Release Notes#

Version 5.1.0 (unreleased)#

Enhancements#

Bug Fixes#

Backward-Incompatible Changes#

Version 5.0.0 (released May 9, 2023)#

Enhancements#

  • As described in the release notes for Vega-Lite 5.0.0, the primary change in this release of Altair is the introduction of parameters. There are two types of parameters, selection parameters and variable parameters. Variable parameters are new to Altair, and while selections are not new, much of the old terminology has been deprecated. See Slider Cutoff for an application of variable parameters (#2528).

  • Grouped bar charts and jitter are now supported using offset channels, see Grouped Bar Chart with xOffset and Strip Plot with Jitter

  • vl-convert is now used as the default backend for saving Altair charts as svg and png files, which simplifies saving chart as it does not require external dependencies like altair_saver does (#2701). Currently, altair_saver does not support Altair 5 and it is recommended to switch to vl-convert. See PNG, SVG, and PDF format for more details.

  • Saving charts with HTML inline is now supported without having altair_saver installed (#2807).

  • The default chart width was changed from 400 to 300 (#2785).

  • Ordered pandas categorical data are now automatically encoded as sorted ordinal data (#2522)

  • The Title and Impute aliases were added for TitleParams and ImputeParams, respectively (#2732).

  • The documentation page has been revamped, both in terms of appearance and content.

  • More informative autocompletion by removing deprecated methods (#2814) and for editors that rely on type hints (e.g. VS Code) we added support for completion in method chains (#2846) and extended keyword completion to cover additional methods (#2920).

  • Substantially improved error handling. Both in terms of finding the more relevant error (#2842), and in terms of improving the formatting and clarity of the error messages (#2824, #2568, #2979, #3009).

  • Include experimental support for the DataFrame Interchange Protocol (through __dataframe__ attribute). This requires pyarrow>=11.0.0 (#2888).

  • Support data type inference for columns with special characters (#2905).

  • Responsive width support using width="container" when saving charts to html or displaying them with the default html renderer (#2867).

Grammar Changes#

  • Channel options can now be set via a more convenient method-based syntax in addition to the previous attribute-based syntax. For example, instead of alt.X(..., bin=alt.Bin(...)) it is now recommend to use alt.X(...).bin(...)`) (#2795). See Method-Based Syntax for details.

  • selection_single and selection_multi are now deprecated; use selection_point instead. Similarly, type=point should be used instead of type=single and type=multi.

  • add_selection is deprecated; use add_params instead.

  • The selection keyword argument must in many cases be replaced by param (e.g., when specifying a filter transform).

  • The empty keyword argument for a selection parameter should be specified as True or False instead of all or none, respectively.

  • The init keyword argument for a parameter is deprecated; use value instead.

Bug Fixes#

  • Displaying a chart not longer changes the shorthand syntax of the stored spec (#2813).

  • Fixed disable_debug_mode (#2851).

  • Fixed issue where the webdriver was not working with Firefox’s geckodriver (#2466).

  • Dynamically determine the jsonschema validator to avoid issues with recent jsonschema versions (#2812).

Backward-Incompatible Changes#

  • Colons in column names must now be escaped to remove any ambiguity with encoding types. You now need to write "column\:name" instead of "column:name" (#2824).

  • Removed the Vega (v5) wrappers and deprecate rendering in Vega mode (save Chart as Vega format is still allowed) (#2829).

  • Removed the Vega-Lite 3 and 4 wrappers (#2847).

  • Removed the deprecated datasets.py (#3010).

  • In regards to the grammar changes listed above, the old terminology will still work in many basic cases. On the other hand, if that old terminology gets used at a lower level, then it most likely will not work. For example, in the current version of Scatter Plot with Minimap, two instances of the key param are used in dictionaries to specify axis domains. Those used to be selection, but that usage is not compatible with the current Vega-Lite schema.

  • Removed the altair.sphinxext module (#2792). The altair-plot Sphinx directive is now part of the sphinxext-altair package.

Maintenance#

  • Vega-Altair now uses hatch for package management.

  • Vega-Altair now uses ruff for linting.

Version 4.2.2 (released Jan 27, 2023)#

Bug Fixes#

  • Fix incompatibility with jsonschema < 4.5 which got introduced in Altair 4.2.1 (#2860).

Version 4.2.1 (released Jan 26, 2023)#

Bug Fixes#

  • Disable uri-reference format check in jsonsschema (#2771).

