The Asterix Data Model (ADM)

Table of Contents

An instance of Asterix data model (ADM) can be a primitive type (boolean, tinyint, smallint, integer, bigint, string, float, double, date, time, datetime, etc.), a special type (null or missing), or a derived type.

The type names are case-insensitive, e.g., both BIGINT and bigint are acceptable.

Primitive Types

Boolean

boolean data type can have one of the two values: true or false.

  • Example:

    { "true": true, "false": false };
    
  • The expected result is:

    { "true": true, "false": false }
    

String

string represents a sequence of characters. The total length of the sequence can be up to 2,147,483,648.

  • Example:

    { "v1": string("This is a string."), "v2": string("\"This is a quoted string\"") };
    
  • The expected result is:

    { "v1": "This is a string.", "v2": "\"This is a quoted string\"" }
    

Tinyint / Smallint / Integer (Int) / Bigint

Integer types using 8, 16, 32, or 64 bits. The ranges of these types are:

  • tinyint: -128 to 127
  • smallint: -32768 to 32767
  • integer: -2147483648 to 2147483647
  • bigint: -9223372036854775808 to 9223372036854775807

int is an abbreviated alias for integer.

  • Example:

    { "tinyint": tiny("125"), "smallint": smallint("32765"), "integer": 294967295, "bigint": bigint("1700000000000000000")};
    
  • The expected result is:

    { "tinyint": 125, "smallint": 32765, "integer": 294967295, "bigint": 1700000000000000000 }
    

Float

float represents approximate numeric data values using 4 bytes. The range of a float value can be from 2^(-149) to (2-2^(-23)·2^(127) for both positive and negative. Beyond these ranges will get INF or -INF.

  • Example:

    { "v1": float("NaN"), "v2": float("INF"), "v3": float("-INF"), "v4": float("-2013.5") };
    
  • The expected result is:

    { "v1": "NaN", "v2": "INF", "v3": "-INF", "v4": -2013.5 }
    

Double (double precision)

double represents approximate numeric data values using 8 bytes. The range of a double value can be from (2^(-1022)) to (2-2^(-52))·2^(1023) for both positive and negative. Beyond these ranges will get INF or -INF.

  • Example:

    { "v1": double("NaN"), "v2": double("INF"), "v3": double("-INF"), "v4": "-2013.593823748327284" };
    
  • The expected result is:

    { "v1": "NaN", "v2": "INF", "v3": "-INF", "v4": -2013.5938237483274 }
    

Double precision is an alias of double.

Binary

binary represents a sequence of bytes. It can be constructed from a hex or a base64 string sequence. The total length of the byte sequence can be up to 2,147,483,648.

  • Example:

    {
      "hex1" : hex("ABCDEF0123456789"),
      "hex2": hex("abcdef0123456789"),
      "base64_1" : base64("0123456789qwertyui+/"),
      "base64_2" : base64('QXN0ZXJpeA==')
    };
    
  • The default output format is in hex format. Thus, the expected result is:

    {
      "hex1": hex("ABCDEF0123456789"),
      "hex2": hex("ABCDEF0123456789"),
      "base64_1": hex("D35DB7E39EBBF3DAB07ABB72BA2FBF"),
      "base64_2": hex("41737465726978")
    }
    

Point

point is the fundamental two-dimensional building block for spatial types. It consists of two double coordinates x and y.

  • Example:

    { "v1": point("80.10d, -10E5"), "v2": point("5.10E-10d, -10E5") };
    
  • The expected result is:

    { "v1": point("80.1,-1000000.0"), "v2": point("5.1E-10,-1000000.0") }
    

Line

line consists of two points that represent the start and the end points of a line segment.

  • Example:

    { "v1": line("10.1234,11.1e-1 +10.2E-2,-11.22"), "v2": line("0.1234,-1.00e-10 +10.5E-2,-01.02") };
    
  • The expected result is:

    { "v1": line("10.1234,1.11 0.102,-11.22"), "v2": line("0.1234,-1.0E-10 0.105,-1.02") }
    

Rectangle

rectangle consists of two points that represent the bottom left and upper right corners of a rectangle.

