Releases: Smithsonian/smax-python
v1.2.3
v1.2.2
Fix a bug in SmaxInt<x>.data, caused by recently added numpy.<Type>.data property having precedence over our own property.
The numpy.<Type>.data property, which returns a numpy.memoryview of the object, is available in Smax types as .memoryview.
v1.2.1
Change the default logging level to INFO from DEBUG.
v1.2.0
This release brings the available types for pushing to SMA-X into line with the latest strict specification. At the same time, it enables a wider range of types for pulling from SMA-X, for backwards compatibility.
New features:
- integers or floats shared to SMA-X that overflow the requested type will automatically be promoted to the necessary width.
- a 'maintain_type' option to
smax_sharehas been added to retrieve the current SMA-X type from the SMA-X server, and cast the input value(s) to that type - a 'raw' option has been added to
smax_pull, to directly return the byte sequence returned by Redis, for debugging purposes.
v1.1.0
This release updates the default mapping of built-in Python int and float to SMA-X int32 and float64. The SMA-X raw type is implemented as SmaxBytes, derived from the built-in Python bytes type, although with some differences (in particular, SmaxBytes(2) == b'2', and not \x00\x00 as for bytes(2)). UTF-8 encoding is assumed for converting bytes to strings and vice versa - this is also true in Redis.
The mapping of built-in int to SMA-X types may change in the future, to use the minimum width necessary to start the value, rather than defaulting to 32 bit.
v1.0.5
Properly fix the previous issue that we tried to fix in v1.0.4
v1.0.4
Fix an issue with subscribing - Retry.call_with_retry() changed signature to require a fail callback function.
v1.0.3
Remove extraneous print statements that were producing strange output.
v1.0.2
Alter the smax command line interface to work with one or both of the -t
v1.0.1
Bump version: 1.0.0 → 1.0.1