  • Replace iteritems with items due to pandas deprecation (#2683).

Maintenance#

  • Add deprecation and removal warnings for Vega-Lite v3 wrappers and Vega v5 wrappers (#2843).

Version 4.2.0 (released Dec 29, 2021)#

Enhancements#

  • Pie charts are now supported through the use of mark_arc. (Examples: eg. Pie Chart and Radial Chart.)

  • Support for the datum encoding specifications from Vega-Lite; see Vega-Lite Datum Definition. (Examples: Line Chart with Datum and Line Chart with Datum for Color.)

  • angle encoding can now be used to control point styles (Example: Wind Vector Map)

  • Support for serialising pandas nullable data types for float data (#2399).

  • Automatically create an empty data object when Chart is called without a data parameter (#2515).

  • Allow the use of pathlib Paths when saving charts (#2355).

  • Support deepcopy for charts (#2403).

Bug Fixes#

  • Fix to_dict() for nested selections (#2120).

  • Fix item access for expressions (#2099).

Version 4.1.0 (released April 1, 2020)#

  • Minimum Python version is now 3.6

  • Update Vega-Lite to version 4.8.1; many new features and bug fixes from Vega-Lite versions 4.1 through 4.8; see Vega-Lite Release Notes.

Enhancements#

  • strokeDash encoding can now be used to control line styles (Example: Multi Series Line Chart)

  • chart.save() now relies on altair_saver for more flexibility (#1943).

  • New chart.show() method replaces chart.serve(), and relies on altair_viewer to allow offline viewing of charts (#1988).

Bug Fixes#

  • Support Python 3.8 (#1958)

  • Support multiple views in JupyterLab (#1986)

  • Support numpy types within specifications (#1914)

  • Support pandas nullable ints and string types (#1924)

Maintenance#

  • Altair now uses black and flake8 for maintaining code quality & consistency.

Version 4.0.1 (released Jan 14, 2020)#

Bug Fixes#

  • Update Vega-Lite version to 4.0.2

  • Fix issue with duplicate chart divs in HTML renderer (#1888)

Version 4.0.0 (released Dec 10, 2019)#

Version 4.0.0 is based on Vega-Lite version 4.0, which you can read about at vega/vega-lite.

It is the first version of Altair to drop Python 2 compatibility, and is tested on Python 3.5 and newer.

Enhancements#

  • Support for interactive legends: (see Interactive Legend)

  • Responsive chart width and height: (see Adjusting Chart Size)

  • Lookup transform responsive to selections: (see Lookup)

  • Bins responsive to selections: (see Histogram with Responsive Bins)

  • New Regression transform: (see Regression)

  • New LOESS transform: (see LOESS)

  • New density transform: (see Density)

  • New pivot transform: (see Pivot)

  • Image mark (see Image)

  • New default html renderer, directly compatible with Jupyter Notebook and JupyterLab without the need for frontend extensions, as well as tools like nbviewer and nbconvert, and related notebook environments such as Zeppelin, Colab, Kaggle Kernels, and DataBricks. To enable the old default renderer, use:

    alt.renderers.enable('mimetype')

  • Support per-corner radius for bar marks: (see Bar Chart with Rounded Edges)

Grammar Changes#

  • Sort-by-field can now use the field name directly. So instead of:

    alt.Y('y:Q', sort=alt.EncodingSortField('x', order='descending'))

    you can now use:

    alt.Y('y:Q', sort="-x")

  • The rangeStep argument to Scale and Chart.configure_scale() is deprecated. instead, use chart.properties(width={"step": rangeStep}) or chart.configure_view(step=rangeStep).

  • align, center, spacing, and columns are no longer valid chart properties, but are moved to the encoding classes to which they refer.

Version 3.3.0 (released Nov 27, 2019)#

Last release to support Python 2

Enhancements#

  • Add inheritance structure to low-level schema classes (#1803)

  • Add html renderer which works across frontends (#1793)

  • Support Python 3.8 (#1740, #1781)

  • Add :G shorthand for geojson type (#1714)

  • Add data generator interface: alt.sequence, alt.graticule, alt.sphere() (#1667, #1687)

  • Support geographic data sources via __geo_interface__ (#1664)

Bug Fixes#

  • Support pickle and copy.deepcopy for chart objects (#1805)

  • Fix bug when specifying count() within transform_joinaggregate() (#1751)

  • Fix LayerChart.add_selection (#1794)

  • Fix arguments to project() method (#1717)

  • Fix composition of multiple selections (#1707)

Version 3.2.0 (released August 5, 2019)#

Upgraded to Vega-Lite version 3.4 (See Vega-Lite 3.4 Release Notes).

