Skip to content

SCIM custom scripts#

With SCIM scripts, custom business logic can be executed when several of the SCIM API operations are invoked. This is useful in many situations, for example:

  • Trigger notifications to external systems when certain resources are deleted
  • Alter the actual responses the service produces
  • Attach additional information to resources when they are created or updated
  • Implement a fine-grained level of access to resources so different callers have permission to handle only a restricted number of resources

Notes:

  • In this document, the term resources refer to those "entities" the service can manage, for instance, users or groups
  • The term operation refers to any SCIM functionality accessible through its HTTP endpoints
  • Basic development skills are assumed. Some grasp of Java and Python are required as well as understanding of the SCIM protocol.

API overview#

Custom scripts adhere to a simple API (ie. a well-defined collection of methods/routines) that is described in the following. It is advised to check the dummy script provided here as you read this section.

Scripts' config properties#

All methods contain a configurationAttributes parameter, this gives access to the configuration properties of the script itself. This is a relevant aspect of Gluu scripts: they are all parameterizable!.

configurationAttributes is a java.util.Map<String, SimpleCustomProperty> and here is how SimpleCustomProperty looks.

Basic methods#

These are methods not related to SCIM operations but still play key roles:

Method name Description Return value
init Called when the (SCIM) service starts and every time the script properties or code changes A boolean value describing success or failure
destroy Called every time the script properties or code changes (called before init) A boolean value describing success or failure
getApiVersion Determines what methods are effectively called when SCIM endpoints are invoked A positive integer

Pre-resource modification#

They are called when the resource is about to be persisted. The second parameter in these methods hold the object that will be persisted to permanent storage, thus any change or manipulation upon the object will be reflected in the underlying database (as well as in the output of the SCIM operation call itself).

These methods are called regardless of the API version used. Names are self explanatory:

Methods 2nd param 2nd param Class/Link
createUser, updateUser, deleteUser user ScimCustomPerson
createGroup, updateGroup, deleteGroup group GluuGroup

Pre-resource modification methods return a boolean. A False value aborts the corresponding SCIM operation and a 500 error is returned. The same applies if the method execution crashes at runtime.

Note that update* methods are called for both SCIM PUT and PATCH operations.

Post-resource modification#

They are called after the resource is persisted. The second parameter in these methods hold the object that was saved. Any change or manipulation upon the object will not be reflected in the underlying database, but may still modify the service response.

These methods are called if getApiVersion returns a number >= 2.

Methods 2nd param 2nd param Class/Link
postCreateUser, postUpdateUser, postDeleteUser user ScimCustomPerson
postCreateGroup, postUpdateGroup, postDeleteGroup group GluuGroup

Post-resource modification methods return a boolean. A False value aborts the corresponding SCIM operation and a 500 error is returned. The same applies if the method execution crashes at runtime.

Note that postUpdate* methods are called for both SCIM PUT and PATCH operations.

Single resource retrieval#

These apply for SCIM operations that retrieve a resource by ID. They are called after the resource has been obtained from the database. The second parameter in these methods hold a reference to such object.

Any change or manipulation upon the object will not be reflected in the underlying database, but may still modify the service response.

These methods are called if getApiVersion returns a number >= 3 (available in Gluu 4.1 onwards).

Methods 2nd param 2nd param Class/Link
getUser user ScimCustomPerson
getGroup group GluuGroup

Single resource retrieval methods return a boolean. A False value aborts the whole SCIM operation and a 500 error is returned. The same applies if the method execution crashes at runtime.

Multiple resources retrieval#

These apply for SCIM search operations. They are called after the results have been obtained from the database. The second parameter in these methods hold a reference to such result set.

Any change or manipulation upon the object will not be reflected in the underlying database, but may still modify the service response.

These methods are called if getApiVersion returns a number >= 4 (available in Gluu 4.2 onwards).

