| ekfind |
|
Table of contents
Procedure
EKFIND ( EK, find data )
SUBROUTINE EKFIND ( QUERY, NMROWS, ERROR, ERRMSG )
Abstract
Find E-kernel data that satisfy a set of constraints.
Required_Reading
EK
Keywords
EK
PARSE
SEARCH
Declarations
IMPLICIT NONE
INCLUDE 'ekquery.inc'
INCLUDE 'ekqlimit.inc'
INCLUDE 'ektype.inc'
INCLUDE 'ekopcd.inc'
INCLUDE 'ekcnamsz.inc'
CHARACTER*(*) QUERY
INTEGER NMROWS
LOGICAL ERROR
CHARACTER*(*) ERRMSG
Brief_I/O
VARIABLE I/O DESCRIPTION
-------- --- --------------------------------------------------
QUERY I Query specifying data to be found.
NMROWS O Number of matching rows.
ERROR O Flag indicating whether query parsed correctly.
ERRMSG O Parse error description.
Detailed_Input
QUERY is a character string that specifies a set of EK
data to select from those present in currently
loaded EK files. The selected data will be
retrievable via the EK fetch routines EKGC, EKGD,
and EKGI.
The query consists of four clauses, the third and
fourth of which are optional. The general form
of a query is
SELECT <column list>
FROM <table list>
[WHERE <constraint list>]
[ORDER BY <ORDER BY column list>]
where brackets indicate optional items. The
elements of the query shown above are called,
respectively, the `SELECT clause', the
`FROM clause', the `WHERE clause', and the
`ORDER BY clause'. The result of a query may be
thought of as a new table, whose columns are those
specified in the SELECT clause, whose rows are
those satisfying the constraints of the WHERE
clause, and whose rows are ordered according to
the ORDER BY clause.
The SELECT clause specifies a list of columns
from which data are to be selected. In a simple
(non-join) query, these columns must belong to
the single table specified in the FROM clause.
The form of a SELECT clause is
SELECT <column name> [ ,<column name>...]
In queries having multiple tables in the FROM
clause, column names are ambiguous if they occur
in more than one table in the FROM clause. Such
column names must be qualified with table
identifiers. These identifiers may be the names of
the tables to which the columns belong, or table
`aliases', names (usually short ones) associated
with tables in the FROM clause. Table aliases have
duration limited to the execution of the query to
which they belong.
The form of a qualified column name is
<table name>.<column name>
or
<table alias>.<column name>
The FROM clause specifies the tables from which
data are to be selected. In simple queries, only
one table is listed. In this case the form of
the FROM clause is
FROM <table name>
In queries involving multiple tables, the form of
the FROM clause becomes
FROM <table name> [<table alias>]
[ , <table name> [<table alias>] ... ]
The aliases associated with the table names must
be distinct and must not be the actual names of
loaded EK tables.
Queries involving multiple tables are called
`joins'.
The meaning of a FROM clause containing multiple
tables is that the output is to be a subset of
the rows of the Cartesian product of the listed
tables. Normally, WHERE clause constraints are
supplied to reduce the selected rows to a set of
interest.
The most common example of a join is a query with
two tables listed in the FROM clause, and a WHERE
clause constraint enforcing equality of members
of a column in the first table with members of
column in the second table. Such a query is
called an `equi-join'. A join in which columns
of different tables are related by an inequality
is called a `non-equi-join'. Any type of join
other than an equi-join may be very slow to
evaluate, due to the large number of elements that
may be contained in the Cartesian
product of the listed tables.
The WHERE clause lists constraints that must
be met by each row satisfying the query. The
constraints are specified as a logical combination
of relational expressions. The form of the
constraint list is
WHERE <constraint expression>
where each <constraint expression> consists of one
or more simple relational expressions of the form
<column name> <operator> <RHS symbol>
where
<RHS symbol>
is a column name, a literal value, or the special
symbol
NULL
and
<operator>
is any of
EQ, GE, GT, LE, LIKE, LT, NE, NOT LIKE, <, <=,
=, >, >=, !=, <>
For comparison with null values, the special
syntaxes
<column name> IS NULL
<column name> IS NOT NULL
are allowed, in addition to the standard
comparison syntaxes using the equality or
inequality operators.
The LIKE operator allows comparison of a string
value against a template. The template syntax
is that allowed by the SPICELIB routine MATCHI.
Templates may include literal characters, the
wild string marker '*', and the wild character
marker '%'. Case is significant in templates.
Templates are bracketed by quote characters, just
as are literal strings.