  • Example:

    { "v1": rectangle("5.1,11.8 87.6,15.6548"), "v2": rectangle("0.1234,-1.00e-10 5.5487,0.48765") };
    
  • The expected result is:

    { "v1": rectangle("5.1,11.8 87.6,15.6548"), "v2": rectangle("0.1234,-1.0E-10 5.5487,0.48765") }
    

Circle

circle consists of one point that represents the center of the circle and a radius of type double.

  • Example:

    { "v1": circle("10.1234,11.1e-1 +10.2E-2"), "v2": circle("0.1234,-1.00e-10 +10.5E-2") };
    
  • The expected result is:

    { "v1": circle("10.1234,1.11 0.102"), "v2": circle("0.1234,-1.0E-10 0.105") }
    

Polygon

polygon consists of n points that represent the vertices of a simple closed polygon.

  • Example:

    {
      "v1": polygon("-1.2,+1.3e2 -2.14E+5,2.15 -3.5e+2,03.6 -4.6E-3,+4.81"),
      "v2": polygon("-1.0,+10.5e2 -02.15E+50,2.5 -1.0,+3.3e3 -2.50E+05,20.15 +3.5e+2,03.6 -4.60E-3,+4.75 -2,+1.0e2 -2.00E+5,20.10 30.5,03.25 -4.33E-3,+4.75")
    };
    
  • The expected result is:

    {
      "v1": polygon("-1.2,130.0 -214000.0,2.15 -350.0,3.6 -0.0046,4.81"),
      "v2": polygon("-1.0,1050.0 -2.15E50,2.5 -1.0,3300.0 -250000.0,20.15 350.0,3.6 -0.0046,4.75 -2.0,100.0 -200000.0,20.1 30.5,3.25 -0.00433,4.75") }
    }
    

Date

date represents a time point along the Gregorian calendar system specified by the year, month and day. ASTERIX supports the date from -9999-01-01 to 9999-12-31.

A date value can be represented in two formats, extended format and basic format.

  • Extended format is represented as [-]yyyy-mm-dd for year-month-day. Each field should be padded if there are less digits than the format specified.
  • Basic format is in the format of [-]yyyymmdd.
  • Example:

    { "v1": date("2013-01-01"), "v2": date("-19700101") };
    
  • The expected result is:

    { "v1": date("2013-01-01"), "v2": date("-1970-01-01") }
    

Time

time type describes the time within the range of a day. It is represented by three fields: hour, minute and second. Millisecond field is optional as the fraction of the second field. Its extended format is as hh:mm:ss[.mmm] and the basic format is hhmmss[mmm]. The value domain is from 00:00:00.000 to 23:59:59.999.

Timezone field is optional for a time value. Timezone is represented as [+|-]hh:mm for extended format or [+|-]hhmm for basic format. Note that the sign designators cannot be omitted. Z can also be used to represent the UTC local time. If no timezone information is given, it is UTC by default.

  • Example:

    { "v1": time("12:12:12.039Z"), "v2": time("000000000-0800") };
    
  • The expected result is:

    { "v1": time("12:12:12.039Z"), "v2": time("08:00:00.000Z") }
    

Datetime (Timestamp)

A datetime value is a combination of an date and time, representing a fixed time point along the Gregorian calendar system. The value is among -9999-01-01 00:00:00.000 and 9999-12-31 23:59:59.999.

A datetime value is represented as a combination of the representation of its date part and time part, separated by a separator T. Either extended or basic format can be used, and the two parts should be the same format.

Millisecond field and timezone field are optional, as specified in the time type.

  • Example:

    { "v1": datetime("2013-01-01T12:12:12.039Z"), "v2": datetime("-19700101T000000000-0800") };
    
  • The expected result is:

    { "v1": datetime("2013-01-01T12:12:12.039Z"), "v2": datetime("-1970-01-01T08:00:00.000Z") }
    

timestamp is an alias of datetime.

Duration/Year_month_duration/Day_time_duration

duration represents a duration of time. A duration value is specified by integers on at least one of the following fields: year, month, day, hour, minute, second, and millisecond.

A duration value is in the format of [-]PnYnMnDTnHnMn.mmmS. The millisecond part (as the fraction of the second field) is optional, and when no millisecond field is used, the decimal point should also be absent.

