2.4.7

New in this release:

1. Log PointCloud() data
2. Export Images and PointCloud data to Comet

2.4.5

New in this release:

1. Don't crash if an image doesn't have annotations

2.4.4

New in this release:

1. Added import from wandb:

kangas import --wandb stacey/mnist-viz/baseline:v4 mnist.datagrid

2.4.3

New in this release:

1. Removed kangas server --root flag: was not used
2. Changed the remote strategy for exported files (need more work to make general)
3. Updated copyright dates
4. Set rows from command line
5. Added ability to see large text fields
6. Don't crash if image is None

2.4.2

New in this release:

1. Added UMAP projection, with docs (use Embedding(projection="umap")); requires pip install umap-learn
2. Remove status check on debounce for filter bar, in UI
3. Put a timeout on downloads (any path that is http or https)
4. Fixed two bugs in dg.read_csv(): handle no headers, if converted raises, use None value
5. Adjust the size of points in thumbnail scatter plots of embeddings
6. Updated notebook Visualizing_embeddings_in_Kangas.ipynb
7. Added Embedding(scale=True) for "umap" and "pca"

2.4.1

New in this release:

1. Pre-compute all embedding points in projection space
2. All kwargs to pass to projection from Embedding()
3. Allow PCA projections to not include all patterns for fit; use Embedding(include=False) for a row

2.4.0

New in this release:

1. New auto dynamic update: a polling system that runs automatically for all DataGrid UIs, checking every 15 seconds if the underlying database has changed. If it has, then it smoothly will update the UI (unless a dialog window, like editing computed columns, is open).
2. Removed the timestamp from the URL
3. Added new expression function: HISTOGRAM({"Column Name"}, 10) that will return a number for that row representing the value in "Column Name" broken into 10 bins. Useful to group by this value to see the distribution of each bar in a histogram chart.
4. Added selectable columns in all Python query APIs. By default, one gets all columns except "row-id". Use select="*" to get all columns, including "row-id". Otherwise, use select=["Column Name1", ...]
5. Spellcheck turned off (again) in computed column editor
6. Row-id cell shows "N/A" when grouping on another column.
7. "About" markdown now automatically turns URLs into links
8. Fixed a bug in "kangas upgrade" and "dg.upgrade()"
9. Fixed a bug in appending data of wrong type (that was in the Getting Started notebook)
10. Fixed loading spinner from flashing, and changing view
11. Refactored embedding projector to handle large embedding vectors

2.3.7

New since last release:

1. Updated all starting URLs to use datagrid name and timestamp
2. Updated all backend endpoints to include timestamp for proper caching
3. Dynamic updates is work-in-progress
4. Floating point cells are color-mapped. Soon you will be able to pick colormap, or disable, per column
5. Values in text columns that have category data now show as colored tags
6. Fixes for running in Kaggle notebooks; uses ngrok. show() now support use_ngrok option in all environments; requires ngrok api token
7. All charts (histogram bar, category bar, and embedding projection scatter) now support mouse click or drag to select items in the chart; usable in filter or in computed column expressions

2.3.4

New in this release:

1. Support for running in Kaggle Notebooks (requires pip install pyngrok)
2. CSS fixes for cells
3. Added DataGrid.set_about_from_script(filename)
4. Various bug fixes

2.3.3

New since last published release:

1. Loading indicator for longer load times
2. Added DataGrid.remove_rows()
3. Add completions to computed columns expression editor
4. Copy and paste from charts for filter or computed columns
5. Notify user with important display messages
6. New santa example notebook
7. Various fixes and enhancements