Read axis with number axisnr as axis axisname from the already-open file header indicated by tag	Multiple calls to F90 API's from_par	N
oaxa	file in tag, axa in axisname, integer in axisnr	Write axis axisname (i.e. its n, o, d) as axis with number axisnr to the already-open file header indicated by tag	Multiple calls to F90 API's from_par	N
raxa	axa in axisname	unformatted write to I/O unit nr. 0 (usually stdout, but system-dependent) of the contents of the axisname structure (n, o, d)	F90 write statement	N
rsf_read	file in tag, multi out array, optional integer in n	If n is specified, the real or complex array must be 1-D and n elements will be read into it from an already-opened file specified by tag. If n is not specified, reads enough to fill the array, and the array can be 1-D, 2-D, 3-D, 4-D or 5-D.	sf_floatread, sf_complexread	Y
rsf_write	file in tag, multi in array, optional integer in n	If n is specified, real or complex array must be 1-D and n elements will be written to an already-opened file specified by tag. If n is not specified, writes to file the entire contents of the array, and the array can be 1-D, 2-D, 3-D, 4-D or 5-D	sf_floatwrite, sf_complexwrite	Y
settype	file in tag, integer in file_type	Write a type for the binary data to the already-open file header indicated by tag. Data type constants are also provided by the F90 API.	sf_settype	N
to_par	file in tag, character in name, multi in value	Writes name=value to already-opened file specified by tag. Value can be integer, integer array, real or character.	sf_putint, sf_putints, sf_putfloat, sf_putstring	N

Functions from the C library

Functions from the C library for which an interface has been defined at the end of RSFSRC/api/f90/fortran.c can be called directly from Madagascar F90 programs. The programmer must know, however, that an argument type mismatch will result in a silent failure, with no informative errors. There may be other sources of problems as well. Caution therefore must be advised when calling these functions. The reason for these issues is that if a FORTRAN90 routine needs to pass a scalar parameter to a C routine, there is no standard way to ensure that a value is passed rather than an address. The success rate of calling C from Fortran is compiler-dependent and special types may have to be defined for portability. If you do have to call a C function from F90, follow the examples of the functions below. Use the C library reference to find argument and output types. A C-to-F90 data type dictionary follows:

C	F90
RSFFILE	type(file)
INT	integer
OFFSETT	integer(kind=OFFKIND)

Void-output functions callable as subroutines

These can be treated as subroutines, and accessed with a call statement. Follow

• sf_error (G)
• sf_init (G) (Can it also be called in subroutines if we want to isolate I/O operations there?)
• sf_seek

Non-void-output functions

It is a good idea to declare in your program the function output type and use the keyword external, i.e.:

real, external :: sf_dosomething

Alternately, you may define a F90 interface block for the procedure(s).

Several examples can be found in RSFSRC/filt/lib/rsf.f90, but they are not enumerated here because F90 wrappers exist for them and they should be used instead of the direct calls to the C library functions.

Matlab API

Data types are indicated in italics and intent in bold italics. In the Wrapper for column, procedure names with no other description belong to the C library. The G? heading means: "Demo-ed in the Guide to the Matlab API?"

Name	Arguments	What it does	Wrapper for	G?
rsf_create	Arg 1: string (row vector) in new_file_name; Arg 2: string (row vector) in old_file_name, or double (column vector) in new_file_cube_dims	Writes to disk RSF header of native_float hypercube of desired dimensions. Takes as input the name of the new file and either the name of another file with a cube of the same size, or a column with cube dimensions	sf_fileflush, or sf_putint and sf_setformat	Y
rsf_dim	Arguments	What it does	sf_filedims	Y
rsf_par	Arguments	What it does	Wrapper for	N
rsf_read	Arguments	What it does	Wrapper for	Y
rsf_write	Arguments	What it does	Wrapper for	Y

Python API

To use the python api, you need to import the rsf package via:

import rsf

The rsf package has the following members:

Overview of the Python RSF API

Class	Use	Constructor	Example	Available Methods
File
Filter
Movie
Input	Provides a basic way to read rsf files.	file = rsf.Input(name)	input = rsf.Input(name); n1 = input.int('n1')	read(data), data is a numpy array to be read into.
				close()
				float(key, default=None), key is a string to be extracted, e.g. 'n1'.
				int(key, default=None), key is a string to be extracted.
				ints(key, num, default=None), key is a string, num is the number of ints.
				put(key, val), key is a string to set the value for, val is a numerical value (int or float).
				... and more, type help(rsf.Input) on a python interactive session for more information.
Output	Provides a way to output numpy arrays to rsf files.	file = rsf.Output(name)	output=rsf.Output(name); output.write(data)
OverUnder
Overlay
Par	Provides a method for accessing rsf parameters passed from the command line.	par = rsf.Par()	par = rsf.Par(); velname = par.string('vel')
SideBySide
Temp
Vplot	View vplot graphics files from within python.	plot = rsf.Vplot(name)	plot = rsf.Vplot(name); plot.show()
Vppen