SPICE Kernel Required Reading

This required reading document is reproduced from the original NAIF document available at https://naif.jpl.nasa.gov/pub/naif/misc/toolkit_docs_N0067/C/req/kernel.html

Note

These required readings documents were translated from documentation for N67 CSPICE. These pages may not be updated as frequently as the CSPICE version, and so may be out of date. Please consult the changelog for more information.

Important

NOTE any functions postfixed by "_" mentioned below are Fortan-SPICE functions unavailable in SpiceyPy as the NAIF does not officially support these with "_c" function wrappers within the CSPICE API. If these functions are necessary for your work please contact the NAIF to request that they be added to the CSPICE API

Abstract

The kernel subsystem loads and unloads kernels, retrieves loaded data, and for text kernels, inserts data into the kernel pool.

Document Outline

This document has five major sections.
  • Introduction to kernels

  • Kernel type identification and kernel naming

  • Binary kernel specifications

  • Text kernel specifications and interfaces, including extra rules for meta-kernels

  • Kernel management

"Introduction to kernels" should be read by anyone new to SPICE or needing a refresher about kernels. "Kernel type identification and kernel naming" contains specifications for kernel architecture and type identification and restrictions and recommendations concerning kernel file naming.

"Binary kernel specifications" points the reader to other SPICE documents for most information on binary kernels.

"Text kernel specifications and interfaces," which includes extra rules for meta-kernels, provides a good deal of technical detail for both producers and consumers (users) of text kernels.

"Kernel management" contains important information about managing and obtaining information about both text and binary kernels.

Appendix A discusses the notion of "competing data."

Appendix B provides definitions of terms used in this document with SPICE-specific meaning.

Appendix C provides a listing of kernel subsystem functions.

Appendix D provides a summary of key text kernel parameter values.

Appendix E provides the revision history of this document.

Introduction to Kernels

Files containing the data used by SPICE are known as kernels (sometimes called "kernel files"). Two kernel architectures exist, referred to as text kernels and binary kernels. Text kernels consist of human readable ASCII text; binary kernels consist of mostly non-ASCII data.

Within each architecture there are several kernel types.

Kernel Types

The SPICE text kernels are:
  • text form of planetary constants (text PCK)

  • leapseconds (LSK)

  • spacecraft clock coefficients (SCLK)

  • instrument geometry (IK)

  • reference frame specifications (FK)

  • meta-kernels (MK)

The SPICE binary kernels are:

  • ephemeris for vehicles, planets, satellites, comets, asteroids (SPK)

  • orientation (attitude) of a spacecraft or other structure (CK)

  • special binary form of planetary constants containing only orientation (binary PCK)

  • shape models or topographic data for extended objects (DSK)

  • mission events (EK)

Text Kernels and the Kernel Pool

Text kernels are used where the amount of data being stored is relatively small, and where easy human readability and revision are important.

Text kernels should contain descriptive information, provided by the kernel producer, describing the sources and intended uses of the kernel data.

Text kernels associate values with variables using a "name = value(s)" form of assignment. The kernel pool is the repository of the information provided in these assignments. Populating the kernel pool occurs in either or both of two ways: by loading text kernels -- by far the most used method -- or by using pool subsystem functions.

Once "name = value(s)" assignments provided in a text kernel have been loaded into the kernel pool the value(s) are said to be associated with the names. You may access these data through kernel pool look-up functions using the names as keys to find the associated values. The kernel pool look-up functions are described in detail a bit later in this document. However, some higher-level and more often used functions also access data loaded into the kernel pool. Two tables in the tutorial named "Summary of Key Points" provide details.

Binary Kernels

Binary kernels store large data sets of primarily non-ASCII data, using either the DAF or DAS format (see the technical reference documents daf.req and das.req for details). For all but EK binary kernels, loading the binary kernel does not cause the subsystem associated with the kernel's type to read the principal kernel data; rather only a small amount of descriptive data are read so the subsystem becomes aware of the existence of the kernel and the nature of the data contained therein. The subsystem physically reads primary binary kernel data only when a data request is made by a kernel reader function.

For EK binary kernels, the descriptive data mentioned above, and some database schema information, are read in at kernel load time. Principal data are read only when an EK query is made by a kernel reader function.

Data from binary kernels do NOT get placed in the kernel pool; the pool is used only for text kernel data.

Binary kernels contain a "comment area" where important descriptive information in ASCII form should be provided by the kernel producer.

On occasion one may be given, or need to make, a "transfer format" file. This is an ASCII-format representation of a binary kernel, used in early versions of CSPICE to port binary kernels between dissimilar computers (e.g. IEEE - Little endian to IEEE - Big endian, or vice-versa). For the most part these transfer format files are no longer needed due to the addition of run-time translation capabilities in the binary kernel readers. But there are some situations when transfer format binary kernels are still needed; refer to the tutorial named "Porting Kernels" for details.

SPICE Kernel Type Identification and Kernel Naming

SPICE Kernel Type Identification

Most SPICE users don't need to know about kernel type identification, but since this aspect of kernels is used later on in this document we have to explain the concept here.

The first 6 to 8 bytes of a SPICE kernel are used for file type identification. In binary and text kernels this identifier consists of two string IDs separated by the "/" character. The first ID, identifying the file architecture of the kernel file ("DAF", "DAS", "KPL"), is always three characters long. The second ID, identifying the file type of the kernel file ("SPK", "PCK", "IK", "SCLK", etc.), is two to four characters long.

In transfer format files this file type identifier consists of a single string ID. See the Convert User's Guide for details.

