|
Download README.md from DataDrivenAlex/concurrency-control-lab: direct link, hf CLI and curl.
- Browser
- Download file 1.53 kB
-
https://huggingface.co/spaces/DataDrivenAlex/concurrency-control-lab/resolve/main/README.md
- Command line
-
hf download hf://spaces/DataDrivenAlex/concurrency-control-lab/README.md
-
curl -L -o README.md https://huggingface.co/spaces/DataDrivenAlex/concurrency-control-lab/resolve/main/README.md
1.53 kB
| title: Concurrency Control Lab | |
| emoji: 🔀 | |
| colorFrom: blue | |
| colorTo: indigo | |
| sdk: static | |
| pinned: false | |
| # Concurrency Control Lab | |
| A hands-on simulator for a classic backend bug: two requests read the same record version, then overwrite each other's updates. Compare last-write-wins with optimistic locking and see when a retry preserves the intended edits. | |
| ## Try the demo | |
| Open the Space and change the initial value, client edits, commit order, or conflict-retry setting. The demo shows the version checks, accepted writes, conflicts, retries, and final value. The example is deterministic so the same request schedule always produces the same trace. | |
| ## Run the Python model | |
| ```bash | |
| python -m unittest -v | |
| ``` | |
| `engine.py` models a set of edits that all read one initial snapshot. Last-write-wins commits each stale computed value, illustrating lost updates. Optimistic locking accepts a write only when its base version matches the current version; a stale write is rejected, and an optional retry re-reads and recomputes against the latest value. | |
| ## Scope and limitations | |
| This is an educational, in-memory simulation, not a database benchmark or production concurrency library. It assumes integer deltas, a single record, a deterministic commit order, and no network failures. Real systems must define transaction isolation, retry limits/backoff, idempotency, validation, and user-facing conflict behavior for their database and workload. The browser demo is static and does not call a backend or persist data. | |