Fix transaction_primitives on Snowflake and BigQuery - #2420
Conversation
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
![chatgpt-codex-connector[bot]](/s/com/githubusercontent/avatars/G.https/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: f1aa2d2f1f
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
dfbb429 to
d8eb5cf
Compare
|
Maintainer review: refreshed onto develop (merge de57ca6) to resolve the compat-doc conflicts; the skip-reference and capability-matrix are regenerated from the merged registry. The BigQuery temp-table skip stands per the live API evidence in b0cdac8, and the Snowflake txn rewrite is unchanged. Zero live unresolved threads. CI re-running on the refreshed head; enqueuing after green. |
de57ca6 to
1b19dbf
Compare
Snowflake and BigQuery lack unnest(generate_series), SAVEPOINT, and SET TRANSACTION ISOLATION LEVEL, so rewrite the catalog integer-series generator to FLATTEN(ARRAY_GENERATE_RANGE) and UNNEST(GENERATE_ARRAY) and register documented execution-filter skips. Repair the BigQuery transaction_with_cte catalog shape (INSERT with a leading WITH clause) and type bare decimal literals as NUMERIC so inserts into money and discount columns no longer fail with FLOAT64 errors. Keep script temp tables unqualified during batch qualification. Live evidence: txn SF0.01 Snowflake 18 pass plus 5 skips; BigQuery 17 pass plus 5 skips with 1 remaining temp-table failure under investigation.
The adapter submits the whole translated script in one connection.query call where script-scoped temporary tables stay visible, and batch qualification keeps temp-table references unqualified. Remove the execution-filter skip so the operation is exercised, and regenerate the compat docs.
Reverts the unskip: live BigQuery API calls just proved a TEMP TABLE created inside BEGIN TRANSACTION is invisible to later statements (Not found: Table ..._script....t_verify_1), and the txn staging dataset has no TXN_ORDERS table. The skip stands.
878791b to
746bafc
Compare
1 participant
Snowflake and BigQuery lack unnest(generate_series), SAVEPOINT, and
SET TRANSACTION ISOLATION LEVEL, so rewrite the catalog integer-series
generator to FLATTEN(ARRAY_GENERATE_RANGE) and UNNEST(GENERATE_ARRAY)
and register documented execution-filter skips. Repair the BigQuery
transaction_with_cte catalog shape (INSERT with a leading WITH clause)
and type bare decimal literals as NUMERIC so inserts into money and
discount columns no longer fail with FLOAT64 errors. Keep script temp
tables unqualified during batch qualification.
Live evidence: txn SF0.01 Snowflake 18 pass plus 5 skips; BigQuery 17
pass plus 5 skips with 1 remaining temp-table failure under
investigation.