NEW:

If a source extjson document contains a document field that is not represented in the target struct and that document contains a nested document or array, unmarshalling can fail or incorrectly skip too many fields.

ORIGINAL:

The (extJSONValueReader).Skip() method fails when it tries to skip a document that has a document as a child. Specifically, the following test fails on the last line with the error "invalid request to read embedded document":

func TestExtJSONValueReader_Skip(t *testing.T) {
    vr, err := NewExtJSONValueReader(strings.NewReader(`{"key": {"one": {"two": "three"}}}`), false)
    if err != nil {
          t.Fatalf("error creating vr: %v", err)
    }

    dr, err := vr.ReadDocument()
    if err != nil {
        t.Fatalf("error reading doc: %v", err)
    }

    _, vr, err = dr.ReadElement()
    if err != nil {
        t.Fatalf("error reading element: %v", err)
    }

    err = vr.Skip()
    if err != nil {
        t.Fatalf("error skipping: %v", err)
    }
}

That error comes from the extJSONParser.readValue() method which cannot read a whole embedded document. I believe the fix for this is to check in Skip() if the type is an embedded document and handle it appropriately there.