Custom, Cypher Based Procedures and Functions

All these procedures (except the list and show ones) are intended to be executed in the system database, therefore they have to be used executed by opening a system database session. There are several ways of doing this: - when using cypher-shell or Neo4j Browser, one can prefix their Cypher query with :use system - when using fabric, one can prefix their Cypher query with USE system - when using the drivers, one can open a session directly against the system database

Moreover, they accept as first parameter the name of the database towards which we want to install/update/remove the automatic UUIDs. Through this implementation, we can use these procedures in a cluster environment, by leveraging the cluster routing mechanism.

Installing, updating or removing a custom Cypher statement is an eventually consistent operation. Therefore, they are not immediately added/updated/removed, but they have a refresh rate handled by the Apoc configuration apoc.custom.procedures.refresh=<MILLISECONDS>.

In case of a cluster environment, the apoc.custom.procedures.refresh also replicate the procedures/functions to each cluster member.

Table 1. Available procedures
Qualified Name Type Release

apoc.custom.dropAll

apoc.custom.dropAll() - Eventually drops all previously added custom procedures/functions and returns info

Procedure

Apoc Extended

apoc.custom.dropFunction

apoc.custom.dropFunction() - Eventually drops the targeted custom function

Procedure

Apoc Extended

apoc.custom.dropProcedure

apoc.custom.dropProcedure() - Eventually drops the targeted custom procedure

Procedure

Apoc Extended

apoc.custom.installFunction

apoc.custom.installFunction() - Eventually registers a custom cypher function

Procedure

Apoc Extended

apoc.custom.installProcedure

apoc.custom.installProcedure() - Eventually registers a custom cypher procedure

Procedure

Apoc Extended

apoc.custom.list

apoc.custom.list() - provide a list of custom procedures/function registered

Procedure

Apoc Extended

apoc.custom.show

apoc.custom.show() - Provides a list of custom procedures/function registered

Procedure

Apoc Extended

Overview

I wanted for a long time to be able to register Cypher statements as proper procedures and functions, so that they become callable in a standalone way.

You can achieve that with the apoc.custom.installProcedure and apoc.custom.installFunction procedure calls. Those register a given Cypher statement, prefixed with the custom.* namespace, overriding potentially existing ones, so you can redefine them as needed.

The first parameter of the apoc.custom.installProcedure and apoc.custom.installFunction procedures, is the signature of the procedure/function you want to create.

This looks similar to the signature results returned by the SHOW PROCEDURES YIELD signature, SHOW FUNCTIONS YIELD signature cypher commands, or by the CALL apoc.help('<fun_or_procedure_name>') YIELD signature procedure, just without the ?. That is: - for a procedure: nameProcedure(firstParam = defaultValue :: typeParam , secondParam = defaultValue :: typeParam, …​.) :: (firstResult :: typeResult, secondResult :: typeResult, …​ ) - for a function: nameFunction(firstParam = defaultValue :: typeParam , secondParam = defaultValue :: typeParam, …​.) :: typeResult

Note that, for both procedures and functions, the = defaultValue are optionals. The default values are parsed as JSON.

If you want to create a procedure/function with a default String parameter with whitespaces, quotes (for example: "my text ' with ' quote") or "null" (as a string), you have to quote the result, e.g CALL apoc.custom.declareProcedure("procWithNullString(param='null'::STRING)::(output::STRING)", 'return $param as output')
Type Names

The typeParam and typeResult in the signature parameter can be one of the following values:

  • FLOAT, DOUBLE, INT, INTEGER, INTEGER | FLOAT, NUMBER, LONG

  • TEXT, STRING

  • BOOL, BOOLEAN

  • POINT, GEO, GEOMETRY

  • DATE, DATETIME, LOCALDATETIME, TIME, LOCALTIME, DURATION

  • NODE, REL, RELATIONSHIP, EDGE, PATH

  • MAP

  • MAPRESULT (valid for declareFunction, does not wrap the result in a further map. see here: [map-vs-map-result])

  • LIST TYPE, LIST OF TYPE (where TYPE can be one of the previous values)

  • ANY

In Neo4j 5.13, the NUMBER type was replaced by INTEGER | FLOAT, but we can still use it for backwards compatibility.
If you override procedures or functions you might need to call call db.clearQueryCaches() as lookups to internal id’s are kept in compiled query plans.

Starting from version 5.11, if we execute a CALL apoc.custom.installFunction('nameFun(…​.)'), or a CALL apoc.custom.installProcedure('nameProc(…​.)'), we cannot execute respectively a RETURN custom.nameFun(..) or a CALL custom.nameProc() within the same transaction, an error will be thrown.

So we must necessarily open a new transaction to execute the custom procedures / functions declared.

The following examples assume that we are on the neo4j database, and we want to create the custom procedures/function in that database.

Custom Procedures with apoc.custom.installProcedure

Procedure, input and output names must have at least 2 characters.

Here is a simple example:

CALL apoc.custom.installProcedure('answerInteger() :: (answer::INT)', 'RETURN 42 as answer')

This registers, after a time defined by the configuration apoc.custom.procedures.refresh, a procedure custom.answer that you then can call.