In binary kernels the kernel type identifier always occupies the first eight bytes. If the combined length of the kernel architecture ID, the "/" character, and the kernel type ID is less than 8 characters, the identifier is padded on the right to eight characters using blanks (e.g. "DAF/SPK ", "DAS/EK "). The correct identifier is written to a binary kernel automatically when the kernel is created by calling the kernel type specific "open new file" function -- spkopn() for SPK files, ckopn() for CK files, etc. If a binary kernel is created by calling an architecture specific "open new file" function -- dafonw_c for DAF files, dasonw() for DAS files, etc., -- it is the caller's responsibility to specify the correct kernel type in the corresponding input argument of these functions to make sure the correct kernel type identifier is written into the kernel.

In text kernels the kernel type identifier occupies the first six to eight characters and is followed by optional trailing blanks and then by the end-of-line terminator character(s), resulting in the identifier appearing on a line by itself. If the combined length of the kernel architecture ID, the "/" character, and the kernel type ID is less than 8 characters, the identifier can, but does not have to be padded on the right to eight characters using blanks (e.g. "KPL/SCLK", "KPL/IK ", etc.). Since most text kernels are created manually using a text editor, it is the responsibility of the person making the kernel to put the correct identifier by itself on the first line of the kernel.

In transfer format files the SPICE kernel type identifier occupies the first six characters of the file and is followed by the expanded name of the format (e.g. "DAFETF NAIF DAF ENCODED TRANSFER FILE"). The correct kernel type identifier is written to a transfer format file automatically when the file is created by the SPICE utility programs TOXFR or SPACIT. See their user guides, toxfr.ug and spacit.ug, for details.

The SPICE kernel type identifiers used in modern SPICE kernels are as follows.

Binary Kernels:

   SPK           DAF/SPK
   CK            DAF/CK
   DSK           DAS/DSK
   PCK           DAF/PCK
   EK            DAS/EK

Text Kernels:

   FK            KPL/FK
   IK            KPL/IK
   LSK           KPL/LSK
   MK            KPL/MK
   PCK           KPL/PCK
   SCLK          KPL/SCLK

Transfer format files:

   DAF           DAFETF
   DAS           DASETF

Some older kernels used an earlier version of the kernel type identifier. In these kernels one would find:

NAIF/DAF
NAIF/DAS

The Toolkit includes the getfat() function to retrieve the kernel file architecture and kernel type encapsulated in the SPICE kernel type identifier. A text kernel not having a kernel type identifier can, in fact, be processed by high-level functions, and by low-level functions other than getfat() that use text kernel data. However, NAIF strongly recommends kernel creators to provide the identifier.

Recommendations on Kernel File Naming

CSPICE places a few restrictions on kernel file names beyond those imposed by your operating system:
  • Kernel file names, including path specifications, must not exceed 255 characters.

  • Use of embedded blanks in kernel file names is not supported by CSPICE. Such names generally will not be recognized when passed as command-line arguments to CSPICE utility programs.

  • Host system "shell variables" or "environment variables" cannot be passed as input arguments to CSPICE functions.

Mission operations teams often include a variety of identifying and user information in kernel names, making them quite long. This practice is probably unavoidable, but kernel producers should be aware that when the mission's SPICE archive is prepared for delivery to the Planetary Data System (PDS), all kernels to be archived must have names consistent with PDS standards, including a limitation to a "36.3" format (1 to 36 alphanumeric characters, followed by the decimal character, followed by 1 to 3 alphanumeric characters) and using only letters, digits and the underscore character. NAIF recommends kernel names use only lower case letters. NAIF further recommends one follows the conventions established for kernel name extensions, shown below.

.bc    binary CK
.bds   binary DSK
.bes   binary Sequence Component EK
.bpc   binary PCK
.bsp   binary SPK
.tf    text FK
.ti    text IK
.tls   text LSK
.tm    text meta-kernel (FURNSH kernel)
.tpc   text PCK
.tsc   text SCLK

Binary Kernel Specifications

Other than the general specifications and recommendations in the section "Kernel type identification and kernel naming" that are applicable to binary kernels, specifications for the various binary kernels are provided in kernel type specific technical reference documents, such as "SPK Required Reading" and "CK Required Reading."

Text Kernel Specifications and Interfaces

The specifications and restrictions discussed below apply to any text kernel. However, the special type of text kernel known as a meta-kernel (sometimes called a "FURNSH kernel") has additional restrictions; these are discussed later in a section on meta-kernels.

Text Kernel Specifications

Often the easiest and best way to create a text kernel is to start with an existing text kernel, editing it to meet your needs. But knowing text kernel rules is still important. Those rules are documented in the remainder of this section.

As the name implies, SPICE text kernels contain printable ASCII text (ASCII code 32-126). Text kernels may not contain non-printing characters, excepting tab (ASCII code 9). However NAIF recommends against use of tabs in text kernels. NAIF also recommends caution be exercised when cutting/pasting text from a formatted document into a text kernel; the text characters displayed in a document may not be in the accepted ASCII range, in which case the text kernel parser will fail when reading those characters.

Assignments in SPICE text kernels have a "name = value(s)" or "name += value(s)" format. We illustrate this format by way of an example using an excerpt from a SPICE text planetary constants kernel (PCK). The format description given below applies to all SPICE text kernels; the specific data names shown in this example apply only to text PCK kernels.

Vectors of values are enclosed in parentheses.

The example begins with a SPICE kernel type identifier and is then filled out with a combination of descriptive information, called comment blocks, and data blocks.

KPL/PCK

Planets first. Each has quadratic expressions for the direction
(RA, Dec) of the north pole and the location and rotation state
of the prime meridian. Planets with satellites (except Pluto)
also have linear expressions for the auxiliary (phase) angles
used in the nutation and libration expressions of their satellites.

