Python API for Vertica Data Science at Scale
What's New in VerticaPy v0.10.x
⚠ Many functions have deprecated (see the list below).
- [New Parameters] Since VerticaPy 0.10.2, vDataFrame.groupby has two new parameters: 'rollup' and 'having'. Rollups and filtering operations are now available.
- vDataFrame has prettier designs when 'interactive' parameter is set to True. Look at set_option function for more information.
- [New Algorithm] IsolationForest is now available.
- VerticaPy 0.10.1 introduces the decision_function method for calculating the anomaly score of a vDataFrame and updates the predict function to account for anomalies.
- [New Parameters] vDataFrame: vDataFrames can now be created directly from a list, an array, a pandas.DataFrame, a tablesample or a dictionary.
- [New Parameters] set_option: You can now toggle displaying vDataFrame counts and footers and display vDataFrames as interactive tables.
- [New Method] vDataFrame.idisplay: This method displays the interactive table. It is used when you don't want to activate interactive table for all vDataFrames.
- [New Function] Since VerticaPy 0.10.2, the function save_to_query_profile is available. This method saves information about the specified VerticaPy method to the QUERY_PROFILES table in the Vertica database. It is used to collect usage statistics on methods and their parameters. This function generates a JSON string.
- [New Function] atan2 : Trigonometric inverse tangent of the arithmetic dividend of the arguments.
- Since VerticaPy 0.10.2, functions usage and statistics are stored in the QUERY_PROFILES table. A specific label is used to track each of them. See save_to_query_profile function for more information.
- VerticaPy Machine Learning Cheat Sheet is now available.
DEPRECATED since 0.9.0
- Functions that previously used a cursor now instead leverage a single connection by VerticaPy; you cannot pass cursors to these functions anymore. The connection functions have been renamed. For a list of the new connection functions, see Connect. For an example of how to connect to a database with VerticaPy, see the connection tutorial.
- Some functions use a temporary schema to create objects. You can now change this temporary schema with set_option. As a result, the method vDataFrame.'set_schema_writing' has been deprecated and will be removed in a later release.
- vModel.predict no longer computes class probabilities. This functionality has been moved to vModel.predict_proba.
- The methods vModel.to_sklearn and vModel.shapExplainer have been deprecated due to integration complexity and the limitations of their respective APIs. A function to compute shap values will be added in a future release. As a substitute for vModel.to_sklearn, you can try instead converting vModels to memModels, which are machine-independent, more flexible, and can be used for in-memory or in-database scoring.
- To simplify usage, the magic command vCharts has been renamed to hchart to simplify usage.
- The function vHelp has been renamed to help_start.
- The function vdf_from_relation has been renamed to vDataFrameSQL, and you can now create a vDataFrame directly from a query. This is the preferred usage.
- The method vDataFrame.set_cursor has been deprecated.
- vDataFrame.count only returns the number of missing elements, rather than the number and the percent. To get the percent of missing elements, use vDataFrame.count_percent.
- The default parameters of vDataFrame.describe have been changed. The cardinality is only computed if the parameter 'unique' is set to True.
- Many parameters of SQL Magic and hchart Magic are deprecated.
GitHub and Unit Tests
- More unit tests have been added.