Methods 2nd param 2nd param Class/Link
postSearchUsers results PagedResult
postSearchGroups results PagedResult

Multiple resources retrieval methods return a boolean. A False value aborts the whole SCIM operation and a 500 error is returned. The same applies if the method execution crashes at runtime.

Note that searching using the root .search SCIM endpoint will trigger calls to both of the methods listed.

Advanced control#

These are alternative methods that allow to tweak the response the service produces. They can be employed to introduce complex business rules when operations are executed.

These methods are called if getApiVersion returns a number >= 5 (available in Gluu 4.3 onwards).

Methods
manageResourceOperation
manageSearchOperation

Later we'll revisit the topic of access and provide concrete examples.

Developer's first steps#

As with any other custom script in Gluu, SCIM scripts are run by the Jython engine. This allows usage of Python-style coding and leverages all the Java classes available in the SCIM web application classpath.

To start, ensure SCIM service is running and a protection mode has been configured. Then in oxTrust enable the sample script:

  1. Navigate to Configuration > Other custom Scripts > SCIM

  2. Expand the only row (labeled scim_event_handler)

  3. Click on the Enabled check box

  4. Click on update at the bottom of the page

SCIM

Inspect (tail) scim_script.log. After some seconds you'll see some lines related to script initialization, ie. method init being called. This means your script is active and properly running. Click here to learn more about SCIM logs.

Learning by doing is a good approach to scripting in Gluu. To start, edit the script by adding some print statements to the following methods: createUser, postCreateUser, manageResourceOperation. Save the script and check the log, wait until destroy and init are called.

Send a user creation request to the service. Here is an example. Note the order in which your prints appear in the log.

Delete the user created (whether via oxTrust or SCIM itself). Edit the return value for each of the three methods one by one and issue a creation request after every script update. Inspect the log output as well as the service outputs, namely, the HTTP responses.

Now that you have got some acquaintance with the edit/test cycle, we can proceed with an example.

Example: modifying search results#

SCIM spec defines the concept of attribute returnability where some attributes should be never be part of a response (like passwords), always be returned (like resource identifiers), or be returned by default unless otherwise stated by the excludedAttributes parameter.

Assume you are maintaining a user base of secret agents that work for your company and need to avoid exposing information such as their physical addresses for safety reasons. To keep it simple let's restrict the scope to user searches only. In practice you should take steps to hide this data on user retrieval and update.

Let's alter postSearchUsers's second parameter (results) to ensure addresses are not leaked:

for user in results.getEntries():
    user.setAttribute("oxTrustAddresses", None)

This is very straightforward code except for the usage of oxTrustAddresses. Shouldn't it be simply addresses as the known SCIM attribute?

Scripts work with entities that are about to be persisted or have already been saved so they kind of resemble the database structure (schema in LDAP terms). It turns out that database attribute names rarely match with SCIM names. An explanation of this fact can be found here.

While it is easy to know the SCIM name of a database attribute, the converse requires checking the code, however since you already have the skill this shouldn't be a problem: in this Java class you'll find the representation of a user resource in SCIM spec terms. Pay attention to the addresses field and its associated StoreReference annotation that contains the attribute where addresses are actually stored.

With that said, save your modifications. You may like the idea of adding some prints for enlightment like:

print "%d entries returned of %d" % (results.getEntriesCount(), results.getTotalEntriesCount())
for user in results.getEntries():
    print "Flushing addresses for user %s" % getUid() 
    user.setAttribute("oxTrustAddresses", None)

Ensure no addresses are returned anymore in your SCIM user searches. Happy testing!

Controlling execution of SCIM operations#

With the manageResourceOperation and manageSearchOperation methods you can make complex decisions on how processing should take place based on contextual data and the incoming payload.

Important

Consider the incoming payload as a read-only variable. Modifying is value may produce unexpected results

manageResourceOperation#

This method is invoked when any following operations are executed: resource creation, modification, removal and retrieval by ID. In case of bulks, the method is called for every operation that fits into these categories.

