|
Документ взят из кэша поисковой машины. Адрес
оригинального документа
: http://www.stsci.edu/~sontag/spicedocs/cspice/ekacld_c.html
Дата изменения: Sat Dec 17 06:08:50 2005 Дата индексирования: Mon Apr 11 00:00:17 2016 Кодировка: |
Add an entire double precision column to an EK segment.
EK
Variable I/O Description -------- --- -------------------------------------------------- handle I EK file handle. segno I Number of segment to add column to. column I Column name. dvals I Double precision values to add to column. entszs I Array of sizes of column entries. nlflgs I Array of null flags for column entries. rcptrs I Record pointers for segment. wkindx I-O Work space for column index.
handle the handle of an EK file that is open for writing.
A "begin segment for fast write" operation must
have already been performed for the designated
segment.
segno is the number of the segment to which data is to be
added. Segments are numbered from 0 to nseg-1, where
nseg is the count of segments in the file.
column is the name of the column to be added. All of
the data for the named column will be added in
one shot.
dvals is an array containing the entire set of column
entries for the specified column. The entries
are listed in row-order: the column entry for the
first row of the segment is first, followed by the
column entry for the second row, and so on. The
number of column entries must match the declared
number of rows in the segment. For columns having
fixed-size entries, a null entry must be allocated
the same amount of space occupied by a non-null
entry in the array dvals. For columns having
variable-size entries, null entries do not require
any space in the dvals* array, but in any case must
have their allocated space described correctly by
the corresponding element of the entszs array
(described below).
entszs is an array containing sizes of column entries.
The Ith element of entszs gives the size of the
Ith column entry. entszs is used only for columns
having variable-size entries. For such columns,
the dimension of entszs must be at least nrows.
The size of null entries should be set to zero.
For columns having fixed-size entries, the
dimension of this array may be any positive value.
nlflgs is an array of logical flags indicating whether
the corresponding entries are null. If the Ith
element of nlflgs is SPICEFALSE, the Ith column entry
defined by dvals and entszs is added to the
current segment in the specified kernel file.
If the Ith element of nlfgls is SPICETRUE, the
contents of the Ith column entry are undefined.
nlflgs is used only for columns that allow null
values; it's ignored for other columns.
rcptrs is an array of record pointers for the input
segment. This array is obtained as an output
from ekifld_c, the routine called to initiate a
fast write.
wkindx is a work space array used for building a column
index. If the column is indexed, the dimension of
wkindx_c must be at nrows, where nrows is the number
of rows in the column. If the column is not
indexed, this work space is not used, so the
dimension may be any positive value.
None. See $Particulars for a description of the effect of this routine.
None.
This routine operates by side effects: it modifies the named EK file by adding data to the specified column. This routine writes the entire contents of the specified column in one shot. This routine creates columns much more efficiently than can be done by sequential calls to ekaced_c, but has the drawback that the caller must use more memory for the routine's inputs. This routine cannot be used to add data to a partially completed column.
1) Suppose we have an E-kernel named order_db.ek which contains
records of orders for data products. The E-kernel has a
table called DATAORDERS that consists of the set of columns
listed below:
DATAORDERS
Column Name Data Type
----------- ---------
ORDER_ID INTEGER
CUSTOMER_ID INTEGER
LAST_NAME CHARACTER*(*)
FIRST_NAME CHARACTER*(*)
ORDER_DATE TIME
COST DOUBLE PRECISION
The order database also has a table of items that have been
ordered. The columns of this table are shown below:
DATAITEMS
Column Name Data Type
----------- ---------
ITEM_ID INTEGER
ORDER_ID INTEGER
ITEM_NAME CHARACTER*(*)
DESCRIPTION CHARACTER*(*)
PRICE DOUBLE PRECISION
We'll suppose that the file ORDER_DB.EK contains two segments,
the first containing the DATAORDERS table and the second
containing the DATAITEMS table.
Below, we show how we'd open a new EK file and create the
first of the segments described above.
#include "SpiceUsr.h"
#include <stdio.h>
void main()
{
/.
Constants
./
#define CNMLEN ( CSPICE_EK_COL_NAM_LEN + 1 )
#define DECLEN 201
#define EKNAME "order_db.ek"
#define FNMLEN 50
#define IFNAME "Test EK/Created 20-SEP-1995"
#define LNMLEN 50
#define LSK "leapseconds.ker"
#define NCOLS 6
#define NRESVC 0
#define NROWS 9
#define TABLE "DATAORDERS"
#define TNMLEN CSPICE_EK_TAB_NAM_LEN
#define UTCLEN 30
/.
Local variables
./
SpiceBoolean nlflgs [ NROWS ];
SpiceChar cdecls [ NCOLS ] [ DECLEN ];
SpiceChar cnames [ NCOLS ] [ CNMLEN ];
SpiceChar fnames [ NROWS ] [ FNMLEN ];
SpiceChar lnames [ NROWS ] [ LNMLEN ];
SpiceChar dateStr [ UTCLEN ];
SpiceDouble costs [ NROWS ];
SpiceDouble ets [ NROWS ];
SpiceInt cstids [ NROWS ];
SpiceInt ordids [ NROWS ];
SpiceInt handle;
SpiceInt i;
SpiceInt rcptrs [ NROWS ];
SpiceInt segno;
SpiceInt sizes [ NROWS ];
SpiceInt wkindx [ NROWS ];
/.
Load a leapseconds kernel for UTC/ET conversion.
./
ldpool_c ( LSK );
/.
Open a new EK file. For simplicity, we will not
reserve any space for the comment area, so the
number of reserved comment characters is zero.
