Plotly

- `px.timeline()` now allows `hover_data` formatting of start and end times [3018](https://github.com/plotly/plotly.py/pull/3018)
- Small change to packaging of `plotlywidget` extension for JupyterLab 3 [3021](https://github.com/plotly/plotly.py/pull/3021)

Updated

- JupyterLab extensions now compatible with JupyterLab 3.x [3016](https://github.com/plotly/plotly.py/pull/3016)
- Updated Plotly.js to version 1.58.4. See the [plotly.js CHANGELOG](https://github.com/plotly/plotly.js/blob/v1.58.4/CHANGELOG.md) for more information. These changes are reflected in the auto-generated `plotly.graph_objects` module. Notable changes include:
- fixes for rendering 3d plots on recent Safari versions
- fixes to inside ticklabels
- regression fixes

Fixed

- `px.histogram()` Y-axis labels now take into account `histnorm` and `barnorm` [2989](https://github.com/plotly/plotly.py/pull/2989)
- `px.histogram()` `marginal` and `facet_*` now work correctly together [3014](https://github.com/plotly/plotly.py/pull/3014)

Updated

- Updated Plotly.js to version 1.58.2. See the [plotly.js CHANGELOG](https://github.com/plotly/plotly.js/blob/v1.58.0/CHANGELOG.md) for more information. These changes are reflected in the auto-generated `plotly.graph_objects` module. Notable changes include:
- fixes for new `ticklabelposition` attribute
- fixes for a regression related to treemaps in the previous version

Added

- `px.imshow` now supports `facet_col` and `animation_frame` arguments for visualizing 3-d and 4-d images [2746](https://github.com/plotly/plotly.py/pull/2746)
- `px.defaults` now supports `color_discrete_map`, `symbol_map`, `line_dash_map`, `labels` and `category_orders` as well as a `.reset()` method [2957](https://github.com/plotly/plotly.py/pull/2957)

Fixed

- axes will now auto-type numeric strings as categorical data rather than linear in the default templates [2951](https://github.com/plotly/plotly.py/pull/2951)

Updated

- Updated Plotly.js to version 1.58.1. See the [plotly.js CHANGELOG](https://github.com/plotly/plotly.js/blob/v1.58.0/CHANGELOG.md) for more information. These changes are reflected in the auto-generated `plotly.graph_objects` module. Notable changes include:
- a new `ticklabelposition` attribute to enable positioning tick labels inside the plotting area
- better support for `scaleanchor` and `matches` on cartesian axes for matched square subplots
- a new `autotypenumbers` attribute which is now set to `strict` in the default templates
- various fixes relating to `automargins` for small figures

Added
- `px.choropleth`, `px.scatter_geo` and `px.line_geo` now support faceting as well as `fitbounds` and `basemap_visible` [2923](https://github.com/plotly/plotly.py/pull/2923)
- `px.scatter_geo` and `px.line_geo` now support `geojson`/`featureidkey` input [2923](https://github.com/plotly/plotly.py/pull/2923)
- `px.scatter_geo` now supports `symbol` [2923](https://github.com/plotly/plotly.py/pull/2923)
- `go.Figure` now has a `set_subplots` method to set subplots on an already
existing figure. [2866](https://github.com/plotly/plotly.py/pull/2866)
- Added `Turbo` colorscale and fancier swatch display functions
[2882](https://github.com/plotly/plotly.py/pull/2882)
- A utility function `image_array_to_data_uri` has been added in
`plotly.utils`, in order to transform NumPy arrays to data b64 URIs (which
can be passed to the source parameter of `go.Image`, or to layout images).
[2879](https://github.com/plotly/plotly.py/pull/2879)
- the `selector` argument to updater/selector functions now accepts `int`s and `str`s
[2894](https://github.com/plotly/plotly.py/pull/2894)


Updated

- the JSON serialization of plotly figures has been accelerated thanks to a
different handling of Infinity and NaN values. For example, a figure with a
1000x1000 Heatmap should now serialize 2x faster. [2880](https://github.com/plotly/plotly.py/pull/2880)
- Coding mistakes with "magic underscores" now return significantly more ergonomic error
messages [2843](https://github.com/plotly/plotly.py/pull/2843)
- Error messages related to impossible subplot geometries are now much more helpful
[2897](https://github.com/plotly/plotly.py/pull/2897)


Fixed

- `px.scatter_geo` support for `text` is fixed [2923](https://github.com/plotly/plotly.py/pull/2923)
- the `x` and `y` parameters of `px.imshow` are now used also in the case where
an Image trace is used (for RGB data or with `binary_string=True`). However,
only numerical values are accepted (while the Heatmap trace allows date or
string values for `x` and `y`). [2761](https://github.com/plotly/plotly.py/pull/2761)

Added

- For `add_trace`, `add_shape`, `add_annotation` and `add_layout_image`, the `row` and/or `col` argument now also accept the string `"all"`. `row="all"` adds the object to all the subplot rows and `col="all"` adds the object to all the subplot columns. ([2840](https://github.com/plotly/plotly.py/pull/2840))
- Shapes that reference the plot axes in one dimension and the data in another dimension can be added with the new `add_hline`, `add_vline`, `add_hrect`, `add_vrect` functions, which also support the `row="all"` and `col="all"` arguments. ([2840](https://github.com/plotly/plotly.py/pull/2840))
- The `add_trace`, `add_shape`, `add_annotation`, `add_layout_image`, `add_hline`, `add_vline`, `add_hrect`, `add_vrect` functions accept an argument `exclude_empty_subplots` which if `True`, only adds the object to subplots already containing traces or layout objects. This is useful in conjunction with the `row="all"` and `col="all"` arguments. ([2840](https://github.com/plotly/plotly.py/pull/2840))
- For all `go.Figure` functions accepting a selector argument (e.g., `select_traces`), this argument can now also be a function which is passed each relevant graph object (in the case of `select_traces`, it is passed every trace in the figure). For graph objects where this function returns true, the graph object is included in the selection. ([2844](https://github.com/plotly/plotly.py/pull/2844))

Added

- Better magic underscore error messages. For example, `some_fig.update_layout(geo_ltaxis_showgrid=True)` shows `Bad property path:\ngeo_ltaxis_showgrid\n ^` and lists the valid properties for `geo`.

Updated

- Updated Plotly.js to version 1.57.1. See the [plotly.js CHANGELOG](https://github.com/plotly/plotly.js/blob/v1.57.1/CHANGELOG.md) for more information. These changes are reflected in the auto-generated `plotly.graph_objects` module.