CALL custom.answerInteger
Table 2. Results
answer

42

Or you can also write in this way:

CALL apoc.custom.installProcedure('answer() :: (row::MAP)', 'RETURN 42 as answer')

In this case the result is wrapped in a stream of maps called row. Therefore, you can do:

CALL custom.answer()
YIELD row
RETURN row.answer
Table 3. Results
answer

42

We can create the procedure custom.powers that returns a stream of the powers of the first parameter, up to and including the power provided by the second parameter:

CALL apoc.custom.installProcedure(
  'powers(input::INT, power::INT) :: (answer::INT)',
  'UNWIND range(0, $power) AS power
   RETURN $input ^ power AS answer'
);

We can use this function, to return 4°, 4¹, 4², and 4³, as shown in the query below:

CALL custom.powers(4,3);
Table 4. Results
answer

1.0

4.0

16.0

64.0

Moreover, via the 3rd parameter, we can create the custom procedure in a different database than neo4j. For example, the following one creates a custom procedure in the foo database:

CALL apoc.custom.installProcedure('foodb() :: (row::INT)', 'RETURN 42 as row', 'foo')

Furthermore, we can pass as the 4th parameter a string to specify the procedure mode (default "WRITE"). It can be: - "READ" - if the procedure will only perform read operations against the graph - "WRITE" - if it may perform both read and write operations against the graph - "SCHEMA" - if it will perform operations against the schema - "DBMS" - if it will perform system operations - i.e. not against the graph

Also, we can pass a description parameter as the 5th parameter, which will be returned by the call apoc.custom.list and SHOW PROCEDURES.

Custom Functions with apoc.custom.installFunction

Function, input and output names must have at least 2 characters.

Here is a simple example:

CALL apoc.custom.installFunction('answerFun() :: INT', 'RETURN 42 as answer')

This registers, after a time defined by the configuration apoc.custom.procedures.refresh, the statement as procedure custom.answer that you then can call.

RETURN custom.answerFun()
Table 5. Results
answer

42

Or you can also write in this way:

CALL apoc.custom.installFunction('answerFunMap() :: MAP', 'RETURN 42 as answer')

In this case the result is wrapped in a stream of maps called row. Therefore, you can do:

WITH custom.answerFunMap() YIELD row
RETURN row.answer
Table 6. Results
answer

42

We can create the function custom.double, that doubles the provided value, by running the following function:

CALL apoc.custom.installFunction(
  'double(input::INT) :: INT',
  'RETURN $input*2 as answer'
);

We can use this function, as shown in the query below:

RETURN custom.double(83) AS value;
Table 7. Results
value

166

Moreover, via the 3rd parameter, we can create the custom procedure in a different database than neo4j. For example, the following one creates a custom procedure in the foo database:

CALL apoc.custom.installFunction('foodb() :: INT', 'RETURN 42', 'foo')

Furthermore, we can pass as a 4th parameter a boolean (with default false) which, if true, in case the function returns a list of a single element, it will return only the single element itself and not the list.

For example:

CALL apoc.custom.installFunction('forceSingleTrue(input::ANY) :: LIST OF INT',
  'RETURN 1',
  true
);
Table 8. Results
value

1

otherwise with false the result will be a singleton list:

CALL apoc.custom.installFunction('forceSingleFalse(input::ANY) :: LIST OF INT',
  'RETURN 1',
  false
);
Table 9. Results
value

[1]

Also, we can pass a description parameter as the 5th parameter, which will be returned by the call apoc.custom.list and SHOW FUNCTIONS.

List of registered procedures/function with apoc.custom.list

The procedure apoc.custom.list provide a list of all registered procedures/function via apoc.custom.installProcedure and apoc.custom.installFunction

Given this call:

CALL apoc.custom.list

The output will look like the following table:

type name description mode statement inputs outputs forceSingle

"function"

"answer"

<null>

<null>

"RETURN $input as answer"

[["input","integer | float"]]

"long"

false

"procedure"

"answer"

"Procedure that answer to the Ultimate Question of Life, the Universe, and Everything"

"read"

"RETURN $input as answer"

[["input","int","42"]]

[["answer","integer | float"]]

<null>

Remove a procedure apoc.custom.dropProcedure

The procedure apoc.custom.dropProcedure allows to delete the targeted custom procedure, from a specific database (with neo4j as a default), after a time defined by the configuration apoc.custom.procedures.refresh.

Given this call:

CALL apoc.custom.dropProcedure(<name>, <databaseName>)

Fields:

argument description

name

the procedure name

databaseName

the database name (default: neo4j)

Remove a procedure apoc.custom.dropFunction

The procedure apoc.custom.dropFunction allows to delete the targeted custom function, from a specific database (with neo4j as a default), after a time defined by the configuration apoc.custom.procedures.refresh.

Given this call:

CALL apoc.custom.dropFunction(<name>)

Fields:

argument description

name

the function name

databaseName

the database name (default: neo4j)

Export metadata

To import custom procedures in another database (for example after a ./neo4j-admin backup and /neo4j-admin restore), please see the apoc.systemdb.export.metadata procedure.