Model.updateAttributes doesn't consider composite primary keys

[shenghu]

Description/Steps to reproduce

1. Create a table w/ composite primary key, e.g.

    CASS_SORTABLE = db.define('CASS_SORTABLE', {
      patBool: {type: Boolean, id: 2},
      str: String,
      patStr: {type: String, id: true},
      num: Number,
      patNum: {type: Number, id: 1},
      yearMonth: {type: String, index: true},
      }, {
      cassandra: {
        clusteringKeys: ['str', 'num DESC'],
        },
      });

2. Insert a record

   var ROW
   CASS_SORTABLE.create({
     str: cassTestString + '10',
     num: 10,
     patBool: true,
     patNum: 100,
     patStr: cassTestString + '100',
   }, function(err, m) {
     ROW = m;
   });

3. Update attributes of the row,

   ROW_1.updateAttributes({ 
     str: ROW_1.str,
     num: ROW_1.num,
     patBool: ROW_1.patBool,
     patNum: ROW_1.patNum,
     patStr: ROW_1.patStr,
     yearMonth: '2018-06'}, function(err, rows) {
     });

The update failed w/ error

     Uncaught AssertionError: expected ResponseError {
  name: 'ResponseError',
  info: 'Represents an error message from the server',
  message: 'PRIMARY KEY part str found in SET part',
  code: 8704,
  query: 'UPDATE "CASS_SORTABLE" SET "str"=?,"num"=?,"yearMonth"=? WHERE "patStr"=?'
} to not exist

Link to reproduction sandbox

Expected result

The operation is expected to succeed.

Additional information

ROW.updateAttributes called these codes, SQLConnector.prototype.updateAttributes = function(model, id, data, options, cb) , which is called by DataAccessObject.prototype.updateAttributes.

In DataAccessObject.prototype.updateAttributes, it calls

              invokeConnectorMethod(connector, 'updateAttributes', Model,
                [getIdValue(Model, inst), Model._forDB(context.data)],
                options, updateAttributesCallback);

getIdValue(Model, inst) doesn't consider the composite primary key.

The ideal fix is in juggler to get an object from getIdValue(Model, inst) instead of a string value. But before that, a workaround can be in SQLConnector.prototype.updateAttributes, to put primary keys inwhere and remove them from data.

[shenghu]

Basically when where is needed, primary ids will be excluded from data. If where doesn't include all primary ids, the operation will fail.

When where is not available, primary ids in data will be used to form where predicate.