\begindata


BODY399_POLE_RA        = (    0.      -0.64061614  -0.00008386  )

BODY399_POLE_DEC       = (  +90.      -0.55675303  +0.00011851  )

BODY399_PM             = (   10.21  +360.98562970  +0.          )

BODY399_LONG_AXIS      = (    0.                                )

BODY3_NUT_PREC_ANGLES  = (  125.045    -1935.53
                            249.390    -3871.06
                            196.694  -475263.
                            176.630  +487269.65
                            358.219   -36000.    )

\begintext


Each satellite has similar quadratic expressions for the pole and

prime meridian. In addition, some satellites have nonzero nutation
and libration amplitudes. (The number of amplitudes matches the
number of auxiliary phase angles of the primary.)

\begindata


BODY301_POLE_RA      = (  270.000   -0.64061614  -0.00008386   )

BODY301_POLE_DEC     = (  +66.534   -0.55675303  +0.00011851   )

BODY301_PM           = (   38.314  +13.1763581    0.           )

BODY301_LONG_AXIS    = (    0.                                 )


BODY301_NUT_PREC_RA  = (  -3.878  -0.120  +0.070  -0.017   0.     )

BODY301_NUT_PREC_DEC = (  +1.543  +0.024  -0.028  +0.007   0.     )

BODY301_NUT_PREC_PM  = (  +3.558  +0.121  -0.064  +0.016  +0.025  )

\begintext

Here we include the radii of the satellites and planets.

\begindata

BODY399_RADII    = (     6378.140    6378.140     6356.755  )
BODY301_RADII    = (     1738.       1738.        1738.     )

\begintext

End of example text kernel. In this example there are several comment blocks providing information about the data. Except for the comments appearing just after the kernel type identifier and before the first data block, all comment blocks are introduced by the control word

\begintext

A comment block may contain any number of comment lines. Once a comment block has begun, no special characters are required to introduce subsequent lines of comments within that block. A comment block is terminated by the control word

\begindata

or by the end of the kernel file. The

\begindata

control word also serves to introduce a block of data that will be stored in the kernel pool. A data block is terminated by the control word

\begintext

or by the end of the kernel file. Each of these control words must appear on a line by itself, and each may be preceded by white space.

Within each data block there are one or more variable assignments. Each variable assignment consists of three components:

  1. A variable name.

  2. An assignment operator. This must be "=" (direct assignment) or "+=" (incremental assignment).

  3. A scalar or vector value.

Variable Name Rules

A variable name can include any printable character except:
  1. " " (space)

  2. "," (comma)

  3. "(" (open parentheses)

  4. ")" (close parentheses)

  5. "=" (equal sign)

  6. TAB character

Variable names must not exceed 32 characters in length. Variable names are case-sensitive. Note that this behavior is different from that of most CSPICE high-level functions, which tend to ignore case in string inputs. Variable names that don't have the expected case will be invisible to CSPICE functions that try to fetch their values. Since high-level CSPICE functions that use kernel variables accept only upper case names, NAIF recommends upper case always be used for variable names.

NAIF recommends you do not use a variable name with "+" as the last character.

Assignment Rules

Direct assignments supersede previous assignments, whereas incremental assignments append the specified values to the set created by previous assignments. For example, the series of assignments
BODY301_NUT_PREC_RA  = -3.878
BODY301_NUT_PREC_RA += -0.120
BODY301_NUT_PREC_RA += +0.070
BODY301_NUT_PREC_RA += -0.017
BODY301_NUT_PREC_RA += 0.

has the same effect as the single assignment

BODY301_NUT_PREC_RA = (  -3.878  -0.120  +0.070  -0.017   0 )

Variable Value Rules

Values may be scalar (a single item) or vectors (two or more items). A value may be a number, a string, or a special form of a date.

Numeric values may be provided in integer or floating point representation, with an optional sign. Engineering notation using an "E" or "D" is allowed. All numeric values, including integers, are stored as double precision numbers. Examples of assignments using valid numeric formats:

BODY399_RADII     = ( 6378.1366     6378.1366     6356.7519   )
BODY399_RADII     = ( 6.3781366D3   6.3781366D3   6.3567519D3 )
BODY399_RADII     = ( 6.3781366d3   6.3781366d3   6.3567519d3 )
BODY399_RADII     = ( 6.3781366E3   6.3781366E3   6.3567519E3 )
BODY399_RADII     = ( 6.3781366e3   6.3781366e3   6.3567519e3 )
BODY399_RADII     = ( 6378          6378          6357        )

String values are supplied by quoting the string using a single quote at each end of the string, for example

DISTANCE_UNITS = 'KILOMETERS'

This quoting convention is independent of the SPICE Toolkit language version being used. All string values, whether part of a scalar or vector assignment, must not exceed 80 characters on a given line. Creating a string value longer than 80 characters is possible through continuation of an assignment over multiple lines; this is described later.

There is no practical limit on the length of a string value other than as mentioned in the section on String Continuation below.

If you need to include a single quote in the string value, use the FORTRAN convention of "doubling" the quote.

MESSAGE = 'You can"t always get what you want.'

Date values may be entered in a wide variety of formats, using two methods. The easiest method is to enter a date as a string, as described above. There are no restrictions on the format of a date string entered as a string, but if you wish to later use that date string in SPICE software the string must conform to SPICE date/time formation rules (see the "Time Required Reading" document for details). A second method for entering dates, unique to text kernels, uses an "@" syntax. Some examples:

CALIBRATION_DATES = ( @31-JAN-1987,
                      @feb/4/1987,
                      @March-7-1987-3:10:39.221 )

Dates entered using the "@" syntax may not contain embedded blanks. Dates entered using the "@" syntax are converted to double precision seconds past the reference epoch J2000 as they are read into the kernel pool.

Note that NO time system specification (e.g. UTC or TDB) is implied by dates using the "@" syntax. Association of a time system with such dates is performed by the software that uses them. For example, in SPICE leapseconds kernels, such dates represent UTC times; in frames kernels, they represent TDB times. You should refer to software user's guides or API documentation to understand the interpretation of these dates for your application.

