This is Alpha code. It has had some reasonable testing but is really just experimental at the moment. I have run and tested this on Linux and MacOSX only. MacOSX is my major development platform (x86 only).
INSTALLATION
first get a copy of c.o from kx.com for your architecture make libkdb.a from c.o for easy linking
ar -r libkdb.a c.o
or
libtool -static -o libkdb.a c.o
perl Makefile.PL
make
make test # you must have a 'q' process running on port 2222 (q -p 2222)
make install
DEPENDENCIES
This module requires these other modules and libraries:
c.o from kx.com for your architecture
q -p 2222 # for make test to actually do anything
NAME
Kx - Perl extension for Kdb+ http://kx.com
SYNOPSIS
use Kx;
DESCRIPTION
Create a wrapper around Kdb+ and Q in Perl using the C interface to Kdb+
EXPORT
None by default.
METHODS
New
$k = Kx->new(name=>'local22', host=>'localhost', port=>2222);
Create a new K object. Set the connection paramaters to conect to 'host' and 'port' as specified.
No connection is made to the server until you call $k->connect()
$k = new K host=>'localhost', port=>2222;
$k->connect(); # defaults to default;
# picks up previous default connection to localhost port 2222 and
# will use it as well.
$k1 = new K;
Environment
There are a number of environment details you can glean from the Kdb+
server you are connected to. They are:
my $arrayref = $k->tables; # The tables defined
my $arrayref = $k->funcs; # The functions defined
my $arrayref = $k->views; # The views defined
my $arrayref = $k->variables; # The variables defined
my $arrayref = $k->memory; # The memory details \w
my $dir = $k->cwd; # The current working directory
my $dir = $k->chdir($newdir); # Set the cwd
my $num = $k->GMToffset; # Offset from GMT for times
TABLES
Create a new table in Kdb+ named mytab with 3 columns col1, col2 and
col3. The keys will be on col1 and col3 This equates to the Q command
# Q command
q)mytab:([col1:;col3:] col2:)
# The long winded Perl way
$k->Tnew(name=>'mytab',keys=>['col1','col3'],cols=>['col2']);
To add data use Tinsert(). Each row is added in the order defined above. This line adds 1 into col1, 2 into col3 and 3 into col2 as the keys are always defined before the other columns.
$k->Tinsert('mytab',1,2,3);
To do a select over a table use Tselect(). Tselect() takes a variable name as its first argument. The select will be executed and assigned to the variable you define. This way no data is passed from Kdb+ to the client until it is needed.
$k->Tselect('a','select from mytab where col1>4');
To get the details of the stored selection
my $numrows = $k->Tnumrows('a');
my $numcols = $k->Tnumcols('a');
You may have run a number of Tselects() and now wish to pull back the data. To do this use Tget()
$k->Tget('table'); # table must be a table in the server
To get access to random values in the returned table from Tget().
$val = $k->Tindex(row,col);
To get the list of column names as Kdb+ knows them.
my $header = $k->Theader();
print "@$header\n";
To get the meta data for a table as defined in KDB do this.
my @meta = $k->Tmeta($table);
foreach(@meta)
{
print "(name type) => (@$_)\n";
}
To get a Perl reference to a column of data from the table (as K is column oriented) do the following:
my $colref = $k->Tcol(0); # get the zeroth column
print "Column 0 data is: @$colref\n";
Finally to delete or remove a table by name from the server:
$k->Tdelete('table');
Here is a list of the complete table methods we have so far:
$k->Tnew(name=>'thename',keys=>[],cols=>[]);
$k->Tinsert('table',1,2,3);
$k->Tbulkinsert('table',col1=>[],col2=>[],...);
$k->Tget('select statement');
$scalar = $k->Tindex($row,$col);
$arref = $k->Tcol(2);
$arref = $k->Theader;
$x = $k->Tnumrows;
$y = $k->Tnumcols;
$k->Tselect('table','select statement');
$k->Tsave('table','file');
$k->Tappend('table','file');
$k->Tload('table','file');
$k->Tdelete('table');
COMMANDS
Execute the code on an already accessable Kdb+ server. The query is
executed and the results are held in K structures in RAM. Example
$return = $k->cmd('b:til 100');
The cmd() function will return a reference to an array if the Q command returns a list. It will return a simple scalar if the result is a scalar response from Q. It will return a hash reference if the return result from Q is either table/keyed table/dictionary. You need to know what you are doing so can know what the result is (or use Perl's ref()).
If you have a Q script that you wish to run against the Kdb+ server you can use the do(file) method. Any error in your script that is caught will stop do(file) from proceeding. If you don't care when it is done then use dolater(file).
Both do() and dolater() don't return anything useful. They just blindly execute each line of Q against the server. If you want to check each command and do stuff as a result then use cmd() and check the result.