The query language also supports the BETWEEN and
NOT BETWEEN constructs
<column> BETWEEN <symbol 1> AND <symbol 2>
<column> NOT BETWEEN <symbol 1> AND <symbol 2>
The tokens
<symbol 1>
<symbol 2>
may be literal values or column names.
The BETWEEN operator considers values that match
the bounds to satisfy the condition: the BETWEEN
operator tests for inclusion in the closed interval
defined by the bounds.
In the WHERE clause, simple relational expressions
may be combined using the logical operators AND,
OR, and NOT, as in the Fortran programming
language. Parentheses may be used to enforce a
desired order of evaluation of logical expressions.
The expression syntax is NOT symmetric: literal
values must not appear on the left hand side of the
operators that apply to them.
The columns named in a constraint clause must
belong to the tables listed in the FROM clause.
If the query is a join, qualifying table names or
aliases are required wherever their omission would
result in ambiguity.
Data types of the columns or constants used on the
right-hand-sides of operators must match the data
types of the corresponding columns on the
left-hand-sides, except that comparison of integer
and double precision quantities is permitted.
Literal strings used in constraints are always
bracketed by quotes. Either single quotes (')
or double quotes (") may be used, but the same
quote character must be used to start and end any
literal string. Within character string values,
quote characters must be doubled in order to be
recognized. Case is significant in character
except in comparisons using the LIKE and NOT LIKE
operators, which ignore case: the expression
ANIMAL LIKE "*A*"
would be considered true when ANIMAL takes the
value
"cat"
Time values are considered to be strings and
require bracketing quotes. Currently, the
only time values allowed are UTC times in ISO
format, UTC times represented in forms accepted by
the SPICELIB routine TPARSE, and SCLK strings in
NAIF format.
The ORDER BY clause indicates which columns to
use to order the output generated by the query.
The columns in the ORDER BY clause define a
dictionary ordering, with the first listed column
acting as a primary key, the second column acting
as a secondary key, and so on.
For each ORDER BY column, the keywords ASC or DESC
may be supplied to indicate whether the items in
that column are to be listed in ascending or
descending order. Ascending order is the default.
The direction in which data items increase is
referred to as the `order sense'.
The ORDER BY clause, if present, must appear
last in the query.
The form of the ORDER BY clause is
ORDER BY <column name> [<order sense>]
[ ,<column name> [<order sense>]...]
Rows satisfying the query constraints will be
returned so that the entries of the first column
specified in the ORDER BY clause will be appear in
the order specified by the order sense keyword,
which is assumed to be ASC if absent. When entries
in the first through Nth ORDER BY column are equal,
the entries in the (N+1)st ORDER BY column
determine the order of the rows, and so on.
As in the WHERE clause, column names must be
qualified by table names or table aliases where
they would otherwise be ambiguous.
The query language is word-oriented, and some
indicate whether the words are reserved. Reserved
words must be separated from other words by white
space. It is not necessary to use white space
to separate words and punctuation characters.
The list of reserved words is
AND
BETWEEN
BY
COLUMN
EQ
FROM
GE
GT
IS
LE
LT
LIKE
NE
NOT
NULL
OR
ORDER
SELECT
WHERE
The left and right parenthesis characters are also
reserved; they may not be used in queries outside
of quoted strings.
Case is not significant in queries, except within
literal strings.
Detailed_Output
NMROWS is the number of rows that match the query
criteria. NMROWS is defined if and only if
ERROR is returned .FALSE.
ERROR is a logical flag indicating whether the query
failed to parse correctly.
ERRMSG is a character string that describes EKFIND's
diagnosis of a parse error, should one occur.
Otherwise, ERRMSG will be returned blank.
Parameters
See the include files.
Exceptions
1) Most of the exceptions that can occur on a call to
EKFIND are caused by errors in the input query. EKFIND
attempts to diagnose these via the output error flag and
error message, instead of signaling errors. The following
classes of errors are detected:
Scanning errors---these result from badly formed query
in which EKFIND could not identify all of the tokens.
When these errors occur, EKFIND may be too confused to
give a helpful diagnostic message.
Parsing errors---these result from a badly formed
query that EKFIND was able to separate into tokens
but that EKFIND determined to be syntactically invalid:
Name resolution errors---these result from referencing
invalid or ambiguous column or table names in a query.
Time resolution errors---these result from use of time
strings that cannot be parsed.
Semantic errors---these result from a syntactically
valid query that violates a limit or a restriction on
values used in a query.
