Since the version 4.0 Report Portal uses a new version of an analyzer. The implementation of the analyzer was moved to the separate service. If the default implementation does not provide a solution it can be extended by a custom implementation of an analyzer. The default analyzer can work together with a custom one. ReportPortal has a client that communicates with analyzers by HTTP and it has a pretty simple interface.

All supported endpoints that could be implemented in a custom analyzer are:

Analyze launch

POST
http://analyzer_host:port/_analyze

Index items. Optional

POST
http://analyzer_host:port/_index

Delete indexed project. Optional

DELETE
http://analyzer_host:port/_index/{project}

Clean indexed items. Optional

PUT
http://analyzer_host:port/_index/delete

In a custom analyzer should be implemented at least one endpoint that consumes request from RP with a launch to be analyzed using the next endpoint:

POST
http://analyzer_host:port/_analyze

It should consume requests in the next json format:

Implementation in Java

[
  {
    "launchId": "5a0d84a8eff46f62cfd9cbe4",                   
    # Launch id in ReportPortal
    "launchName": "test-results",  
    # Launch name in ReportPortal. In default implementation issues from the launch with
    # the same launch name have a higher priority
    "project": "analyzer",                       
    # Project name. Default implementation categorises items by project  
    "testItems": [                                            
    # Test items to be analyzed
      {
        "autoAnalyzed": false,
        # Optional, used for marking items as analyzed. In default implementation it means 
        # that the item that was analyzed by human ('false') has a higher priority
        "testItemId": "5a0d84b6eff46f62cfd9cd9f",             
        # Test item id in ReportPortal       
        "issueType": "TI001",         
        # Only items with RP status 'TO_INVESTIGATE' can be analyzed        
        "logs": [
        # Test item's logs with level "ERROR" and higher          
          {
            "logLevel": 40000,
            "message": "java.lang.AssertionError: expected..."
          } 
        ],                                                    
        "uniqueId": "auto:f9377b76a3ebede22df09fa76400788b"   
        # Test Item's unique id in RP. In default implementation issue from item 
        # with the same unique id has a higher priority
      }
    ]
  }
]

In the current default realization it actually sends only one launch in any case.

Report Portal accepts the analyzed items back as a response in the next json format:

Implementation in Java

[
    {
        "test_item": "5a0d84b6eff46f62cfd9cd9f",
        # Test item id in RP        
        "issue_type": "PB001",
        # Investigated issue type that will be applied to the item
        "relevant_item": "5a1be11deff46f8b838fc5e5"
        # The most relevant item id. It is used for taking a comment 
        # and an external system issue from the most relevant item in RP
    }
]

As a result items are updated in ReportPortal database with a new issue, comment and reference in BTS if any are found. Also new analyzed items are indexed.

If an analyzing alghorithm is based on the previous results, the analyzer interface also provides possibility to collect information about updated items in RP. To have that logic there should be implemented one more endpoint:

POST
http://analyzer_host:port/_index

The request contains the same list of json objects as in "_analyze" higher except of the "issueType" sould be provided by user and it should be different from "TO_INVESTIGATE". In the current implementation RP doesn't really use the response from index so it could be ignored (or just 'null').

There are a few more optional endpoints. They are required for the default analyzer implementation. The first one is for deleting all the accumulated information about specified project:

DELETE
http://analyzer_host:port/_index/{project}

The other is for cleaning information about the specified test items from the specified project:

PUT
http://analyzer_host:port/_index/delete

{
  "ids": [
    "5a1be11deff46f8b838fc5e5"
  ],
  "project": "analyzer"
}

For correct interaction with Report Portal the analyzer must have several required tags in it's consul service configuration.

• analyzer=custom

  Marks that the current service is analyzer implementation with name 'custom'
  

• analyzer_priority=0

  If the service is going to work together with default analyzer there should be specified the serivce's priority. 
  For basic analyzer it is 10. So if the current analyzer's resutls are more important than default's the priority 
  parameter should be lower than 10. If the value is not specified or incorrect than the priority is the 
  lowest by default.
  

• analyzer_index=true

  If the service processes previous results it should have the tag 'analyzer_index=true'. It does indexing of logs.
  If value is not specified or incorrect than 'false' by default.
  

Java configuration example

All included analyzers could be found by sending a request to the next endpoint:

GET
http://reportportal_host:port/composite/info

This repository is an example of all endpoints' implementation using Java