Parameters are described in the table below:

Name Description Class/Link
context Provides contextual information about the SCIM operation being called such as type of resource involved, HTTP verb, request headers, query params, etc. OperationContext
entity An non-null object representing the resource involved A descendant of Entry. If the resource is a user, it will be an instance of ScimCustomPerson. In case of a group, it will be a GluuGroup
payload The payload sent in the invocation; null when the operation is removal or retrieval by ID. Altering the payload is discouraged The datatype depends on the operation called. Check the interface that suits best and inspect the first parameter's datatype. The class will belong to some subpackage inside org.gluu.oxtrust.model.scim2

This method is expected to return an instance of javax.ws.rs.core.Response that supersedes the output of the operation itself. In other words, the actual processing of the operation is skipped in favor of the code supplied here. However note that minor validations may take place in the payload before your code is actually called.

Returning None transfers the control of the operation for normal processing (default Gluu implementation). If the method execution crashes at runtime, a 500 HTTP error is sent.

Notes:

  • Possible values for context.getResourceType() are: User, Group, FidoDevice, Fido2Device
  • context.getTokenDetails().getValue() is a shortcut that will give you the access token the caller employed to issue the service call
  • Both context.getTokenDetails().getTokenType() and context.getTokenDetails().getScope() return non null values when the protection mechanism of the API is OAuth or test mode
  • Note that for resource creation operation, entity basically contains the same data supplied in the POST payload. In this case, entity has not originated from the database and has not been persisted either
  • For the case of modification, retrieval and removal, entity contains the data currently stored in the database for the resource in question
  • Since many values come from Java code, you can always do getClass().getName() to get an idea of what type of variables you are dealing with
  • To build custom error responses your can reuse some of the getErrorResponse methods of class BaseScimWebService

This method offers a high degree of flexibility. Perform careful testing of your code and account all potential scenarios.

manageSearchOperation#

This method is invoked when resource searches are performed. Parameters are described in the table below:

Name Description Class/Link
context Provides contextual information about the SCIM operation being called such as type of resource involved, HTTP verb, request headers, query params, etc. OperationContext
searchRequest An object representing the search parameters provided in the call (applies for both GET and POST) SearchRequest

Unlike manageResourceOperation, no entity parameter is passed. This is so because making decisions based on already executed searches would have a performance impact. Instead you can use context.setFilterPrepend(...) to help restrict the search against the database: here you can pass a String value that will be interpreted as an SCIM filter expression (see section 3.4.2.2 of RFC 7644). When the search being performed already contains a search filter (ie. searchRequest.getFilter()is non-empty), a new filter is created by appending both "subfilters" with an and operator.

As in the case of manageResourceOperation this method is expected to return an instance of javax.ws.rs.core.Response.

Returning None transfers the control of the operation for normal processing (the Gluu implementation). If the method execution crashes at runtime, a 500 HTTP error is sent.

The same recommendations given for manageResourceOperation apply here. If you build filter expressions in your method, ensure they are syntactically valid to avoid your callers getting unexpected "invalidFilter" 400 errors.

Example: segmenting the user base#

Let's assume you make use of the SCIM attribute userType so that your user base is partitioned into three disjoint segments according to such attribute and you have designated its possible values to be: Contractor, Employee, or Intern.

Suppose your company has three different applications that make management of users in every category, that is, only one application is devoted to manage contractors, another employees, and other interns. You are interested in consistently granting access so that no application can create, query or modify users that don't belong to its focus.

To properly handle this multi-tenancy scenario, you decide to use the contextual information coming from every request to determine if the operation should be allowed or not. For this purpose you communicate every application developer to send an additional HTTP header in their call with a value that only you (the administrator) and the developer knows. Let's call it the "secret". For the sake of simplicity let's assume developers are external to the company and only you know their identities. They don't know each other so they cannot exchange secrets.

