Identifier¶

The "identifier" transform adds a runtime-unique identifier (__uniqueId) to each data object.

GenomeSpy uses these identifiers internally for picking, including tooltips and point selections. It inserts an identifier transform automatically when a mark participates in picking and the dataflow branch does not already contain one.

Identifiers are runtime values. Do not use them as persistent record keys. For point selections that need to survive state persistence, use encoding.key, as described in the non-visual channels section, with fields that identify the data objects independently of the current data flow.

Explicit identifiers¶

Most specifications do not need to add this transform themselves. Add one explicitly when several downstream views must refer to the same input object after the data has been expanded or branched. Place it before the expansion or branching step:

{
  "transform": [
    { "type": "identifier" },
    {
      "type": "flattenCigar",
      "copyFields": ["__uniqueId", "chrom", "name", "cigar"]
    }
  ],
  "layer": [
    {
      "transform": [{ "type": "filter", "expr": "datum.cigarType == 'skip'" }],
      "mark": "rule"
    },
    {
      "transform": [
        { "type": "filter", "expr": "datum.cigarType == 'aligned'" }
      ],
      "mark": "rect"
    }
  ]
}

Both layers receive the same identifier for all CIGAR elements derived from one input alignment. When a transform limits the fields copied to generated rows, include "__uniqueId" explicitly, as in the example above.