An example file name foo.txt holds the lines:
t:([]a:();b:())
insert[`t](`a;10.70)
insert[`t](`b;-5.6)
insert[`t](`c;21.73)
You can run that file by doing this:
$k->do("foo.txt");
ATOMS and STRUCTURES
To create Kdb+ atoms locally in RAM use the following calls.
$k = new K;
my $d;
$d=$k->bool(0); # boolean
$d=$k->byte(100); # char
$d=$k->char(ord('a')); # char
$d=$k->short(20);
$d=$k->int(70);
$d=$k->long(93939);
$d=$k->real(20.44); # remember 20.44 may look close as a real
$d=$k->float(20.44); # should look closer to 20.44 as a float
$d=$k->sym('mysymbol'); # A Kdb+ symbol
$d=$k->date(2007,4,22); # integer encoded date year, month, day
$d=$k->dt(time()); # Kdb+ datetime from epoch
$d=$k->tm(100); # Time type in milliseconds
These allow for fine grained control over the 'type' of K object you want. If you don't mind particularly about the type conversions then you can use perl2K() like this.
$d = $k->perl2K('mysymbol');
$d = $k->perl2K([qw/this will be a K list of symbols/]);
$d = $k->perl2K({this => 1, that => 2, 'is a' => 'dict'});
To get a Perl value back from a Kdb+ atom try this;
my $val = $d->val();
To get the internal value back from a Kdb+ atom try this;
my $val = $k->kval; # used in $x->cmd('func',$val)
As a further comment on the date() method. When you look at the value
# Simple create
my $k = new K;
my $bool = $k->bool(0);
print "My boolean in K is ",$bool->val,"\n";
LISTS
These list functions create in memory local lists outside of any 'q'
running process. These will allow you to create very large simple lists
without blowing out all your memory.
To create a simple Kdb+ list of a single type use the listof() function. The type of the list is passed in as the second aregument and can be one
Kx::KC() char
Kx::KD() date yyyy mm dd
Kx::KE() real
Kx::KF() float
Kx::KG() byte
Kx::KH() short
Kx::KI() integer
Kx::KJ() long
Kx::KM() month
Kx::KS() symbol (internalised string)
Kx::KT() time
Kx::KU() minute
Kx::KV() second
Kx::KZ() datetime epoch seconds
Example simple lists:
my $k = new K;
my $list = $k->listof(20,Kx::KS()); # List of 20 symbols
for( my $i=0; $i < 20; $i++)
{
$list->at($i,"symbol$i");
}
# To get at the 4th element
my $sym = $list->at(3); # symbol3
my $perl_list = $list->list;
print "Symbols are @$perl_list\n";
# dates
$d = $k->listof(20,Kx::KD());
for( my $i=0; $i < 20; $i++)
{
$d->at($i,2007,4,$i+1); # 20070401 -> 20070421
}
# Add an extra date to the end of the list
my $day = $k->date(2007,4,30);
$d->joinatom($day->kval);
Kx::LIST
You may wish to tie a Perl array to a Kdb+ variable. Well, you can do that as well. Try something like this:
use K;
my %config = (
host=>"localhost",
port=>2222,
type=>'symbol',
list=>'d',
create=>1
);
tie(@a, 'Kx::LIST', %config);
# push lost of stuff on an array
my @array = (qw/aaaa bbbbb ccccc ddddddddd e f j h i j k l/) x 30000
;
push(@a,@array);
push(@a,@array);
push(@a,@array);
print "\@a has ", scalar(@a)," elements\n";
# Store
$a[3] = "Help me";
print "Elementt 3 is ",$a[3],"\n";
All the functions defined in perltie for lists are included.
Note: 'type' is a Kdb+ type as defined in Types below - it is the type for the array. Only simple types are allowed at the moment.
Kx::HASH
You may wish to tie a Perl hash to a Kdb+ variable. Well, you can do that as well. Try something like this:
use K;
my %config = (
host=>"localhost",
port=>2222,
ktype=>'symbol',
vtype=>'int',
dict=>'x',
create=>1
);
tie(%x, 'Kx::HASH', %config);
print "Size of hash x is :". scalar %x ."\n";
for(0..5)
{
$x{"a$_"} = $_;
}
%y = %x;
for(0..5)
{
print $y{"a$_"}," " if exists $y{"a$_"};
}
print "\n";
while(($k,$v) = each %x)
{
print "Key=>$k is $v\n";
}
untie(%x);
All the functions defined in perltie for hashs are included.
Note: ktype is a Kdb+ type as defined in Types below - it is the 'key' type for the hash. vtype is also defined in Types - it is the value type. Only simple types are allowed at the moment.
SEE ALSO
See the test code under the 't' directory of this module for more details on how to call each method.
AUTHOR
Mark Pfeiffer, <markpf at mlp-consulting dot com dot au>
COPYRIGHT AND LICENSE
Copyright (C) 2007 by Mark Pfeiffer
This library is free software; you can redistribute it and/or modify it under the same terms as Perl itself, either Perl version 5.8.6 or, at your option, any later version of Perl 5 you may have available.