Wip for using validator within sequelize, when creating or modifying.

sequelize_dao.js

  // if an array with field names is passed to save()
  // only those fields will be updated
  DAO.prototype.save = function(fields) {
    var self          = this
      , values        = fields ? {} : this.values
      , updatedAtAttr = this.__options.underscored ? 'updated_at' : 'updatedAt'
      , createdAtAttr = this.__options.underscored ? 'created_at' : 'createdAt'

// --- validation START
      
    var validation_errors = self.validate();

    if (validation_errors !== null) {
      return new Utils.CustomEventEmitter(function(emitter) {
        setTimeout( function() {
          emitter.emit('error', validation_errors);
        }, 100);
      }).run();
    }

// --- validation END

    if(fields) {
      if(self.__options.timestamps) {
        if(fields.indexOf(updatedAtAttr) === -1) {
          fields.push(updatedAtAttr)
        }

        if(fields.indexOf(createdAtAttr) === -1) {
          fields.push(createdAtAttr)
        }
      }

      fields.forEach(function(field) {
        if(self.values[field] !== undefined) {
          values[field] = self.values[field]
        }
      })
    }


    if(this.__options.timestamps && this.hasOwnProperty(updatedAtAttr)) {
      var now = new Date()
      this[updatedAtAttr] = now
      values[updatedAtAttr] = now
    }

    if(this.isNewRecord) {
      return this.QueryInterface.insert(this, this.__factory.tableName, values)
    } else {
      var identifier = this.__options.hasPrimaryKeys ? this.primaryKeyValues : this.id
        , tableName  = this.__factory.tableName
        , query      = this.QueryInterface.update(this, tableName, values, identifier)

      return query
    }
  }

The event emitter is for handling asynchronous callbacks, and since validations are not async you shouldn't create one. I was thinking something like:

var validation_errors = self.validate();

if (validation_errors !== null) {
  return new Utils.CustomEventEmitter(function(emitter) {
    emitter.emit('erorr', validation_erorrs);
  }).run();
};

The reason you are returning an eventEmitter is that the user expects something that he can call done, error etc. on. But the emitter should only be created if there is an actual error to emit.

I'm not sure about race-conditions here - emit might be called before the .error callback can be attached, but try this and if it doesn't work wrap the body of the emit function in a setTimeout