Wsrtjtyk

I'm investigating the use of YAML for a somewhat complicated metadata language. It would help to make the documents smaller and less complex if we could use YAML's anchors and references. I've written some test code that seems to show that Jackson's YAML implementation doesn't support this feature (and/or doesn't surface SnakeYAML's support for this feature).

Here is my test YAML file:

set_one:

  bass: tama rockstar 22x16

  snare: &ludwig ludwig supralight 6.5x15

  tom1: tama rockstar 12x11

  tom2: tama rockstar 16x16



set_two:

  snare: *ludwig

I'm parsing this file like so:

    ObjectMapper mapper = new ObjectMapper(new YAMLFactory());

    FileInputStream fis = null;



    try

    {

        fis = new FileInputStream(file);

        JsonNode nodeTree = mapper.readTree(fis);

        examineObject(nodeTree, 0);

    }

    ...

Here is the out from my "examineObject()" method (you can probably guess what it does):

key = "set_one", type = OBJECT

  key = "bass", type = STRING, value = "tama rockstar 22x16"

  key = "snare", type = STRING, value = "ludwig supralight 6.5x15"

  key = "tom1", type = STRING, value = "tama rockstar 12x11"

  key = "tom2", type = STRING, value = "tama rockstar 16x16"

key = "set_two", type = OBJECT

  key = "snare", type = STRING, value = "ludwig"

Clearly something knew enough to omit the anchor value from "set_one.snare" but, in the debugger, I can't find that value anywhere in the JsonNode for this element. The real problem is that the value of "set_two.snare" is just "ludwig". The reference symbol ('*') has been stripped, but the value is that of the reference and not the element it is referring to.

I'm using Jackson version 2.8.3 and SnakeYaml version 1.17. I am constrained to using Jackson as this is only part of a much bigger project which already uses Jackson for JSON.

What I would really like is if Jackson could automatically resolve references and make a copy of the referenced value. In my example this would mean that the value of "set_two.snare" would be "ludwig supralight 6.5x15".

If I can't get my first choice then I would like Jackson to preserve both the anchors and the references so that I could manually post-process the node tree and resolve the references myself. For example, when I saw that the value of "set_two.snare" was "*ludwig", I could search the tree for a node with an anchor of "&ludwig" and make a copy of that node.

If there is an answer, I have the feeling that it will probably involve the "com.fasterxml.jackson.dataformat.yaml.YAMLParser.Feature" class somehow. Unfortunately I can't find any documentation on the features (if they exist) that will enable the behavior I am looking for.

I just tried to make anchors/aliases work in Jackson... and failed.
You can see here that alias based id support is not implemented. The _currentAnchor instance variable is set and exposed by getObjectId(), but I didn't find a practical way to hook into Jackson to use that method. This is not my first fight with Jacksons Object-Id resolution architecture. Be advised not the spend too much time with it.

My solution was using the snakeyaml library directly.

First, Jackson actually does support YAML anchors and references, at least to degree they work with how Jackson supports Object Id references with @JsonIdentityInfo: limitation being that you can not -- for example -- refer to one key/value pair of am Object.

But identity id/reference handling is only enabled for types and properties specified by annotating then with @JsonIdentityInfo.
So you have to annotate either types that may be referenced, or properties (no need to do both).

One thing that may help here is to consider that Object Id handling by Jackson is very similar for all formats: so although jackson-dataformat-yaml does expose "native" Object (and Type) Ids that YAML has (and JSON does not have), handling at databinding level is identical. So if you can make Object Id/References work with JSON (which adds extra id property), it will work with YAML as well.

There is one extra thing that is related: YAMLParser.Feature.USE_NATIVE_OBJECT_ID which determines how references and ids are expressed when writing YAML -- by default, it uses native anchors, but it can be turned off to use "JSON-like" plain properties.

I hope this helps. For additional help the best place would be jackson-users mailing list.