Vector values, whether of numeric, string or date types, are enclosed in parentheses, and adjacent components are separated by either white space (blank or carriage return, but not TAB) or commas. Multiple components can be placed on a single line. Multiple lines may be used to continue a list of values. Individual numeric, date, and string values may not be split across lines, but a long string may be continued using multiple substrings. See the section "Additional Text Kernel Syntax Rules" below for details.

MISSION_UNITS = ( 'KILOMETERS','SECONDS'
                  'KILOMETERS/SECOND' )

The types of values assigned to a given kernel pool variable must all be the same. If you attempt to make an assignment such as the one shown here:

ERROR_EXAMPLE = ( 1, 2, 'THREE', 4, 'FIVE' )

the kernel pool reader will regard the assignment as erroneous and reject it. |

Additional Text Kernel Syntax Rules

Line Length

All assignments, or portions of an assignment, occurring on a line must not exceed 132 characters, including the assignment operator and any leading or embedded white space.

Blank Lines

Blank lines in data blocks are ignored.

String Continuation

It is possible to treat specified, consecutive elements of a string array as a single "continued" string. String continuation is indicated by placing a user-specified sequence of non-blank characters at the end (excluding trailing blanks) of each string value that is to be concatenated to its successor. The string continuation marker can be any positive number of printing characters that fit in a string value (except not true for meta-kernels).

For example, if the character sequence

//

is used as the continuation marker, the assignment

CONTINUED_STRINGS = ( 'This //  ',
                      'is //  ',
                      'just //',
                      'one long //',
                      'string.',
                      'Here"s a second //',
                      'continued //'
                      'string.'              )

allows the string array elements on the right hand side of the assignment to be treated as the two strings

This is just one long string.
Here's a second continued string.

Everything between the single quotes, including white space and the continuation marker, counts towards the limit of 80 characters in the length of each string element. The SPICE function stpool(), and ONLY that function, provides the capability of retrieving continued strings from the kernel pool. See the discussion below under "Fetching Data from the Kernel Pool" or the header of stpool() for further information.

Maximum Numbers of Variables and Variable Values

All variable values from all text kernels loaded into your program are stored in the kernel pool. There are upper bounds on the total numbers of variables and variable values.

See Appendix D for the numeric values of these limits.

Treatment of Invalid Text Kernels

If during a call to furnsh(), an error is detected in a text kernel, CSPICE will signal an error. By default, a diagnostic message will be displayed to standard output and the program will terminate.

If the SPICE error handling subsystem is in RETURN mode, furnsh() will return control to the calling program. RETURN mode is typically used in interactive programs.

In the latter case, all data loaded from the text kernel prior to discovery of the error will remain loaded.

If, in RETURN mode, an error occurs while a meta-kernel is being loaded, all files listed in that meta-kernel that have already been loaded will remain loaded. Files listed in the meta-kernel later than the file for which the failure occurred will not be loaded.

Note that continuing program operation after a load failure could, due to changes in the availability of competing data, result in performing computations with data that were not planned to be used.

Additional Meta-kernel Specifications

A meta-kernel (also known as a "FURNSH kernel") is a special instance of a text kernel. Its use has been discussed earlier in this document. In addition to the text kernel specifications above, a meta-kernel has the following restrictions.
  • When continuing the value field (a file name) over multiple lines, the continuation marker must be a single "+" character.

  • The maximum length of any file name, including any path specification, is 255 characters.

  • Embedded blanks are not allowed in path or file names.

Text Kernel Interfaces - Fetching Data from the Kernel Pool

Note

For most SPICE users the accessing of text kernel data occurs inside of high-level CSPICE functions, so you may choose to skip the rest of this section. But if you need to work with text kernel variables that are not present in traditional text kernels, and thus are not accessed by high-level SPICE functions, read on.

The values of variables stored in the kernel pool may be retrieved using the functions:

gcpool()

Used to fetch character data from the kernel pool.

gdpool()

Used to fetch double precision data from the kernel pool.

gipool()

Used to fetch integer data from the kernel pool. Within the kernel pool all numeric data are stored as double precision values. This interface is provided as a convenience so that users may insert and retrieve integer data from the kernel pool without having to worry about converting between double precision values and integers.

Non-integer, numeric kernel variable values retrieved by calling gipool() are rounded by gipool to the nearest integer. Kernel creators must ensure that values to be read using gipool() are within the range representable by integers.

stpool()

Used to fetch continued strings from the kernel pool.

See function documentation for specifics on function parameters.

Informational Functions

Four routines are provided for retrieving general information about the contents of the kernel pool.
dtpool()

Returns information about the existence, dimension and type of a specified kernel pool variable.

expool()

Returns information on the existence of a numeric kernel pool variable.

gnpool()

Allows retrieval of names of kernel pool variables that match a string pattern.

szpool()

Returns information about the size of various structures used in the implementation of the kernel pool.

These routines are discussed at length in their respective source code headers and referenced NAIF CSPICE documentation.

Section 5 -- Kernel Management

The kernel subsystem provides functions_c to load and unload SPICE files, known as kernels, and provides other kernel management and information functions. These functions_c are part of the "KEEPER" subsystem.

Loading Kernels

For the SPICE system to use kernels, they must be made known to the system and opened at run time. This activity is called "loading" kernels. SPICE provides a simple interface for this purpose.

The principal kernel loading function is named furnsh() (pronounced "furnish"). A kernel database stores the existence information for any kernel (text or binary) loaded by furnsh(). The subsystem provides a set of functions that enable an application to find the names and attributes of kernels stored in the database.