Following are changes to Altair in addition to those that came with VL 3.4:

Enhancements#

  • Selector values can be used directly in expressions (#1599)

  • Top-level chart repr is now truncated to improve readability of error messages (#1572)

Bug Fixes#

  • top-level add_selection methods now delegate to sub-charts. Previously they produced invalid charts (#1607)

  • Unsupported mark_*() methods removed from LayerChart (#1607)

  • New encoding channels are properly parsed (#1597)

  • Data context is propagated when encodings are specified as lists (#1587)

Backward-Incompatible Changes#

  • alt.LayerChart no longer has mark_*() methods, because they never produced valid chart specifications) (#1607)

Version 3.1.0 (Released June 6, 2019)#

Update includes full compatibility with version 3.3 of Vega-Lite.

Enhancements#

  • Added support for vega themes via alt.themes.enable(theme_name) (#1539)

  • Added an alt.renderers.disable_max_rows() method for disabling the maximum rows check (#1538)

  • Improved user-facing warnings/errors around layering and faceting (#1535).

  • data argument is now properly handled by Chart.properties (#1525)

  • Compound charts (layer, concat, hconcat, vconcat) now move data to the top level by default. In particular, this means that the facet() method can now be called directly on a layered chart without having to change how data is specified. (#1521)

  • alt.LayerChart now supports mark_*() methods. If a layer specifies a mark at the top level, all child charts will inherit it (unless they override it explicitly).

  • alt.Chart.facet() now handles wrapped facets; for example: python   chart.facet('column_name', columns=5) See altair/examples/us_population_over_time_facet.py for a more complete example.

Bug fixes#

  • Make chart.serve() and chart.save() respect the data transformer setting (#1538)

  • Fixed a deserialization bug for certain chart specs in schemapi (#1543)

Backward-Incompatible Changes#

  • alt.Chart.facet() now accepts a wrapped facet encoding as a first positional argument, rather than a row encoding. The following are examples of old invocations, and the equivalent new invocations:

  • chart.facet(row='col1', column='col2'): unchanged

  • chart.facet('col1', 'col2'): change to chart.facet(row='col1', column='col2')

  • chart.facet('col1'): change to chart.facet(row='col1')

In each case, the new invocations are compatible back to Altair 2.X.

  • Several of the encoding channels added in 3.0 have had their capitalization corrected to better match the names used in the schema:

  • alt.Fillopacity -> alt.FillOpacity

  • alt.Strokeopacity -> alt.StrokeOpacity

  • alt.Strokewidth -> alt.StrokeWidth

  • alt.Xerror -> alt.XError

  • alt.Xerror2 -> alt.XError2

  • alt.Yerror -> alt.YError

  • alt.Yerror2 -> alt.YError2

Version 3.0.1 (Released May 1, 2019)#

Fix version info bug for HTML output and Colab & Kaggle renderers.

Version 3.0.0 (Released April 26, 2019)#

Update to Vega-Lite 3.2 and Vega 5.3 & support all new features. See vega/vega-lite for Vega-Lite feature lists.

Highlights:#

  • new compound marks: mark_boxplot(), mark_errorband(), mark_errorbar()

  • new transforms: transform_impute(), transform_joinaggregate(), transform_flatten() transform_fold(), transform_sample(), transform_stack()

  • new facet encoding that is similar to the row and column encoding, but allows for wrapped facets

  • new alt.concat() function that is similar to alt.hconcat and alt.vconcat, but allows for more general wrapped concatenation

  • new columns keyword that allows wrapped faceting, repeating, and concatenation.

  • many, many bug fixes

  • tooltips can now be automatically populated using the tooltip mark configuration.

  • ability to specify initial conditions for selections

Version 2.4.1 (Released February 21, 2019)#

Enhancements#

  • Several documentation cleanups & new examples

Bug Fixes#

  • Fix incompatibility with pandas version 0.24 (#1315)

Version 2.3.0 (Released December 7, 2018)#

Includes many reworked examples in the example gallery.