Some problems with queries are not trapped by EKFIND but
instead cause errors to be signaled. These are listed below.
2) If no E-kernels are loaded at the time this routine is called,
an error is signaled by a routine in the call tree of this
routine.
3) If a leapseconds kernel is is not loaded before this routine
is called, UTC time values may not be used in queries. If they
are, an error is signaled by a routine in the call tree of
this routine.
4) If an SCLK kernel for the appropriate spacecraft clock has not
been loaded before this routine is called, SCLK values for
that clock may not be used in queries. If they are, an error
is signaled by a routine in the call tree of this routine.
Files
None.
Particulars
This routine operates almost entirely by side effects: it
prepares the EK fetch routines to return event data that
satisfy the input query. See the EK Required Reading for
examples of use of this routine in conjunction with the EK
fetch routines.
Examples
The numerical results shown for these examples may differ across
platforms. The results depend on the SPICE kernels used as
input, the compiler and supporting libraries, and the machine
specific arithmetic implementation.
1) Perform a query on an EK file that contains a database with
the different commands of the Deep Impact spacecraft
subsystem, and a table with the subsystem id, the parameter
name, and the description of that parameters, ordered by
subsystem name. Print the number of records of that table.
Use the EK kernel below to load the Deep Impact spacecraft
subsystem commands dictionary.
dif_cmdict_128_20050620.bdb
Example code begins here.
PROGRAM EKFIND_EX1
IMPLICIT NONE
C
C Include the EK Maximum length of an input query,
C MAXQRY, and the maximum length of literal string
C values, MAXSTR, from eklimit.inc.
C
INCLUDE 'ekqlimit.inc'
C
C Local parameters
C
CHARACTER*(*) EKNAME
PARAMETER ( EKNAME =
. 'dif_cmdict_128_20050620.bdb' )
INTEGER ERRLEN
PARAMETER ( ERRLEN = 1840 )
C
C Local variables
C
CHARACTER*(ERRLEN) ERRMSG
CHARACTER*(MAXQRY) QUERY
INTEGER NMROWS
LOGICAL ERROR
C
C Open an EK file.
C
CALL FURNSH ( EKNAME )
C
C The EK file contains the table 'DIF_COMMANDS',
C and that 'DIF_COMMANDS' contains columns named:
C
C SUBSYSTEM, COMMAND, PARAMETER_NAME, DESCRIPTION
C
C Define a set of constraints to perform a query on all
C loaded EK files (the SELECT clause).
C
QUERY = 'Select SUBSYSTEM, COMMAND, PARAMETER_NAME, '
. // 'DESCRIPTION from DIF_COMMANDS '
. // 'order by SUBSYSTEM'
C
C Query the EK system for data rows matching the
C SELECT constraints.
C
CALL EKFIND ( QUERY, NMROWS, ERROR, ERRMSG )
C
C Check whether an error occurred while processing the
C SELECT clause. If so, output the error message.
C
IF ( ERROR ) THEN
WRITE(*,*) 'SELECT clause error: ', ERRMSG
ELSE
C
C If no error, NMROWS contains the number of rows
C matching the constraints specified in the query
C string.
C
WRITE(*,*) 'Number of matching rows: ', NMROWS
END IF
END
When this program was executed on a Mac/Intel/gfortran/64-bit
platform, the output was:
Number of matching rows: 5798
2) Examples of strings containing syntactically valid queries:
SELECT COL1 FROM TAB1
select col1 from tab1 where col1 gt 5
SELECT COL2 FROM TAB1 WHERE COL2 > 5.7D0 ORDER BY COL2
SELECT COL2 FROM TAB1 WHERE COL1 != 5
SELECT COL2 FROM TAB1 WHERE COL1 GE COL2
SELECT COL1, COL2, COL3 FROM TAB1 ORDER BY COL1
SELECT COL3 FROM TAB1 WHERE COL5 EQ "ABC"
SELECT COL3 FROM TAB1 WHERE COL5 = 'ABC'
SELECT COL3 FROM TAB1 WHERE COL5 LIKE 'A*'
SELECT COL3 FROM TAB1 WHERE COL5 LIKE 'A%%'
SELECT COL4 FROM TAB1 WHERE COL4 = '1995 JAN 1 12:38:09.7'
SELECT COL4 FROM TAB1 WHERE COL4 = "1995 JAN 1 12:38:09.7"
SELECT COL4 FROM TAB1 WHERE