The strategy to implement segmentation is rather simple:

  • Alter the default SCIM script by supplying a custom implementation for the methods that control execution

  • Make the HTTP header name be a configuration property of the script so that it is not hard-coded

  • Add a configuration property that contains the mapping of userType value vs. expected header value in JSON format

Adding and parsing config properties#

In oxTrust expand the SCIM script and add properties:

  • custom_header with value USER-SEGMENT-SECRET
  • access_map with value { "<random_string>":"Contractor", "<random_string>":"Employee", "<random_string>":"Intern" }

Save the changes.

In the init method this properties should be parsed. To start, in the script context textarea let's add some imports:

from org.gluu.oxtrust.ws.rs.scim2 import BaseScimWebService
import json
import sys

Here is how init would look like:

def init(self, configurationAttributes):
    self.custom_header = configurationAttributes.get("custom_header").getValue2()
    json = configurationAttributes.get("access_map").getValue2()    
    self.access_map = json.loads(json)
    print "ScimEventHandler (init): Initialized successfully"
    return True

Note no validations took place here: we assumed the script contains the properties, that they are non empty and have sensible values.

Allow/Deny resource operations#

The first step is to know the kind of application that is calling our service. For this purpose let's create a method that given incoming request headers returns the matching userType

# headers params is an instance of javax.ws.rs.core.MultivaluedMap<String, String>
def getUserType(self, headers):
    secret = headers.getFirst(self.custom_header)
    if secret in self.access_map:
       return self.access_map[secret]
    else:
       return None

Now let's code manageResourceOperation. We should allow access only under the following conditions:

  • The getUserType method does not return None
  • The entity object (ScimCustomPerson instance) has a proper userType value. This means that for user creation, the incoming payload comes with a matching userType and for the other cases, the already stored attribute matches as well

Assume that if the operation invoked is not user-related, we should allow access freely. Here is how the implementation might look:

def manageResourceOperation(self, context, entity, payload, configurationAttributes):

    print "manageResourceOperation. SCIM endpoint invoked is %s (HTTP %s)" % (context.getPath(), context.getMethod()) 
    if context.getResourceType() != "User":
        return None

    expected_user_type = self.getUserType(context.getRequestHeaders())

    if expected_user_type != None and entity.getAttribute("oxTrustUserType") == expected_user_type:
        return None
    else:
        return BaseScimWebService.getErrorResponse(403, None, "Attempt to handle a not allowed user type")

Note no usage of the payload took place. A case you may like to evaluate is where mistakenly using an update operation, the userType is set to an unexpected value.

Allow/Deny searches#

This time instead of inspecting an entity, we ought to make a filter expression to restrict the search when the database is queried. For your reference, a valid filter expression is for instance userType eq "Contractor".

def manageSearchOperation(self, context, searchRequest, configurationAttributes):

    print "manageSearchOperation. SCIM endpoint invoked is %s (HTTP %s)" % (context.getPath(), context.getMethod())

    resource_type = context.getResourceType()
    print "manageSearchOperation. This is a search over %s resources" % resource_type

    if resource_type != "User":
        return None

    expected_user_type = self.getUserType(context.getRequestHeaders())

    if expected_user_type != None:
        context.setFilterPrepend("userType eq \"%s\"" % expected_user_type)
        return None
    else:
        return BaseScimWebService.getErrorResponse(403, None, "Attempt to handle a not allowed user type")

Recall this method must return a javax.ws.rs.core.Response. A None value makes continue the operation processing normally.

Working with more than one script#

You may have already noticed that in oxTrust it is possible to have several scripts under the SCIM tab. This is how execution takes place when there are several scripts enabled:

The applicable method is called in the first script. If the return value was True, the method is called again but this time in the subsequent script. If at any point a False return value is encountered, the SCIM operation is aborted with error 500. This means that a normal operation execution requires all involved methods across different scripts to be successful.

There is an important exception to the above and is related to the manage* methods. In this case, only one script takes effect (the first script found). Note that in most cases having a single SCIM script suffices for all needs.