0.10.0

[rcap107]

✨ Skrub version 0.9.0 has been released ✨

Main Changes

• New transformers are now available: SessionEncoder groups timestamped data by sessions, DropSimilar helps with dropping redundant columns.
• ToFloat now includes decimal and thousand as parameters to parse numerical columns that use formatting different from the python default (such as "1 234,5").
• The TableReport can now be exported in text format (Markdown formatting), or as a dictionary.
• DataOps have been improved with better caching, as well as easier access and search of nodes in the graph. It is also possible to filter which columns are shown in the parallel coordinate plot.

In detail

TableReport

• A summarized version of the TableReport can now be exported in Markdown. By @rcap107 in ENH - Adding a function that exports the TableReport in Markdown format #2048
• The TableReport can now be exported as a dictionary. By @m4nn2609-dot in Add dict method for converting report data to dictionary #2188
• It is now possible to specify how many rows should be shown by the TableReport via the table_report_n_rows configuration option. By @m4nn2609-dot in Add table_report_n_rows configuration option to TableReport #2193

Transformers

• The SessionEncoder is now available. This encoder adds a session_id column that groups together events that happen within the same session gap. Events can also be grouped by a column or list of columns to distinguish between users or devices (for example). By @rcap107 in Adding the SessionEncoder #1930
• The ToFloat transformer now allows to specify decimal and thousand separators to parse numerical columns that use formatting different from the default (such as "1 234,5"). Additionally, negative numbers indicated in parentheses are converted to negative numbers ("(135)" becomes "-135"). By @gabrielapgomezji in FEAT - Adding decimal as parameter for ToFloat32 #1772
• The HTML diagram of skrub estimators now includes a link to the online documentation. By @rcap107 in ENH - adding doc link to html repr of estimators #2036
• A new transformer, DropSimilar, allows dropping columns that are strongly correlated with each other. By @emassoulie in Issue 1249 add DropSimilar transformer #2023

Data Ops

• Dataop nodes now have a unique ID. This makes it easier to search for specific nodes in the DataOp to extract and set their values. By @jeromedockes in dataop nodes have a unique ID #2062, Set params by name #2090, support other kinds of predicate than name in find_fitted_estimator #2194
• It is now possible to set the variable preview values on a DataOp. By @jeromedockes in allow setting the variable preview values on a dataop #2081
• skrub.var now has the becomes_default parameter, to indicate that the provided preview value should be treated as a default for this variable in all contexts. By @jeromedockes in Add store_default parameter to skrub.var #2082
• Predict, predict_proba, etc. are now cached during SkrubLearner.score() by @jeromedockes in cache predict, predict_proba, etc. during SkrubLearner.score() #2195
• It is now possible to choose which columns of the parallel coordinate plot are shown. By @jeromedockes in allow filtering coords of the parallel coord plot #2202

Bugfixes

• Datetime-based transformers had inconsistent behaviors when they encountered duration columns. This has been fixed by @rcap107 in FIX - adding a transformer that converts durations to seconds #2069
• The TableReport could not plot dataframes that contained $$ as value. This was fixed by @Michenina-Lab in Fix - tablereport plots fail when dataframe contains $$ as values  #2154
• Fixed an error in the TableReport that occurred when the data range is very narrow. By @jeromedockes in fix tablereport error in histogram when data range is very narrow #2189

Misc

• The fetcher for the dataset on electricity forecasting in France has been added by @lisaleemcb in FEAT - Add a fetcher for the electricity forecasting dataset #2013
• The object selector has been added to skrub.selectors by @omkar-334 in add an object selector to skrub.selectors #2171

Deprecations/dependencies

• The parameter order_by of the TableReport is now deprecated by @uniheko in #1464 Deprecate the parameter order_by of the TableReport #2101
• Matplotlib's minimum required version has been increased from 3.4.3 to 3.6.1 by @rcap107 in MAINT - bumping matplotlib, unpinning pyparsing #2159

New Contributors

• @sandrineh made their first contribution in fix issue 1828 about a warning #1908
• @VeraCurly made their first contribution in Alphabetical order of autosummary lists #2095
• @4navi made their first contribution in add docstring for datetimeencoder module #2099
• @AlicjaKo made their first contribution in Added message and and test for associations tab with one column in _data/templates column-assosiations.html and tests/test_table_report.py #2094
• @siddiqua-tamk made their first contribution in DOC Fix Cleaner docstring formatting #2105
• @uniheko made their first contribution in #1464 Deprecate the parameter order_by of the TableReport #2101
• @omkar-334 made their first contribution in fix: rename DATAFAME variable name #2119
• @aidbar made their first contribution in list(outcomes) in choose_from #2100
• @hafsaelibrahimi made their first contribution in doc: fixing documentation in textencoder #2151
• @Maryahcee made their first contribution in Add drop method to selector base class and include unit tests #2108
• @uju-ds made their first contribution in Added docsting to drop #2104
• @rajalakshmikr made their first contribution in Removed spaces #2156
• @eve2705 made their first contribution in Removed old FIXME files #2157
• @maureen-githaiga made their first contribution in grouping examples into sections #2102
• @Michenina-Lab made their first contribution in Fix - tablereport plots fail when dataframe contains $$ as values  #2154
• @glitch-ux made their first contribution in Fix MinHashEncoder(hashing='fast') dropping the largest n-gram size #2168
• @m4nn2609-dot made their first contribution in Add dict method for converting report data to dictionary #2188

Full Changelog: 0.9.0...0.10.0

---

This discussion was created from the release 0.10.0.