Early versions of CSPICE loaded kernels using functions specific to each kernel type. Code written for the binary kernels also supported a kernel unload facility. CSPICE continues to support the original kernel loaders and unloaders, but anyone writing new code should use the furnsh() function instead of the kernel-specific functions.

NAIF recommends loading multiple kernels using a "meta-kernel" rather than by executing multiple calls to furnsh(). ("Meta-kernels" are sometimes called "furnsh kernels.") A meta-kernel is a SPICE text kernel that lists the names of the kernels to load. At run time, the user's application supplies the name of the meta-kernel as an input argument to furnsh(). For example, instead of loading kernels using the code fragment:

from spiceypy import *

furnsh("leapseconds.tls")
furnsh("mgs.tsc")
furnsh("generic.bsp")
furnsh("mgs.bc")
furnsh("earth.bpc")
furnsh("mgs.bes")

one may now write

from spiceypy import *

furnsh("kernels.tm")

where the file "kernels.tm" is a SPICE text meta-kernel containing the lines

KPL/MK
\begindata

KERNELS_TO_LOAD = ( 'leapseconds.tls',
                    'mgs.tsc',
                    'generic.bsp',
                    'mgs.bc',
                    'earth.bpc',
                    'mgs.bes'           )

\begintext

This technique has the important advantage of enabling a user to easily change the set of kernels to be loaded without modifying his source code. While far less robust, it is also possible to provide the names of kernels to be loaded as input arguments via a list or other iterable to furnsh(). For example, one may write

kernels = [
    "leapseconds.tls",
    "mgs.tsc",
    "generic.bsp",
    "mgs.bc",
    "earth.bpc",
    "mgs.bes",
]

furnsh(kernels)

Kernel Priority

It is fairly common that two kernels of the same type - for example two SPKs - to have "competing data." "Competing" means that both kernels could provide an answer to the user's request for data, even though the numeric results would likely be different. This usually occurs when the two kernels were produced using different input data and mostly contain non-competing data, but do have some overlap in time. When two or more kernels contain competing data a kernel loaded later has higher priority than kernel(s) loaded earlier. This is true whether using separate calls to furnsh() for each kernel to be loaded, or a single call to furnsh_c with a list of kernels to be loaded, or a call to furnsh() that loads a meta-kernel. See Appendix A for a more complete discussion on competing data.

If orientation data for a given body-fixed frame are provided in both a text PCK and a binary PCK, data from the binary PCK always have higher priority.

Path Symbols in Meta-kernels

Inside a meta-kernel it is sometimes necessary to qualify kernel names with their path names. To reduce both typing and the need to continue kernel names over multiple lines, meta-kernels allow users to define symbols for paths. This is done using two kernel variables:
PATH_VALUES
PATH_SYMBOLS

To create symbols for path names, one assigns an array of path names to the variable PATH_VALUES. Next, one assigns an array of corresponding symbol names to the variable PATH_SYMBOLS. The nth symbol in the second array represents the nth path name in the first array. Then you can prefix with path symbols the kernel names specified in the KERNELS_TO_LOAD variable. Each symbol is prefixed with a dollar sign to indicate that it is in fact a symbol.

Suppose in our example above the MGS kernels reside in the path

/flight_projects/mgs/SPICE_kernels

and the other kernels reside in the path

/generic/SPICE_kernels

Then we can add paths to our meta-kernel as follows:

\begindata

PATH_VALUES  = ( '/flight_projects/mgs/SPICE_kernels',
                 '/generic/SPICE_kernels'              )

PATH_SYMBOLS = ( 'MGS',
                 'GEN' )


KERNELS_TO_LOAD = ( '$GEN/leapseconds.tls',
                    '$MGS/mgs.tsc',
                    '$GEN/generic.bsp',
                    '$MGS/mgs.bc',
                    '$GEN/earth.bpc',
                    '$MGS/mgs.bes'           )

\begintext

It is not required that paths be abbreviated using path symbols; it's simply a convenience available to you. Caution: the symbols defined using PATH_SYMBOLS are not related to the symbols supported by a host shell or any other operating system interface.

Specifying Kernels Using Relative Paths

When a kernel is specified with a relative path, this path should be valid at the time when furnsh() is called and stay valid for the rest of the application run. This is required because SPICE stores kernel names as provided by the caller and uses them to open and close binary kernels as needed by the DAF/DAS handle manager subsystem (behind the scenes, to allow reading many more binary kernels than available logical units), and to automatically reload into the POOL the rest of text kernels that should stay loaded when a particular text kernel is unloaded.

Changing the working directory from within an application during an application run after calling furnsh() to load kernels specified using relative paths is likely to invalidate stored paths and prevent open/close and unload operations mentioned above. A simple workaround when this is needed is to specify kernels using absolute paths.

Keeping Track of Loaded Kernels

The KEEPER subsystem maintains a database of the load operations that furnsh() has performed during a program run. This is implemented using data structures of fixed size, so there is a limit on the maximum number of loaded kernels that the KEEPER subsystem can accommodate.

When a kernel is loaded using furnsh(), a new entry is created in the database of loaded kernels, whether or not the kernel is already loaded.

All load and unload operations (see the discussion of unload() below) affect the list of loaded kernels and therefore affect the results returned by the functions ktotal(), kdata(), and kinfo(), all of which are discussed below under "Finding Out What's Loaded."

Reloading Kernels

Reloading an already loaded kernel creates another (duplicate) entry in the database of loaded kernels, and thus decreases the available space in that list. furnsh()'s treatment of reloaded kernels is thus slightly different from that performed by the CSPICE low-level kernel loaders, which handle a reload operation by first unloading the kernel in question, then loading it.

Changing Kernel Priority

The recommended method of increasing the priority of a loaded binary kernel, or of a meta-kernel containing binary kernels, is to unload it using unload() (see below), then reload it using furnsh(). This technique helps reduce clutter in furnsh()'s kernel list.