The constant IFNAME is the internal file name.
./
ekopn_c ( EKNAME, IFNAME, NRESVC, &handle );
/.
Set up the table and column names and declarations
for the DATAORDERS segment. We'll index all of
the columns. All columns are scalar, so we omit
the size declaration. Only the COST column may take
null values.
./
strcpy ( cnames[0], "ORDER_ID" );
strcpy ( cdecls[0], "DATATYPE = INTEGER, INDEXED = TRUE" );
strcpy ( cnames[1], "CUSTOMER_ID" );
strcpy ( cdecls[1], "DATATYPE = INTEGER, INDEXED = TRUE" );
strcpy ( cnames[2], "LAST_NAME" );
strcpy ( cdecls[2], "DATATYPE = CHARACTER*(*),"
"INDEXED = TRUE" );
strcpy ( cnames[3], "FIRST_NAME" );
strcpy ( cdecls[3], "DATATYPE = CHARACTER*(*),"
"INDEXED = TRUE" );
strcpy ( cnames[4], "ORDER_DATE" );
strcpy ( cdecls[4], "DATATYPE = TIME, INDEXED = TRUE" );
strcpy ( cnames[5], "COST" );
strcpy ( cdecls[5], "DATATYPE = DOUBLE PRECISION,"
"INDEXED = TRUE,"
"NULLS_OK = TRUE" );
/.
Start the segment. We presume the number of rows
of data is known in advance.
./
ekifld_c ( handle, TABLE, NCOLS, NROWS, CNMLEN,
cnames, DECLEN, cdecls, &segno, rcptrs );
/.
At this point, arrays containing data for the
segment's columns may be filled in. The names
of the data arrays are shown below.
Column Data array
"ORDER_ID" ordids
"CUSTOMER_ID" cstids
"LAST_NAME" lnames
"FIRST_NAME" fnames
"ORDER_DATE" odates
"COST" costs
The null flags array indicates which entries are null.
It is ignored for columns that don't allow null
values. In this case, only the COST column allows
nulls.
Fill in data arrays and null flag arrays here. This code
section would normally be replaced by calls to user functions
returning column values.
./
for ( i = 0; i < NROWS; i++ )
{
ordids[i] = i;
cstids[i] = i*100;
costs [i] = (SpiceDouble) 100*i;
sprintf ( fnames[i], "Order %d Customer first name", i );
sprintf ( lnames[i], "Order %d Customer last name", i );
sprintf ( dateStr, "1998 Mar %d", i );
utc2et_c ( dateStr, ets+i );
nlflgs[i] = SPICEFALSE;
}
nlflgs[1] = SPICETRUE;
/.
The sizes array shown below is ignored for scalar
and fixed-size array columns, so we need not
initialize it. For variable-size arrays, the
Ith element of the sizes array must contain the size
of the Ith column entry in the column being written.
Normally, the sizes array would be reset for each
variable-size column.
Add the columns of data to the segment. All of the
data for each column is written in one shot.
./
ekacli_c ( handle, segno, "order_id", ordids,
sizes, nlflgs, rcptrs, wkindx );
ekacli_c ( handle, segno, "customer_id", cstids,
sizes, nlflgs, rcptrs, wkindx );
ekaclc_c ( handle, segno, "last_name", LNMLEN,
lnames, sizes, nlflgs, rcptrs, wkindx );
ekaclc_c ( handle, segno, "first_name", FNMLEN,
fnames, sizes, nlflgs, rcptrs, wkindx );
ekacld_c ( handle, segno, "order_date", ets,
sizes, nlflgs, rcptrs, wkindx );
ekacld_c ( handle, segno, "cost", costs,
sizes, nlflgs, rcptrs, wkindx );
/.
Complete the segment. The rcptrs array is that
returned by ekifld_c.
./
ekffld_c ( handle, segno, rcptrs );
/.
At this point, the second segment could be
created by an analogous process. In fact, the
second segment could be created at any time; it is
not necessary to populate the first segment with
data before starting the second segment.
The file must be closed by a call to ekcls_c.
./
ekcls_c ( handle );
}
1) Only one segment can be created at a time using the fast
write routines.
2) No other EK operation may interrupt a fast write. For
example, it is not valid to issue a query while a fast write
is in progress.
1) If handle is invalid, the error will be diagnosed by routines
called by this routine.
2) If column is not the name of a declared column, the error will
be diagnosed by routines called by this routine.
3) If column specifies a column of whose data type is not
double precision, the error SPICE(WRONGDATATYPE) will be
signaled.
4) If the specified column already contains ANY entries, the
error will be diagnosed by routines called by this routine.
5) If an I/O error occurs while reading or writing the indicated
file, the error will be diagnosed by routines called by this
routine.
6) If the string pointer for column is null, the error
SPICE(NULLPOINTER) will be signaled.
7) If the input string column has length zero, the error
SPICE(EMPTYSTRING) will be signaled.
See the EK Required Reading for a discussion of the EK file format.
N.J. Bachman (JPL)
None.
-CSPICE Version 1.1.1, 09-JAN-2002 (NJB)
Documentation change: instances of the phrase "fast load"
were replaced with "fast write."
-CSPICE Version 1.1.0, 07-JUL-1998 (NJB)
Bug fix: now uses dynamically allocated array of type logical
to interface with underlying f2c'd function ekacld_.
Now maps segno from C to Fortran range.
Added "undef" of masking macro. Changed input pointer types
to pointers to const objects.
-CSPICE Version 1.0.0, 01-APR-1998 (NJB)
Based on SPICELIB Version 1.0.0, 08-NOV-1995 (NJB)
write entire double precision column to EK segment