kdb+ Data Types

Kdb+ provides a large number of data types, in this tutorial we will consider defining the various types. This tutorial forms part of the online kdb+ training course

kdb+ Data Types

For reference the data types available in kdb+ together with their space usage etc. are shown below.
You may also want to download and print our kdb+ cheat sheet that shows the data types and other useful snippets.

type	size	char	num	notation	Null Value	Positive Infinity
Mixed List			0
boolean	1	b	1	1b	0b
byte	1	x	4	0x26	0x00
short	2	h	5	42h	0Nh	0Wh
int	4	i	6	42	0N	0W
long	8	j	7	42j	0Nj	0Wj
real	4	e	8	4.2e	0Ne	0We
float	8	f	9	4.2	0n	0w
char	1	c	10	"z"	" "
symbol		s	11	`zaphod	`
timestamp	8	p	12	2011.07.08D21:48:48.703125000	0Np	0Wp
month	4	m	13	2006.07m	0Nm	0Wm
date	4	d	14	2006.07.21	0Nd	0Wd
datetime	4	z	15	2006.07.21T09:13:39	0Nz	0Wz
timespan	8	n	16	0D21:56:26.421875000	0Nn	0Wn
minute	4	u	17	12:11	0Nu	0Wu
second	4	v	18	12:11:17	0Nv	0Wv
time	4	t	19	09:01:02:042	0Nt	0Wt
enum	4	*	20-77	`u$v
table			98	([] c1:ab`c; c2:10 20 30)
dictionary			99	`a`b`c!!10 20 30

Defining Data Types

Defining Bytes / Booleans

q)1b 1b q)a:1b q)type a -1h q)type 0x12 -4h

The "char" for the given type is placed after the value to force it to be that type. Then when we call the type function it returns the type "num" for its argument. In the code above you can see booleans are type num 1 and bytes are type nums 4, this corresponds with what is shown in the data type table above.

Defining Numeric Types

q)type 1200300i -6h q)type 300400 -7h q)realNum:2.674635e q)floatNum:1.27263736f q)type realNum -8h q)type floatNum -9h q)floatNum:1.27263736262536f q)floatNum 1.272637 q)\P 7i q)\P 16 q)floatNum 1.27263736262536 q)\P 7

Similar to before, write the "char" at end to force type. Notice that when you do not explicitly specify a type long/float is assumed depending on if your number had decimal places. This behaviour depends on your kdb version, older 2.x defaulted to integers instead of longs.

For floating point numbers, the IEEE standard is used for reals (4 bytes) and floats (8 bytes). Therefore the standard accuracy you would expect is stored. The accuracy shown in the console itself depends on the system P setting.

Defining Character Types

q)type "a" -10h q)c:"\n" q)type c -10h q)c:"\t" q)type c -10h q)c "\t" q)s:`yahoo q)type s -11h q)`space space 'space q)sym:`$"space space" q)type sym -11h

Character data can be stored as char or symbols. Char stores unicode-8/ascii values. Symbols store a sequence of characters as a single entity.

Date Time kdb+ Data Types

Define DateTime TimeStamps etc.

q)d:2011.02.22 q)d:2011.2.22 '2011.2.22 q)`int$2000.02.01 31i q)t:11:02:58.000 q)dt:2006.07.04T09:05:01.00 q)type d -14h q)d 2011.02.22 q)t 11:02:58.000 q)type t -19h q)dt 2006.07.04T09:05:01.000 q)type dt -15h q)d.year 2011i q)d.mm 2i q)d.dd 22i q)d.month 2011.02m q)mon:2011.04m q)type mon -13h q)mm:11:05 q)mm 11:05 q)type mm -17h q)sec:11:05:59 q)type sec -18h

Dot Notation for Accessing Datetime Components or Casting

q)t 11:02:58.000 q)t.hh 11i q)t.mm 2i q) q)t.ss 58i q)t.minute 11:02 q)t.second 11:02:58 q) q)dt 2006.07.04T09:05:01.000 q)dt.date 2006.07.04 q)dt.time 09:05:01.000 q)dt.month 2006.07m

Using the shorter dot notations dt.dd, dt.mm, dt.ss returns that component

Using the longer dot notations dt.minute, dt.second casts the value to that type, truncating it's value.

Getting the Current Date, Time, TimeStamp

q).z.t 09:39:03.891 q).z.t 09:39:06.000 q).z.d 2013.08.20 q).z.p 2013.08.20D09:39:15.777185000 q).z.p 2013.08.20D09:39:18.065316000 q).z.n 0D09:39:29.395964000 q).z.n 0D09:39:31.831103000

To access the current date, time, timestamp etc. type '.z.' followed by the temporal type's char in the table above; for example .z.t will give the current time, .z.d the current date and so forth.

0N Nulls and 0W Infinities

q)0Ni 0Ni q)type 0Ni -6h q)null 0Ni 1b q)null 1i 0b q)0n 0n q)0Nz 0Nz q)0Wi 0Wi q)0We 0we q)-0Wj -0W q)0W > 1 1b q)0W < 1 0b q)4%0 0w q)null ` 1b q)null "" `boolean$() q)null " " 1b

• kdb+ provides values for most types to repsent Positive Infinity and Negative Infinity
• Infinities are useful in relational operations.
• Null represents a piece of unknown missing data.
• The null function returns true, if it's argument is null. However since most types in kdb+ have a null value, we can use equality to test for null, in contrast to standard SQL.

Summary

The important information to take from this tutorial is

• How to define:
  • Numeric Data
  • booleans / bytes
  • Characters
  • Date, Times, Timestamps, Months, Hours
• Accessing / Casting datetime values using dot notation
• That most types in kdb+ have nulls and infinities