mongotrainer

Documentation

Index

• Variables
• type DataSource
  • func NewDataSource(c *pb.DataSourceConfig, s *mgo.Session) (DataSource, error)
• type MongoTrainer
  • func (m *MongoTrainer) Loop()

Variables

var (
	// MongoServer to connect to
	MongoServer = flag.String("mongo_server", "localhost", "")
	// MongoDatabase to use
	MongoDatabase = flag.String("mongo_database", "test", "")
	// MongoCollection with active tasks
	MongoCollection = flag.String("mongo_collection", "decisiontrees", "")
	// MongoFs with GridFS files
	MongoFs = flag.String("mongo_fs", "fs", "")
	// MongoPollTime is the interval between polling the task collection for
	// unclaimed tasks
	MongoPollTime = flag.Duration("mongo_poll_time", 30*time.Second, "")
)

Types

type DataSource

type DataSource interface {
	GetTrainingData() (*pb.TrainingData, error)
}

DataSource is an interface for extracting training data from an arbitrary location TODO(tulloch) - make this an iterator?

func NewDataSource

func NewDataSource(c *pb.DataSourceConfig, s *mgo.Session) (DataSource, error)

NewDataSource is a factory function that returns a new DataSource given the input DataSourceConfig.

type MongoTrainer

type MongoTrainer struct {
	Collection *mgo.Collection
}

MongoTrainer polls a MongoDB collection for changes and spins up training jobs based on these changes

func (*MongoTrainer) Loop

func (m *MongoTrainer) Loop()

Loop starts the polling thread, and selects on the channel of potential tasks