- Add
tp.from_polars
andtp.to_polars
. @akshatvishu - Add
EventSet.filter_moving_count
. - Add
tp.input_node_from_schema
. - Add
tp.from_struct
.
- Add
before_first
andafter_last
parameters toEventSet.tick
andEventSet.tick_calendar
- Added
bokeh_webgl
as a possiblebackend
fortp.plot
. - Binary operations maintain the name of the feature on the left.
- Implement more operators in beam (still in beta).
- Fixed a bug in
EventSet.tick_calendar
where it would miss the first tick under certain conditions - Fixed a bug in
EventSet.tick
where it would miss the last tick under certain conditions - Upgrade to tensorflow 2.13 for compatibility with macOS on arm
Adding missing wheels for macos, no changes to the library
- Add
EventSet.moving_product()
andEventSet.cumprod()
operators. @akshatvishu - Add
to.to_numpy()
. @nagavenkateshgavini - Add trigonometric functions
EventSet.arccos()
,EventSet.arcsin()
,EventSet.arctan()
,EventSet.cos()
,EventSet.sin()
, andEventSet.tan()
. @akshatvishu
- Speed up of calendar operations (now implemented in c++)
- Add
force_garbage_collector_interval
totp.compile
. - Improved worst case time complexity from quadratic to linear for moving min/max operator.
- Show first and last events instead of only first ones when displaying an EventSet. @jtaylor205
- Exposed
tp.FeatureSchema
andtp.IndexSchema
EventSet.propagate
now works even if both arguments don't have the same index values.- Add
font_scale
param totp.plot
- Speed-up
tp.plot
on evsets with a large number of index values. tp.plot
now works even when evsets have different index values.- Allow
tp.from_tensorflow_record
to consume a list of paths. - Add parameters
num_parallel_reads
andbuffer_size
totp.from_tensorflow_record
. - Check that timestamps are sorted on
tp.from_tensorflow_record
- Fixed a bug with
EventSet.tick_calendar
and daylight savings time. - Fixed a bug with calendar operations and daylight savings time.
In adition to the contributors mentioned above, thanks to @umbr4g3, @jsoref, and @tanaysd for improvements to the Github Actions and profiling.
- Add
tp.from_parquet()
andtp.to_parquet()
. - Add
EventSet.fillna()
operator.
- Add support for pip build on Windows.
- Documentation improvements.
- Add
timestamps
parameter totp.from_pandas()
. - Add implicit casting in
EventSet.where()
operator. - Add support for list argument in
EventSet.rename()
operator.
- Support for
timezone
argument in all calendar operators. - Add
drop()
operator to drop features. - Add
assign()
operator to assign features. - Add
before()
andafter()
operators.
- Improve error messages for type mismatch in window operators.
- Improve structure of docs site.
- Support exporting timestamps as datetimes in
tp.to_pandas()
. - Remove inputs limit in
glue()
andcombine()
.
- Use
wday=0
for Mondays intick_calendar
(likecalendar_day_of_week
). - Support bool in
DType.missing_value()
. - Show
EventSet
's magic methods in docs.
- Added
EventSet.filter_moving_count()
operator. - Added
EventSet.map()
operator. - Added
EventSet.tick_calendar()
operator. - Added
EventSet.where()
operator. - Added all moving window operators to Beam execution backend.
- Print
EventSet
timestamps as datetimes instead of float. - Support
sampling
argument inEventSet.cumsum()
operator. - Using utf-8 codec to support non-ascii in string values.
- New
tp.types
module to facilitate access to types used throughout the API. - Relaxed version requirements for protobuf and pandas.
- Fixed issues when loading timestamps from
np.longlong
and other dtypes.
- Added
EventSet.select_index_values()
operator. - Added
steps
argument toEventSet.since_last()
operator. - Added variable
window_length
option to moving window operators. - Added unsupervised anomaly detection tutorial.
- Add
until_next
operator. - Added Beam execution tutorial.
- Added changelog to docs site.
- Added
display_max_feature_dtypes
anddisplay_max_index_dtypes
options totp.config
. - Improved HTML display of an
EventSet
. - Improvements in Beam execution backend.
- Fixed tutorials opening unreleased versions of the notebooks.
This is the first operational version of Temporian for users. The list whole and detailed list of features is too long to be listed. The main features are:
- PyPI release.
- 72 operators.
- Execution in eager, compiled mode, and graph mode.
- IO Support for Pandas, CSV, Numpy and TensorFlow datasets.
- Static and interactive plotting.
- Documentation (3 minutes intro, user guide and API references).
- 5 tutorials.