Negative durations are also supported for the arithmetic operations between time instance types (date, time and datetime), and is used to roll the time back for the given duration. For example date("2012-01-01") + duration("-P3D") will return date("2011-12-29").

There are also two sub-duration types, namely year_month_duration and day_time_duration. year_month_duration represents only the years and months of a duration, while day_time_duration represents only the day to millisecond fields. Different from the duration type, both these two subtypes are totally ordered, so they can be used for comparison and index construction.

Note that a canonical representation of the duration is always returned, regardless whether the duration is in the canonical representation or not from the user’s input. More information about canonical representation can be found from XPath dayTimeDuration Canonical Representation and yearMonthDuration Canonical Representation.

  • Example:

    { "v1": duration("P100Y12MT12M"), "v2": duration("-PT20.943S") };
    
  • The expected result is:

    { "v1": duration("P101YT12M"), "v2": duration("-PT20.943S") }
    

Interval

interval represents inclusive-exclusive ranges of time. It is defined by two time point values with the same temporal type(date, time or datetime).

  • Example:

    {
      "v1": interval(date("2013-01-01"), date("20130505")),
      "v2": interval(time("00:01:01"), time("213901049+0800")),
      "v3": interval(datetime("2013-01-01T00:01:01"), datetime("20130505T213901049+0800"))
    };
    
  • The expected result is:

    {
      "v1": interval(date("2013-01-01"), date("2013-05-05")),
      "v2": interval(time("00:01:01.000Z"), time("13:39:01.049Z")),
      "v3": interval(datetime("2013-01-01T00:01:01.000Z"), datetime("2013-05-05T13:39:01.049Z"))
    }
    

UUID

uuid represents a UUID value, which stands for Universally unique identifier. It is defined by a canonical format using hexadecimal text with inserted hyphen characters. (E.g.: 5a28ce1e-6a74-4201-9e8f-683256e5706f). This type is generally used to store auto-generated primary key values.

  • Example:

    return { "v1":uuid("5c848e5c-6b6a-498f-8452-8847a2957421") }
    
  • The expected result is:

    { "v1": uuid("5c848e5c-6b6a-498f-8452-8847a2957421") }
    

Incomplete Information Types

Null

null is a special value that is often used to represent an unknown value. For example, a user might not be able to know the value of a field and let it be null.

  • Example:

    { "field": null };
    
  • The expected result is:

    { "field": null }
    

Missing

missing indicates that a name-value pair is missing from an object. If a missing name-value pair is accessed, an empty result value is returned by the query.

As neither the data model nor the system enforces homogeneity for datasets or collections, items in a dataset or collection can be of heterogeneous types and so a field can be present in one object and missing in another.

  • Example:

    { "field": missing };
    
  • The expected result is:

    {  }
    

Since a field with value missing means the field is absent, we get an empty object.

Derived Types

Object

An object contains a set of fields, where each field is described by its name and type. An object type may be defined as either open or closed. Open objects (instances of open object types) are permitted to contain fields that are not part of the type definition, while closed objects do not permit their instances to carry extra fields. An example type definition for an object is:

    create type SoldierType as open {
        name: string?,
        rank: string,
        serialno: int
    };

Syntactically, object constructors are surrounded by curly braces “{…}”. Some examples of legitimate instances of the above type include:

    { "name": "Joe Blow", "rank": "Sergeant", "serialno": 1234567 }
    { "rank": "Private", "serialno": 9876543 }
    { "name": "Sally Forth", "rank": "Major", "serialno": 2345678, "gender": "F" }

The first instance has all of the type’s prescribed content. The second instance is missing the name field, which is fine because it is optional (due to the ?). The third instance has an extra field; that is fine because the type definition specifies that it is open (which is also true by default, if open is not specified). To more tightly control object content, specifying closed instead of open in the type definition for SoldierType would have made the third example instance an invalid instance of the type.

Array

An array is a container that holds a fixed number of values. Array constructors are denoted by brackets: “[…]”.

An example would be

    ["alice", 123, "bob", null]

Multiset

A multiset is a generalization of the concept of a set that, unlike a set, allows multiple instances of the multiset’s elements. Multiset constructors are denoted by two opening curly braces followed by data and two closing curly braces, like “{{…}}”.

An example would be

    {{"hello", 9328, "world", [1, 2, null]}}