Data Types


      This section describes the data types which Couchbase Analytics can operate on.

      An instance of Couchbase Analytics data model can be a primitive type (BOOLEAN, STRING, BIGINT, or DOUBLE), a special type (NULL or MISSING), or a composite type (OBJECT, ARRAY, or MULTISET).

      The type names are case-insensitive; for example, both BIGINT and bigint are acceptable.

      Primitive Types


      The boolean data type has only two values: true and false.


      An expression that compares two values, such as weight > limit, might return a boolean value, or it might return one of the special values null or missing. For details, see Comparison Operators.


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

      A literal string can be enclosed in single or double quotes.


      "This is a string."
      'Have you read "War and Peace"?'
      "I don't think so."
      For more details on string literals, see Literals.


      A bigint is a 64-bit integer. The range of bigint is from -9223372036854775808 to 9223372036854775807.



      In general, arithmetic operations on two bigint values return a bigint result. However, two kinds of division operators are provided:

      • 5 / 2 returns 2.5, a value of type double.

      • 5 div 2 returns 2, a value of type bigint, by truncating the result.

      Double (double precision)

      The double type 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 numbers. Expressions that compute values outside these ranges will return INF or -INF. The special value NaN (not a number) is also supported.

      A numeric literal is treated as a double if it contains a decimal point or an exponent.


      All numbers in the incoming JSON which are not integers are parsed as double values.

      Temporal Data

      Temporal data (dates, times, and timestamps) are important for many applications. Since JSON does not provide data types for temporal data, SQL++ for Analytics uses the string and bigint types for this purpose. SQL++ for Analytics provides many builtin functions, called temporal functions, that operate on string or bigint values, interpreting them as dates or times. For a list of these functions, see Temporal Functions.

      Using bigint for Temporal Data

      Some temporal functions take bigint values and interpret them as representing a number of milliseconds since the beginning of the "Unix Epoch" at 00:00:00 UTC on January 1, 1970. Here are two of these temporal functions and some examples of their use.

      • now_millis( ) returns the current point in time as a bigint representing Unix Epoch milliseconds.

      • date_diff_millis(t1, t2, unit) returns the difference between two points in time (t1 and t2), in terms of the designated units (years, months, weeks, days, hours, seconds, or milliseconds).


      • now_millis( ) returns 1655230396621 at the start of an experiment.

      • now_millis( ) returns 1655231924715 at the end of the experiment.

      • date_diff_millis(1655231924715, 1655230396621, "minute") returns 25, showing that the experiment lasted for 25 minutes.

      • date_diff_millis(1655231924715, 1655230396621, "second") returns 1528, providing a finer measurement: the experiment lasted for 1528 seconds.

      Using string for Temporal Data

      A string in ISO-8601 format can represent a date, time, or timestamp. The ISO format is illustrated by the following example, which represents the 30.55 seconds after 15:10 (3:10pm) on June 14, 2022 in a timezone that (like San Francisco) is eight hours behind Coordinated Universal Time (UTC).


      When using a string to represent temporal data, various parts of the ISO format can be omitted. For example:

      • "2022-06-14" represents the day June 14, 2022.

      • "15:10:30" represents a time 30 seconds past 15:10 (3:10pm).

      In general, two temporal values can be compared only if they have the same components, for example two dates or two times. Violations of this rule usually give null results. These rules are illustrated by the following examples, which use the date_diff_str function to compare two temporal values and return the difference in a designated unit:


      • date_diff_str("2022-06-22", "2022-06-15", "day") returns 7.

      • date_diff_str("09:20:00", "08:00:00", "hour") returns 1.

      • date_diff_str("09:20:00", "08:00:00", "minute") returns 80.

      • date_diff_str("2022-06-22", "10:00:00", "hour") returns null.

      For more information about the ISO-8601 format for temporal data, see Date Formats.

      Incomplete Information Types


      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 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.

      A missing value is converted to null when converted to JSON in the query results.

      Composite Types


      An object (a.k.a., JSON object) contains a set of fields, where each field is described by its name.

      Syntactically, object constructors are surrounded by curly braces "{…​}". For example:

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


      An array is a container that holds a fixed number of values. An array is an ordered collection of items. Array constructors are denoted by brackets: "[…​]".

      An example would be:

      ["alice", 123, "bob", null]
      An array can appear in the incoming JSON and can be constructed by the query. Each SELECT statement with an ORDER BY clause returns an array.


      A multiset is an unordered collection of items. A multiset allows multiple instances of its elements.

      A multiset cannot appear in the incoming JSON. A multiset is converted into a JSON array with an undefined order of elements in the query results. The order of the items in the result might change from one query execution to another. Each SELECT statement without an ORDER BY clause returns a multiset.