COL4 NE 'GLL SCLK 02724646:67:7:2'
SELECT COL1 FROM TAB1 WHERE COL1 != NULL
SELECT COL1 FROM TAB1 WHERE COL1 IS NULL
SELECT COL1 FROM TAB1 WHERE COL1 IS NOT NULL
SELECT COL1, COL2, COL3 FROM TAB1
WHERE (COL1 BETWEEN 4 AND 6) AND (COL3 NOT LIKE "A%%")
ORDER BY COL1, COL3
SELECT COL4 FROM TAB1
WHERE COL4 BETWEEN "1995 JAN 1 12:38" AND
"October 23, 1995"
SELECT COL1, COL2 FROM TAB1 WHERE
NOT ( ( ( COL1 < COL2 ) AND ( COL1 > 5 ) ) OR
( ( COL1 >= COL2 ) AND ( COL2 <= 10 ) ) )
SELECT T1.COL1, T1.COL2, T2.COL2, T2.COL3
FROM TABLE1 T1, TABLE2 T2
WHERE T1.COL1 = T2.COL1
AND T1.COL2 > 5
ORDER BY T1.COL1, T2.COL2
3) Examples of syntactically invalid queries:
SELECT TIME WHERE TIME
LT 1991 JAN 1 {FROM clause is absent}
select time from table1 where
time lt 1991 jan 1 {time string is not
quoted}
select time from table1
where time .lt. '1991 jan 1' {operator should be lt}
select cmd from table1
where "cmd,6tmchg" != cmd {value is on left side
of operator}
select event_type from table1
where event_type eq "" {quoted string is empty
---use " " to indicate
a blank string}
select event_type from table1
where event_type = "COMMENT"
order TIME {ORDER BY phrase is
lacking BY keyword}
select COL1 from table where
where COL1 eq MOC_EVENT {literal string on
right-hand-side of
operator is not quoted}
In the following examples, we'll assume that the program
calling EKFIND has loaded an EK containing two segments
having columns having the following names and attributes:
TABLE1:
==========
Column name Data type Size Indexed?
----------- --------- ---- --------
EVENT_TYPE CHARACTER*32 1 YES
EVENT_PARAMETERS CHARACTER*(*) 1 NO
COMMENT CHARACTER*80 VARIABLE NO
TABLE2:
==========
Column name Data type Size Indexed?
----------- --------- ---- --------
EVENT_TYPE CHARACTER*32 1 YES
EVENT_PARAMETERS CHARACTER*80 1 NO
COMMENT CHARACTER*80 VARIABLE NO
COMMAND CHARACTER*80 1 YES
Then the following queries are semantically invalid:
SELECT EVENT_PARAMETERS
FROM TABLE1
WHERE EVENT_DURATION = 7.0 {No column called
EVENT_DURATION
is present in a loaded
EK}
SELECT COMMENT FROM TABLE2
WHERE COMMENT EQ "N/A" {The COMMENT column does
not have size 1 and
therefore cannot be
referenced in a query}
Restrictions
1) A leapseconds kernel must be loaded before this routine may
be called, if UTC time values are used in input queries.
2) An appropriate SCLK kernel must be loaded before this routine
may be called, if SCLK values are used in input queries.
3) Data found in response to a query become unavailable
when a fast load is initiated via EKIFLD. Any desired
fetches of the data must be performed before a fast
load or any other operation that modifies the EK scratch
area is initiated.
Literature_References
None.
Author_and_Institution
N.J. Bachman (JPL)
J. Diaz del Rio (ODC Space)
B.V. Semenov (JPL)
Version
SPICELIB Version 1.1.0, 13-AUG-2021 (JDR)
Added IMPLICIT NONE statement.
Edited the header to comply with NAIF standard.
Added complete code example.
SPICELIB Version 1.0.4, 18-MAY-2010 (BVS)
Removed "C$" marker from text in the header.
SPICELIB Version 1.0.3, 19-DEC-2001 (NJB)
$Restrictions section was updated.
SPICELIB Version 1.0.2, 14-JAN-1997 (NJB)
Syntax descriptions for comparisons using null values have been
added. The $Examples section was augmented with sample queries
demonstrating use of the IS NULL and IS NOT NULL comparison
operators.
SPICELIB Version 1.0.1, 16-AUG-1996 (NJB)
$Exceptions section of header was updated to indicate that
calling this routine while no E-kernels are loaded will cause
an error to be signaled. Previous version line was changed
from "Beta" to "SPICELIB," and the previous version was
corrected to 1.0.0.
SPICELIB Version 1.0.0, 24-OCT-1995 (NJB)
|
Fri Dec 31 18:36:18 2021