The items Correlator

Last modified 18 Oct 2022 14:14 +02:00
Since 4.6
This functionality is available since version 4.6.
Table of Contents

Introduction

In earlier versions of midPoint, the only correlation mechanism available was the filter-based one. The administrator had to specify a query based (typically) on the values of the resource object and executed it in order to find its owner.

The newly-introduced items correlator makes this process much easier by constructing the filter(s) automatically from the correlator configuration.

Simple Examples

Here is the simplest configuration. It matches users based on the value of employeeNumber property. (We assume that the value of the property for accounts being correlated is determined by correlation-time evaluation of inbound mappings, as described in the overview document.)

Listing 1. Matching on employeeNumber
<correlators>
    <items>
        <item>
            <ref>employeeNumber</ref>
        </item>
    </items>
</correlators>

The following configuration matches on both givenName and familyName values, requiring that both do match.

Listing 2. Matching on both given and family name
<correlators>
    <items>
        <item>
            <ref>givenName</ref>
        </item>
        <item>
            <ref>familyName</ref>
        </item>
    </items>
</correlators>

Sometimes, we need to fine-tune the search based on individual items. There are the following configuration options available:

Table 1. Configuration options for search fine-tuning
Option Meaning Default

index

What index (normalization) to use when matching this item?

Default index (for indexed items) or original value (otherwise).

matchingRule

What matching rule to use? (For non-fuzzy search.)

Rule of mr:default (for indexed items; note they are normalized using polyStringNorm algorithm by default); or the rule specified in the object template for given item (for non-indexed items).

fuzzy

Specification of the fuzzy search to be used.

confidence

Expression for computing the resulting confidence value. The confidence values for individual items are multiplied to provide the overall value for given candidate.

1 for non-fuzzy search

1/(d+1) for Levenshtein distance of d

s for trigram similarity of s

The matchingRule and fuzzy elements are mutually exclusive.

Specifying the Index to Use

It is possible to provide the name of the normalization (index) to be used when correlating by the specified item by the index option. (Usually, it is not necessary: either there is only one normalization defined for the particular item, or there is one marked as the default. But there are cases when it is needed.)

An example:

Listing 3. Specifying the normalization and then using it for correlation
<objectTemplate>
    ...
    <item>
        <ref>extension/nationalId</ref>
        <indexing>
            <normalization>
                <default>true</default>
                ... (a normalization) ...
            </normalization>
            <normalization>
                <name>digits</name> (1)
                <steps>
                    <custom>
                        <expression>
                            <script>
                                <code>
                                    basic.stringify(input).replaceAll("[^\\d]", "")
                                </code>
                            </script>
                        </expression>
                    </custom>
                </steps>
            </normalization>
        </indexing>
    </item>
    ...
    <correlation>
        <correlators>
            <items>
                <item>
                    <ref>extension/nationalId</ref>
                    <search>
                        <index>digits</index> (2)
                    </search>
                </item>
            </items>
        </correlators>
    </correlation>
</objectTemplate>
1 Declares the name of digits for the particular normalization.
2 Specifies that the digits normalization should be used in this correlation.

Specifying Matching Rule

It may be useful to specify the matching rule to be applied when searching for the item value. This is especially true when dealing with values that are not indexed. (Although it is possible to specify matching rules also for indexed items.)

For example, we may want to specify that when correlating on employeeNumber we want to use mr:polyStringNorm matching rule:

Listing 4. Specifying the matching rule used when correlating
<objectTemplate>
    ...
    <correlation>
        <correlators>
            <items>
                <item>
                    <ref>employeeNumber</ref>
                    <search>
                        <matchingRule>mr:polyStringNorm</matchingRule>
                    </search>
                </item>
            </items>
        </correlators>
    </correlation>
</objectTemplate>
It is possible to define a default matching rule attached to a focus item as such - by providing matchingRule property right on the definition of the item. However, this definition is currently not used for correlation search. It may change in the future.

The fuzzy search can be specified for both custom-indexed and regular items.

An example:

Listing 4. Specifying the fuzzy search used when correlating
<objectTemplate>
    ...
    <correlation>
        <correlators>
            <items>
                <item>
                    <ref>familyName</ref>
                    <search>
                        <fuzzy>
                            <levenshtein>
                                <threshold>3</threshold>
                            </levenshtein>
                        </fuzzy>
                    </search>
                </item>
            </items>
        </correlators>
    </correlation>
</objectTemplate>

For more information please see fuzzy searching document. It contains also information on specifying custom confidence values, as these are the most useful with the connection with the fuzzy search logic.

Default Search Options for an Item

It is also possible to specify correlation search options that are to be applied when given item is used in any items correlator.

For example, let us specify the default search options used when correlating by givenName.

Listing 5. Specifying the fuzzy search used when correlating
<objectTemplate>
    ...
    <item>
        <ref>givenName</ref>
        <indexing> ... </indexing>
        <correlation>
            <search> (1)
                <index>name-normalization</index>
                <fuzzy>
                    <levenshtein>
                        <threshold>3</threshold>
                    </levenshtein>
                </fuzzy>
            </search>
        </correlation>
    </item>
    ...
    <correlation>
        <correlators>
            <items>
                ...
                <item>
                    <ref>givenName</ref> (2)
                </item>
                ...
            </items>
            <items>
                ...
                <item>
                    <ref>givenName</ref> (2)
                </item>
                ...
            </items>
        </correlators>
    </correlation>
</objectTemplate>
1 These search options are specified at one place only.
2 Item givenName can be now used in various correlation rules.
These search options are completely overridden when search element is present in a correlator. (Meaning there is no partial override.)

Short (Attribute-Bound) Form

In simple scenarios, it is possible to define single-item correlator right on the resource object attribute. An example:

Listing 6. ri:employeeNumber serving as a correlation attribute
<schemaHandling>
    <objectType>
        ...
        <attribute>
            <ref>ri:employeeNumber</ref>
            <correlator>
                <search> ... </search> <!-- optional -->
            </correlator>
            <inbound>
                <target>
                    <path>employeeNumber</path>
                </target>
            </inbound>
        </attribute>
        ...
    </objectType>
</schemaHandling>

The correlator element may be empty. Its mere existence is sufficient to enable correlation based on that attribute and its corresponding focus item.

The correlator element is translated into single-item items correlator, just like the one described in Listing 1. It is added to other correlators defined for the object type - either attribute-bound or object-type-bound.

Requirements

In order for the attribute-bound correlator be used, the following conditions must be fulfilled:

  1. There must be at least one inbound mapping defined for the attribute.

  2. Corresponding focus item path must be determinable.

    1. Either it is specified explicitly (using focusItem property of the attribute-bound correlator);

    2. or there is exactly one inbound mapping, and it has a target path that points to the focal object.

The correlator created has no customizations (like weight, ordering, tier number, and so on). It contains only the reference to the focus item and the (optional) search parameters.

Limitations

  • Matching rules defined at the level of focus items in the object template are ignored during the correlation search. (See a note in Specifying Matching Rule section.)

  • When using attribute-bound form, the Requirements described above must be met.

Was this page helpful?
YES NO
Thanks for your feedback