A release for large datasets and shared servers, with a simpler way to load panel sets and a calmer layout.
Large datasets
- Plots are budgeted from the browser's memory, measured per point, up to 200 million points (the largest tested). Positions are kept as typed arrays, which removes the out-of-memory crash at 200M.
- Recolouring a 1M-cell plot no longer goes through Plotly's colour strings: the redraw after a gene recolour went from 1.13 s to 0.38 s in our profile (#82).
- Automatic point size and opacity follow the points in view when you zoom or pan (#85).
- Subset reads, numeric columns and sparse gene columns are read without copying whole columns on the server (#77).
- Any number of categories can be coloured: up to 64 colour groups, each category keeping its colour, the last group covering the rest; the first ten colours are clearly distinct. Above about 500,000 labels the Hover picker starts at "No hover"; picking a hover column loads the labels. The Equals filter lists the categories.
Panel sets and layout
- Load Panel Set: one blue Load restores the set as a share link would (its dataset and its open panels, in their layout). Two icons do less: load on the current dataset, or only add the panels to the closed list. A badge says whether the set's dataset is on this server; if not, Load uses the open dataset and a picker chooses another. The ? explains the three options.
- Saved links and panel sets open on any deployment that has the same store, and figures export reproducibly.
- Close all closes every panel. Optionally it also clears what this site stored in the browser and reloads as a first visit; saved panel sets, datasets and your login are not touched.
- Panels grow with their content instead of scrolling inside; panels in a row share its height. Only table rows, legends and gene-set lists scroll. The page scroll takes over when an inner list reaches its end.
- Tables fill their panel with the pager at the bottom, and the search row sits below the advanced SearchBuilder. Numeric columns with more than 10,000 values keep their number conditions.
- A cell table's text filters (contains, starts with, ends with, empty, AND/OR groups) can define a subset of every cell.
Shared servers
- Refresh no longer holds a server worker while it waits for the rate limit: it answers at once and the browser collects the result. gunicorn runs threaded workers (#83).
- Requests are held to the data directories; admins may open any path by default (
server.arbitrary_paths). A link like?dataset_path=name.zarropens the listed dataset under its own name.
Desktop apps
- The Linux packages contain no GPL libraries; every release file is licence-checked, and third-party notices are shipped with the apps.
Other
- Citation metadata for Zenodo and GitHub (
.zenodo.json,CITATION.cff). - Gene Set Analysis says which way a service failed; a rerun keeps the chosen category. Fixes to colour maps, the log toggle, search, focus columns and button corners.
Install
pip install annzarro==0.4.1 (PyPI), or a desktop app below.
Desktop apps
Desktop apps for Windows, macOS and Linux. The builds are not signed with a developer identity (the macOS app is ad-hoc signed, not notarized), so the first launch needs one extra step. macOS 15 and 26: macOS says "AnnZarro" Not Opened; click Done, then System Settings > Privacy & Security > Security > Open Anyway, and confirm (once). macOS 14 and earlier: right-click the app > Open. Windows: More info > Run anyway. Details: https://annzarro.readthedocs.io/en/latest/getting-started/desktop-app.html (First launch). Verify downloads against SHA256SUMS.txt.
The .json and .txt files list the licences of everything bundled in each package.