Data Types

All data manipulated and stored using frames and graphs must fit into one of the supported Python data types.

>>> ta.valid_data_types

float32, float64, ignore, int32, int64, unicode, vector(n), datetime
(and aliases: float->float64, int->int32, list->vector, long->int64, str->unicode)

datetime	[ALPHA] object for date and time; equivalent to python’s datetime.datetime class. Converts to and from strings using the ISO 8601 format. Inside the server, the object is represented by the nscala/joda DateTime object. When interfacing with various data sources and sinks that use different data types for datetime, the datetime value will be converted to a string by default.
float32	32-bit floating point number; equivalent to numpy.float32
float64	64-bit floating point number; equivalent to numpy.float64
ignore	type available to describe a field in a data source that the parser should ignore
int32	32-bit integer; equivalent to numpy.int32
int64	32-bit integer; equivalent to numpy.int64
unicode	Python’s unicode representation for strings.
vector(n)	[ALPHA] Ordered list of n float64 numbers (array of fixed-length n); uses numpy.ndarray

Note: Numpy values of positive infinity (np.inf), negative infinity (-np.inf) or nan (np.nan) are treated as Python’s None when sent to the server. Results of any user-defined functions which deal with such values are automatically converted to None. Any further usage of those data points should treat the values as None.