Enhancements#

  • Better errors for non-string column names, as well as automatic conversion of pandas.RangeIndex columns to strings (#1107)

  • Renderers now have set_embed_options() method (#1203)

  • Added kaggle renderer & more HTML output options (#1123)

Backward-incompatible changes#

Maintenance#

  • fix typing requirement in Python 3.6+ (#1185)

  • Added support & CI testing for Python 3.7 (#1008)

Bug fixes#

  • Selection predicates now recognize all valid entries (#1143)

  • Python 2 support for chart.save() (#1134)

Version 2.2.2 (Released August 17, 2018)#

Bug Fixes#

  • fix missing JSON resource in altair.vega.v4 (#1097)

Version 2.2.1 (Released August 15, 2018)#

Bug Fixes#

  • appropriate handling of InlineData in dataset consolidation (#1092)

  • fix admonition formatting in documentation page (#1094)

Version 2.2.0 (Released August 14, 2018):#

Enhancements#

  • better handling of datetimes and timezones (#1053)

  • all inline datasets are now converted to named datasets and stored at the top level of the chart. This behavior can be disabled by setting alt.data_transformers.consolidate_datasets = False (#951 & #1046)

  • more streamlined shorthand syntax for window transforms (#957)

Maintenance#

Backward-incompatible changes#

  • alt.SortField renamed to alt.EncodingSortField and alt.WindowSortField renamed to alt.SortField (#3741)

Bug Fixes#

  • Fixed serialization of logical operands on selections within transform_filter(): (#1075)

  • Fixed sphinx issue which embedded chart specs twice (#1088)

  • Avoid Selenium import until it is actually needed (#982)

Version 2.1.0 (Released June 6, 2018):#

Enhancements#

  • add a scale_factor argument to chart.save() to allow the size/resolution of saved figures to be adjusted. (#918)

  • add an add_selection() method to add selections to charts (#832)

  • add chart.serve() and chart.display() methods for more flexibility in displaying charts (#831)

  • allow multiple fields to be passed to encodings such as tooltip and detail (#830)

  • make timeUnit specifications more succinct, by parsing them in a manner similar to aggregates (#866)

  • make to_json() and to_csv() have deterministic filenames, so in json mode a single datasets will lead to a single on-disk serialization (#862)

Breaking Changes#

  • make data the first argument for all compound chart types to match the semantics of alt.Chart (this includes alt.FacetChart, alt.LayerChart, alt.RepeatChart, alt.VConcatChart, and alt.HConcatChart) (#895).

  • update vega-lite to version 2.4.3 (#836)

  • Only API change is internal: alt.MarkProperties is now alt.MarkConfig

Maintenance#

  • update vega to v3.3 & vega-embed to v3.11 in html output & colab renderer (#838)

Version 2.0.0: May 2, 2018#

  • Complete rewrite of Altair, focused on supporting Vega-Lite 2.X

Version 1.2.1: October 29, 2017#

This version of Altair is based on Vega-Lite 1.2.1.

Major additions#

  • Support for JupyterLab/nteract through MIME based rendering. Enable this by calling enable_mime_rendering() before rendering visualizations (#216).

  • Change default import in all code and docs to import altair as alt

  • Check for missing and misspelled column names upon exporting or rendering, and raise FieldError (#399) if any problems are found. This can be disabled by setting Chart.validated_columns=False.

  • Raise MaxRowsExceeded if the number of rows in the dataset is larger than Chart.max_rows to guard against sending large datasets to the browser.

  • Move the Vega-Lite 1.x api into altair.v1 to make it easier for us to migrate to Vega-Lite 2.x and continue to support 1.x. No import change are needed as altair.v1 is aliased to altair in this releasealtair.v1 (#377).

  • Moved the example notebooks into a separate repository (https://github.com/altair-viz/altair_notebooks) that has Binder support (#391).

  • Add $schema to top-level JSON spec (#370).

  • Minor documentation revisions.

Bug fixes#

  • Make sure default mark is a point (#344).

Version 1.2: Nov 7, 2016#

Major additions#

Bug fixes#

  • Countless minor bug fixes

maintenance:#

  • Update to Vega-Lite 1.2.1 and add its supported features

  • Create website: http://altair-viz.github.io/

  • Set up Travis to run conda & pip; and to build documentation

Version 1.0: July 11, 2016#

  • Initial release of Altair