This commit is contained in:
Alexander Wang 2023-03-18 16:18:00 -07:00
parent cd47fc8dd7
commit 3b85f4d965
No known key found for this signature in database
GPG key ID: D89FA31966BDBECE

View file

@ -0,0 +1,34 @@
This release brings massive improvements in performance. Screenshot from our internal benchmarks on a batch of tests, comparing this release vs the last:
<img width="540" alt="Screen Shot 2023-03-18 at 4 03 09 PM" src="https://user-images.githubusercontent.com/3120367/226144882-c6746a65-1f60-4a55-8382-9e5ea2a686a2.png">
As a result of this, TALA by default now executes 3 runs in parallel with different seeds, and picking the best one for you. This should produce better results at a much higher rate!
#### Features 🚀
- You can now pass in multiple seeds, and TALA will race multiple layout runs with
different seeds, choosing the best outcome. See the TALA Manual for more.
#### Improvements 🧹
- Improves subgraph compaction giving space for better edges
- Improved edge routing for clustered nodes
- Significantly speeds up edge routing with many edges
- Significant performance improvements
- Uses up to 56% less memory for layouts with `sql_table`s
#### Bugfixes ⛑️
- Fixes a rare condition that allowed nodes with self loops to escape their containers
- Fixes edge from container going through descendants
- Fixes a bug that could cause some asymmetrical edge routing
- Fixes container shapes sometimes not having enough padding to fit icons
- Fixes the alignment of nodes connected to sequence steps
- Fixes the placement of clustered nodes connected to sequences
#### Breaking Changes
- `tala-seed` changed to `tala-seeds`. You can still continue to pass in a single number,
same as before. But the name change is reflective that it can now take in multiple
seeds.