Load Limits

furnsh() can currently keep track of up to 5000 kernels. The list of loaded kernels may contain multiple entries for a given kernel, so the number of distinct loaded kernels would be smaller if some have been reloaded. Unloading kernels using unload() frees room in the kernel list, so there is no limit on the total number of load and corresponding unload operations performed in a program run.

The DAF/DAS handle manager system imposes its own limit on the number of DAF binary kernels that may be loaded simultaneously. This limit is currently set to a total of 5000 DAF kernels.

Finding Out What's Loaded

SPICE-based applications may need to determine at run time which files have been loaded. Applications may need to find the DAF or DAS handles of loaded binary kernels so that the kernels may be searched. Some applications may need to unload kernels to make room for others, or change the priority of loaded kernels at run time.

SPICE provides kernel access functions to support these needs. For every loaded kernel, an application can find the name of the kernel, the kernel type (text or one of SPK, CK, DSK, PCK, or EK), the kernel's DAF or DAS handle if applicable, and the name of the meta-kernel used to load the kernel, if applicable.

The function ktotal() returns the count of loaded kernels having their types on a caller-supplied list of one or more types. The function kdata() returns information on the nth kernel of the set having the types named in the list. The two functions are normally used together. The following example shows how an application could retrieve summary information on the currently loaded SPK files:

#!/usr/bin/env python
"""
This script uses SpiceyPy to list the names of loaded SPK kernel files.
"""

import spiceypy as spice


def main():
    # Get the total number of loaded SPK kernels.
    count = spice.ktotal("spk")
    if count == 0:
        print("No SPK files loaded at this time.")
    else:
        print("The loaded SPK files are:\n")
    # Loop over each loaded kernel and retrieve its data.
    for which in range(count):
        # kdata returns a tuple: (file, file type, source, handle)
        file, file_type, source, handle = spice.kdata(which, "spk")
        print(file)


if __name__ == "__main__":
    main()

Above, the input argument "spk"

is a kernel type specifier. More generally, a blank-delimited list of types may be provided as the input argument. The set of types that may appear in the list is shown below.

SPK  --- All SPK kernels are counted in the total
CK   --- All CK kernels are counted in the total
PCK  --- All binary PCK kernels are counted in the
         total
DSK  --- All DSK kernels are counted in the total
EK   --- All EK kernels are counted in the total
TEXT --- All text kernels that are not meta-
         kernels are included in the total
META --- All meta-kernels are counted in the
         total
ALL  --- Every type of kernel is counted in the
         total

In this example, 'filtyp' is a string indicating the type of kernel. 'handle' is the file handle if the file is a binary SPICE kernel. 'source' is the name of the meta-kernel used to load the kernel, if applicable.

CSPICE also contains the function kinfo() that returns summary information about a kernel whose name is already known. kinfo() is called as follows:

# will throw a NotFoundError if file is not found
filtyp, source, handle = kinfo(file)

Unloading Kernels

CSPICE-based applications may need to remove loaded kernels. Possible reasons for this are:
  • to make room to load other kernels

  • to change the priority of loaded kernel data

  • to change the set of kernel data visible to CSPICE

The function unload() acts as an inverse to furnsh(): passing a kernel name to unload() undoes the effect of the previous load operation performed on that kernel using furnsh(). For binary kernels that have been loaded just once, the meaning of this is simple: the kernel is closed and the database referring to the file is adjusted to reflect the absence of the kernel. Text kernels are unloaded by clearing the kernel pool and then reloading the other text kernels not designated for removal.

Note that unloading text kernels has the side effect of wiping out any kernel variables and associated values that had been entered in the kernel pool using any of the kernel pool assignment functions, such as pcpool(). It is important to consider whether this side effect is acceptable when writing code that may unload text kernels or meta-kernels.

Call unload() as follows:

unload(kernel)

Unloading a meta-kernel involves unloading all the kernels referenced by the meta-kernel.

Loading of Non-native Text and Binary Kernels

The various platforms supported by CSPICE use different end-of-line (EOL) indicators in text files:
Environment                  Native End-Of-Line
                             Indicator
___________                  _____________________

PC DOS/Windows                <CR><LF>
Unix                          <LF>
Linux                         <LF>
Mac OS X                      <LF>

As of CSPICE version N0059, the SPICE text kernel loader furnsh() (and the deprecated loader ldpool()) can read and parse non-native text files. (Caution: the FORTRAN SPICELIB text kernel readers do not include this capability.) The CSPICE text file reader, rdtext(), does not possess the capability to read non-native text files.

Starting with the version N0052 release of the SPICE Toolkit (January, 2002), supported platforms are able to read DAF-based binary kernels (SPK, CK and binary PCK) that were written using a non-native binary representation. This access is read-only; any operations requiring writing to the file--for example, adding information to the comment area, or appending additional ephemeris data-- require prior conversion of the kernel to the native binary file format. See the "Convert User's Guide" for details.

Manipulating Kernel Pool Contents

The main way one adds to or changes the contents of the kernel pool is by "loading" a SPICE text kernel using the function furnsh(). However, the kernel subsystem also provides several other functions that allow one to change the contents of the kernel pool.
clpool()

Clears (initializes) the kernel pool, deleting all the variables in the pool.

kclear()

Clears (empties) the kernel pool, the kernel database (same effect as unloading all kernels), and re-initializes the subsystem. Use of kclear() also clears programmatic kernel pool assignments from the "put-pool" routines, e.g. pipool(), pdpool(), pcpool().

dvpool()

Deletes a specific variable from the kernel pool.

lmpool()

Similar in effect to loading a text kernel using furnsh(), but the data being loaded into the pool come from an array of strings instead of a text kernel.

pcpool()

Programmatically inserts a single character variable and its associated values into the kernel pool. The assignment is direct (the values replace any previously existing set of values associated with the variable.)

pdpool()

Programmatically inserts a single double precision variable and its associated values into the kernel pool. The assignment is direct.

pipool()

Programmatically inserts a single integer variable and its associated values into the kernel pool. The assignment is direct.

