diff --git a/README.md b/README.md index e10f01f..9723a37 100644 --- a/README.md +++ b/README.md @@ -16,9 +16,11 @@ Arrow-native ADBC driver — columnar record batches out, bulk ingest in — fro Rust, Go, Java, C#, R and anything else that speaks the ADBC driver manager. **Integrated in [omniload](https://github.com/panodata/omniload):** the dlt-based loader ships an -`adbcbridge` SQL backend since its [0.17 release](https://github.com/panodata/omniload/releases) -(`omniload ingest --sql-backend adbcbridge`), requested, reviewed and merged by its maintainers on -2026-09-17. [How it was built and tested](https://adbcbridge.org/notes/omniload-adbcbridge-backend/) · +`adbcbridge` SQL backend since [v0.17.0](https://github.com/panodata/omniload/releases/tag/v0.17.0), +released 2026-09-17 — `pip install "omniload[adbcbridge]"`, then +`omniload ingest --sql-backend adbcbridge`. Requested, reviewed and merged by its maintainers the +same day, and carried by every release since. It is a declared extra of the published package +(`adbcbridge>=0.1.3,<0.2`), and of omniload's own `full` and `test` extras. [How it was built and tested](https://adbcbridge.org/notes/omniload-adbcbridge-backend/) · [omniload's handbook page](https://omniload.readthedocs.io/getting-started/sql-backends.html). Site and docs: · Launch write-up with the numbers: diff --git a/scripts/bump_version.py b/scripts/bump_version.py index 98fbdfb..8b4620f 100755 --- a/scripts/bump_version.py +++ b/scripts/bump_version.py @@ -120,12 +120,21 @@ def looks_like_ours(text, pos): 19.0.0, Maven plugins ...). Ours sit on a line that names the package -- a wheel, jar or nupkg file name, ``adbcbridge = "..."``, ``AdbcBridge --version`` -- or on the line right after ``adbcbridge`` in a Maven snippet. + + A markdown link decides on its own, whatever else the line says: a version inside + ``[v0.17.0](https://github.com/panodata/omniload/releases/tag/v0.17.0)`` belongs to + the project the link points at, not to us. Without that, a sentence naming our own + backend *and* citing the host project's release -- which the omniload and dlt + integration lines do -- would be read as an adbcbridge version. """ start = text.rfind("\n", 0, pos) + 1 prev_start = text.rfind("\n", 0, max(start - 1, 0)) + 1 end = text.find("\n", pos) line = text[start:end if end != -1 else len(text)].lower() prev = text[prev_start:start].lower() + for m in re.finditer(r"\[[^\]]*\]\(([^)]*)\)", line): + if m.start() <= pos - start < m.end(): + return "adbcbridge" in m.group(1) if "adbcbridge" in line: return True return "adbcbridge" in prev and "" in line