Skip to main content

Developing Record UDFs

A Record UDF is a User-Defined Function that is applied to a single record. Record UDFs can be used to augment both read and write behavior.

Typical uses for Record-Based UDFs

  • Record UDFs can be used to implement an atomic operation that does not currently exist in the client. As a UDF will not perform at the same speed and scale as native operations, it is important to consider whether the rich

    List and Map API methods cannot achieve the same effect. Combining multiple native operations through the `operate()` command should also be considered, before using a UDF.
  • Record UDFs can be used for single-record, multi-operation transactions where data from one bin is used to determine the value of another bin.

  • Use record UDFs to extend the existing list of predicate filtering operators for scan and query.

  • Background UDFs are good for large scale maintenance and data normalization, with a record UDF applied to all the records returned by a scan or query.

General Comments on Record UDFs:

  • The first argument of the function refers to the database record (e.g. rec). Do not name the variable record as this is one of the types in the Aerospike Lua API.
  • Each subsequent argument is specific to the UDF, and must be one of the types supported by the database: numeric (integer or double), string, list or map.
  • A Record UDF should have one or more parameters (the record and optionally more). If a UDF has N parameters, and only (N-k) arguments passed in, the last k will automatically be assigned a value of nil.
  • A Record UDF may return one of the types supported by the database: numeric (integer or double), string, bytes, list or map.
  • As of server version 4.7, metadata predicate filters may be applied to Record UDF transactions.
  • A background UDF, where a single User-Defined Function is applied by scan or query, modifies records without returning a result (write only).

Record TTL and UDF

When creating or updating a record, its new TTL value adheres to the following hierarchy:

Example: Record Create or Update

In this simple example, Annotated Record UDF Example, we show how the UDF can create or update an Aerospike record.

Example: String Slice Operator

Aerospike has two atomic operations for the String data type - append and prepend. The following record UDF adds an atomic string slice operator.

function slice(rec, bin, a, b)
local s = rec[bin]
if type(s) == 'string' then
return s:sub(a, b)
return nil

In aql

aql> register module 'util.lua'
OK, 1 module added.

aql> insert into (PK, x) values ('1', "Alright I like the beat except the snare, kick and keys")
OK, 1 record affected.

aql> execute util.slice('x', 9, 23) on where PK='1'
| slice |
| "I like the beat" |
1 row in set (0.001 secs)

Example: Using Protobuf

In this example, a UDF makes use of an external Lua module, which in turns calls a C shared object. The Protobuf Module Example gives more details about the code involved in this example.

aql> select * from where PK='1'

aql> execute pbuf.vartix(123, 'JDoe', '') on where PK='1'
| vartix |
| |
1 row in set (0.001 secs)

aql> select * from where PK='1'
| person |
| 12 03 57 69 6C 08 7B 1A 16 77 69 6C 2E 6A 61 6D 69 65 73 6F 6E 40 67 6D 61 69 6C 2E 63 6F 6D |
1 row in set (0.000 secs)

aql> execute pbuf.velkor() on where PK='1'
| velkor |
| 123 |
1 row in set (0.000 secs)

Example: Using a Background UDF to Reset TTLs

The Background UDF Example resets the TTL of records accidentally set to never expire.