The following code fragment shows how the data provided in a leapseconds kernel (LSK) could be loaded using lmpool().

#!/usr/bin/env python
"""
This script uses SpiceyPy to demonstrate using lmpool.
"""

import spiceypy as spice


def main():
    text = [
        "DELTET/DELTA_T_A = 32.184",
        "DELTET/K = 1.657D-3",
        "DELTET/EB  = 1.671D-2",
        "DELTET/M = ( 6.239996 1.99096871D-7 )",
        "DELTET/DELTA_AT = ( 10, @1972-JAN-1",
        "                     11, @1972-JUL-1",
        "                     12, @1973-JAN-1",
        "                     13, @1974-JAN-1",
        "                     14, @1975-JAN-1",
        "                     15, @1976-JAN-1",
        "                     16, @1977-JAN-1",
        "                     17, @1978-JAN-1",
        "                     18, @1979-JAN-1",
        "                     19, @1980-JAN-1",
        "                     20, @1981-JUL-1",
        "                     21, @1982-JUL-1",
        "                     22, @1983-JUL-1",
        "                     23, @1985-JUL-1",
        "                     24, @1988-JAN-1",
        "                     25, @1990-JAN-1",
        "                     26, @1991-JAN-1",
        "                     27, @1992-JUL-1",
        "                     28, @1993-JUL-1",
        "                     29, @1994-JUL-1",
        "                     30, @1996-JAN-1",
        "                     31, @1997-JUL-1",
        "                     32, @1999-JAN-1",
        "                     33, @2006-JAN-1",
        "                     34, @2009-JAN-1  )",
    ]
    #
    # Add the contents of the buffer to the kernel pool:
    #
    spice.lmpool(text)


if __name__ == "__main__":
    main()

See the docstrings of the kernel subsystem functions for specific details regarding their use.

Detecting Changes in the Kernel Pool Using Watchers

Since loading SPICE text kernels often happens only at program initialization, a function that relies on data in the kernel pool may run more efficiently if it can store a local copy of the values needed and update these only when a change occurs in the kernel pool. Two functions are available that allow a quick test to see whether kernel pool variables have been updated.
swpool()

Sets up a watcher on a a list of variables so that a specified agent can be notified when any variables on the list have been updated.

cvpool()

Indicates whether or not any of an agent's variables have been updated since the last time the agent checked with the pool.

See the docstrings of these functions for details and examples of their use.

Appendix A -- Discussion of Competing Data

Binary Kernels

For binary kernels, the conditions resulting in competing data depend on the kernel type.

SPKs

For SPKs, a segment contains data of a single SPK type, providing ephemeris for a single target measured relative to a single center and given in a single reference frame, spanning between specified start and stop times. If ephemeris data from any two segments, whether found in a single SPK file or in two SPK files, are for the same target and have an overlap in the time spans covered, then the two kernels are said to have some competing data. Note that centers play no role in the competition: two segments with the same target and different centers may compete.

By definition, SPKs contain continuous data during the time interval covered by a segment, so there is no chance for a "data gap" in a segment within a higher priority file (later loaded file) leading to a state lookup coming from a segment in a lower priority file.

SPK segment chaining may lead to a problem. It may happen that you have loaded into your program sufficient SPK data to compute the desired state or position vector, but CSPICE nevertheless returns an error message saying insufficient ephemeris data have been loaded. This can occur if a higher priority SPK segment, for which there are not sufficient additional SPK data to fully construct your requested state or position vector, is masking (blocking) a segment that is part of a viable (complete) chain. See the BACKUP section of the SPK tutorial for further discussion about this.

Having competition between two SPKs can be a relatively common occurrence when using mission operations kernels, but is far less likely when using PDS-archived SPICE data sets because of the clean-up and consolidation actions usually taken when an archive delivery is produced.

CKs

For CKs, a segment contains data of a single CK type providing the orientation of a reference frame associated with one object or structure, such as a spacecraft or instrument (sometimes called the "to" reference frame), relative to a second reference frame, generally referred to as the base reference frame (sometimes called the "from" reference frame), spanning between specified start and stop times.

If transformation data from any two segments, whether found in a single CK file or in two CK files, are for the same object/structure (are for the same "to" frame) and have an overlap in the time span covered, then the two kernels may have competing data. But read on.

However, unlike for SPKs, competition between CK files goes beyond segment-level considerations. The so-called "continuous" CK types (Types 2 through 5) do not necessarily provide orientation results for any epoch falling within a segment--there may be real data gaps. And the now little used Type 1 CK, containing discrete instances of orientation data, can be thought of as containing mostly data gaps.

While some of the Toolkit software used to compute orientation obtained from CKs can provide an orientation result within a gap, this is usually not the case. See the CK tutorial and the "CK Required Reading" document for discussions on interpolation intervals, tolerance, and how the various CK readers work.

CK segment chaining may lead to a problem. It may happen that you have loaded into your program sufficient CK data to compute the desired rotation matrix, but CSPICE nevertheless returns an error message saying insufficient data have been loaded. This can occur if a higher priority CK segment, for which there are not sufficient additional CK data to fully construct your requested rotation matrix, is masking (blocking) a segment that is part of a viable (complete) chain.

Having competition between two CKs can be a relatively common occurrence when using mission operations kernels, but is far less likely when using PDS-archived SPICE data sets because of the clean-up and consolidation actions usually taken when an archive delivery is prepared.

