RumbleML: Missing support for vectors in parameters other than featuresCol

[CanBerker]

let $data := annotate(
    json-file("./src/main/resources/queries/rumbleML/sample-ml-data-flat.json"),
    { "label": "integer", "binaryLabel": "integer", "name": "string", "age": "double", "weight": "double", "booleanCol": "boolean", "nullCol": "null", "stringCol": "string", "stringArrayCol": ["string"], "intArrayCol": ["integer"],  "doubleArrayCol": ["double"],  "doubleArrayArrayCol": [["double"]] }
)

let $est := get-estimator("BucketedRandomProjectionLSH")
let $tra := $est(
    $data,
    { "inputCol": "weight" }
)
for $result in $tra(
    $data,
    { }
)
return {
    "label": $result.label,
    "name": $result.name,
    "age": $result.age,
    "weight": $result.weight,
    "result": $result
}

⚠️ ️Error [err: RBML0003]LINE:6:COLUMN:12:Invalid Param; Parameter provided to BucketedRandomProjectionLSH causes the following error: requirement failed: Column weight must be of type struct<type:tinyint,size:int,indices:array,values:array> but was actually double.

BucketedRandomProjectionLSH requires a vector in its 'inputCol' param. Currently we only support vectors with an internal trick on the featuresCol

Same applies to the following classes:

Estimators expecting 'inputCol' as a vector:

• BucketedRandomProjectionLSH
• IDF
• MaxAbsScaler
• MinHashLSH
• MinMaxScaler
• PCA
• StandardScaler
• VectorIndexer

Transformers expecting 'inputCol' as a vector:

• DCT
• ElementwiseProduct
• Normalizer
• PolynomialExpansion
• VectorSizeHint
• VectorSlicer

Edit: These will be extracted to a new issues: ~Transformers 'inputCols' as a vector (each column in inputCols can be a vector):~

• ~Interaction~

~ElementwiseProduct also requires a vector in its scalingVec param.~