Binary PCKs

For binary PCKs, a segment contains data of a single binary PCK type providing orientation of a reference frame associated with a single object (a body-fixed frame), relative to a second reference frame, which is always an inertial frame, spanning between specified start and stop times. If orientation data from any segment in one binary PCK and orientation data from any segment in a second binary PCK are for the same body-fixed frame and overlap in time, then the two kernels are said to have competing data.

At present binary PCKs produced by NAIF exist only for the earth and the moon. Having competition between the latest high precision, short term earth orientation binary PCK and the lower precision, long term predict earth orientation binary PCK is a clear possibility -- be sure to load the long term predict file first to ensure any higher precision files also loaded have higher priority.

Orientation data provided in any loaded binary PCK have priority over what would have otherwise been competing data provided in any loaded text PCK.

Text Kernels

If a given variable name has two or more assignments, with the final assignment made using the "=" operator, whether within a single loaded text kernel, or from multiple loaded text kernels, or achieved using CSPICE functions, the last such assignment supersedes all previous occurrences of the assignment. This superseding happens no matter how many values are contained in the last assignment. (It's as if all previous assignments for the subject name had never occurred.)

It is generally best to unload a text kernel before loading another one containing competing data.

Appendix B -- Glossary of Terms

Agent
A string associated with a list of kernel variables to be watched for updates. The string can be passed to the update checking function cvpool() to determine whether any of the variables on the list have been updated.

Often the string is the name of a function that needs to be informed if any of a specified set of kernel variables has had a change made to its associated value(s).

Assignment
What appears inside data blocks of a text kernel. Each assignment consists of three parts: a variable (also called variable name), an operator, and a scalar or vector value. For example,
BODY399_RADII = ( 6378.14   6378.14   6356.75 )

is an assignment with a vector value. Once a text kernel is loaded, the value(s) on the right hand sides of the assignments become associated with the variable names on the corresponding left hand sides. See "direct assignment" and "incremental assignment" below.

Continued string
A string value composed of two or more pieces--called elements--each of which is no longer than 80 characters.
Control words
Markers indicating the start of data or comment blocks, specifically
\begindata
\begintext
Direct assignment
A text kernel assignment, made using the "=" operator. When a direct assignment is processed during text kernel loading, it associates one or more values with a variable name, and in so doing, replaces any previous such associations.
Element
Within the kernel pool the length of a string value is limited to 80 characters. A string value that is longer than 80 characters may be stored in and extracted from the pool by chunking it into pieces--called elements--each of which is no longer than 80 characters. Such a string is referred to as a "continued string."
Incremental assignment
A text kernel assignment made using the "+=" operator. When an incremental assignment is processed during text kernel loading, it appends one or more values to the list of values already associated with a variable name. Any previous such associations are NOT replaced; rather they are supplemented with the new value(s). Incremental assignments may be made to variables that didn't previously exist in the kernel pool; in such cases incremental assignments are equivalent to direct assignments.
Keeper (subsystem)
The SPICE subsystem used to keep track of (manage) loaded kernel files. In this sense it is also involved with the unloading of kernels.
Kernel pool (sometimes just called "the pool")
A specially managed area of program memory where data from text kernel assignment statements are stored.
Kernel variable
Often a synonym for "variable name," but may refer to the combination of a variable name and its associated values.
Meta-kernel (also known as "FURNSH kernel")
A special kind of text kernel, used to name a collection of kernels that are to be loaded into a user's application at run-time. May include the path names for the kernels as well as the file names.
Operator
Within SPICE text kernels, an operator is either "=" or the sequence of "+" and "=", written as "+=". The former is used to make direct assignments, the latter is used to make incremental assignments.
Principal data
This term occurs only within this document. It is used to refer to the "elemental" data contained in a kernel, as opposed to meta-data or bookkeeping data. For instance, within an SPK the principal data are the polynomials or other numeric data providing ephemeris information. Not part of the principal data are the descriptive information placed in the comment area, the file architecture IDs, and the indexes that help the subsystem quickly find the principal data needed to return a state vector.
Value
That which appears on the right-hand side of an assignment. May be a single value or a vector of values.

variable name = value(s)

Variable name
That which appears on the left-hand side of an assignment.

variable name = value(s)

Vector value
Two or more values associated with a single variable name.

Appendix C -- Summary of Routines

Each of the function names is a mnemonic that translates into a short description of the function's purpose.
clpool()

Clear the pool of kernel variables

cvpool()

Check variable in the pool for update

dtpool()

Return information about a kernel pool variable

dvpool()

Delete a variable from the kernel pool

expool()

Confirm the existence of a pool kernel variable

furnsh()

Furnish a program with SPICE kernels

gcpool()

Get character data from the kernel pool

gdpool()

Get double precision values from the kernel pool

gipool()

Get integers from the kernel pool

gnpool()

Get names of kernel pool variables

kclear()

Clear and re-initialize the kernel database

kdata()

Return information about the nth loaded kernel

kinfo()

Return information about a specific loaded kernel

ktotal()

Return the number of kernels loaded using KEEPER

lmpool()

Load variables from memory into the pool

pcpool()

Put character strings into the kernel pool

pdpool()

Put double precision values into the kernel pool

pipool()

Put integers into the kernel pool

stpool()

Return a string associated with a kernel variable

swpool()

Set watch on a pool variable

szpool()

Get size parameters of the kernel pool

unload()

Unload a kernel

Appendix D -- Summary of Key Text Kernel Parameter Values

Text kernel limits
Maximum variable name length:                         32
Maximum length of any element of a string value:      80
Maximum number of distinct variables:              26003
Maximum number of numeric variable values:        400000
Maximum number of character strings
 stored in the kernel pool as values:              15000
Maximum length of a file name, including any
 path specification, placed in a meta-kernel:        255

Other applicable limits

Maximum total number of kernel files of any
type that can be loaded simultaneously:             5000