[LP Bispo (phelipelui)]

Work Item / Issue Reference

GitHub Issue: #863

---

Summary

Cursor.nextset() calls SQLMoreResults and refreshes description and messages, but it leaves rowcount at the value execute() set for the first statement. In a multi-statement batch, the affected-row count of every later INSERT/UPDATE/DELETE was lost. This hits hardest with T-SQL control flow: a WHILE loop sends one result per INSERT, so a tool that sums rowcount over nextset() gets a stale number.

Change (mssql_python/cursor.py, nextset()). After the describe step, rowcount is set the same way execute() sets it:

• -1 when the new result has columns; fetch*() then updates it as it already does;
• DDBCSQLRowCount(self.hstmt) otherwise.

The SQL_NO_DATA path (nextset() returning False) is unchanged. This matches pyodbc, which calls SQLRowCount in Cursor_nextset after SQLMoreResults.

Tests (tests/test_004_cursor.py):

• test_nextset_updates_rowcount runs a batch of INSERT (1 row), INSERT (2 rows), SELECT, UPDATE (3 rows), and DELETE (0 rows). It checks rowcount after each nextset(): 1, 2, -1, 3, 0.
• test_nextset_rowcount_in_loop_batch runs a WHILE loop whose iterations insert 1, 2 and 3 rows. It expects per-result counts [1, 2, 3]. The counts differ on purpose, so a stale rowcount cannot reach the right total by accident.

CHANGELOG: entry under Unreleased → Fixed.

Verification. I ran against SQL Server 2022 (16.0.4305.5) on Python 3.12, Debian 12, with the native bindings from the published 1.16.0 wheel, whose cursor.py is identical to main before this change.

tests/test_004_cursor.py	Without the fix	With the fix
New tests	2 failed (assert 1 == 2, [1, 1, 1] == [1, 2, 3])	2 passed
Whole file	2 failed, 544 passed, 4 skipped	546 passed, 4 skipped

The only failures without the fix are the two new tests, and no other test changes result. Black 26.5.1, the version pinned in .pre-commit-config.yaml, reports both files unchanged.

Out of scope: AsyncCursor.nextset() goes through py-core and was not changed or checked.

🤖 Generated with Claude Code

[azure-pipelines]

Azure Pipelines:
There may be pipelines that require an authorized user to comment /azp run to run.

[github-actions]

PR Performance Report

✅ No regression detected

No consistent slowdowns detected across all 2 environments.

0 IMPROVEMENTS 0 SLOWDOWNS 2/2 ENVIRONMENTS

Coverage: 2 of 2 environments completed. Advisory result; does not block merging.

Performance diagnostics

Phase times are inclusive diagnostics and must not be added together. They identify where measured time changed, not why it changed.

No affected phases or call-count changes were recorded.

All database tasks and timings

Unix / SQL Server 2022

Database task	Before	After	Paired change	Result
Connection opening	10.859 ms	10.787 ms	-0.4%	no signal
SELECT queries	1.294 ms	1.091 ms	-15.0%	no signal
Row insertion	31.498 ms	31.482 ms	+0.2%	no signal
Executemany inserts	136.656 ms	138.094 ms	+0.3%	no signal
Fetch-all queries	116.997 ms	117.861 ms	+0.8%	no signal
Row-by-row fetching	12.695 ms	12.548 ms	-0.9%	no signal
Batched row fetching	110.367 ms	110.806 ms	+0.4%	no signal
Transaction commit and rollback	100.852 ms	99.766 ms	-1.1%	no signal
Arrow row fetching	90.241 ms	87.839 ms	-3.1%	no signal
100,000-row insertion	404.615 ms	400.005 ms	-0.6%	no signal
Row fetching in batches of 100	106.997 ms	107.648 ms	+0.2%	no signal
Row fetching in batches of 10,000	130.018 ms	118.466 ms	-8.1%	no signal
Repeated positional queries	30.134 ms	30.072 ms	-0.3%	no signal
Repeated named-parameter queries	32.281 ms	32.805 ms	+1.6%	no signal
Legacy 100,000-row insertion	308.605 ms	311.153 ms	+1.4%	no signal
Insertion with explicit input sizes	435.290 ms	440.170 ms	+1.1%	no signal
Joined aggregation queries	185.291 ms	183.791 ms	-0.5%	no signal
Large joined-result fetching	183.204 ms	178.662 ms	-2.0%	no signal
1.2-million-row fetching	3460.605 ms	3478.392 ms	+0.5%	no signal
Common table expression queries	5.761 ms	5.701 ms	-0.9%	no signal
256 KiB VARCHAR(MAX) / fetchall()	1.259 ms	1.313 ms	+2.8%	no signal
10,000 scalar values / fetchval() (debug disabled)	100.813 ms	100.513 ms	-0.9%	no signal

Unix / SQL Server 2025

Database task	Before	After	Paired change	Result
Connection opening	96.190 ms	96.277 ms	+0.6%	no signal
SELECT queries	1.249 ms	1.120 ms	-2.6%	no signal
Row insertion	32.216 ms	32.544 ms	+0.9%	no signal
Executemany inserts	131.874 ms	131.960 ms	+0.1%	no signal
Fetch-all queries	120.463 ms	121.775 ms	+1.5%	no signal
Row-by-row fetching	12.743 ms	12.686 ms	-1.1%	no signal
Batched row fetching	110.042 ms	112.625 ms	+0.3%	no signal
Transaction commit and rollback	104.451 ms	104.000 ms	-0.6%	no signal
Arrow row fetching	90.493 ms	88.964 ms	-0.5%	no signal
100,000-row insertion	396.350 ms	424.020 ms	+3.3%	no signal
Row fetching in batches of 100	107.127 ms	107.389 ms	+0.1%	no signal
Row fetching in batches of 10,000	136.806 ms	126.252 ms	-5.9%	no signal
Repeated positional queries	30.412 ms	30.494 ms	+1.1%	no signal
Repeated named-parameter queries	32.954 ms	32.978 ms	+0.1%	no signal
Legacy 100,000-row insertion	313.275 ms	334.676 ms	+5.9%	no signal
Insertion with explicit input sizes	438.347 ms	475.596 ms	+5.0%	no signal
Joined aggregation queries	164.407 ms	164.757 ms	+1.4%	no signal
Large joined-result fetching	196.313 ms	198.652 ms	+1.2%	no signal
1.2-million-row fetching	3466.679 ms	3485.073 ms	+1.0%	no signal
Common table expression queries	5.405 ms	5.406 ms	-1.5%	no signal
256 KiB VARCHAR(MAX) / fetchall()	1.360 ms	1.375 ms	+0.7%	no signal
10,000 scalar values / fetchval() (debug disabled)	100.412 ms	100.075 ms	-0.1%	no signal

Build and measurement details

ADO build 186626

PR head: 683a85f216bd64c13f13fe4a274d5bec1ad451ef
Base: 21187de03cb86e82a19c6363e5243d7b18daacc4
Measured merge: c3c77a76a0a3bd521c997e6a906175562cf16c18

• Unix / SQL Server 2022: Python 3.12.3, x86_64, SQL 16.0.4305.5; 5 paired comparisons and 1 warmup.
• Unix / SQL Server 2025: Python 3.12.3, x86_64, SQL 17.0.5035.5; 5 paired comparisons and 1 warmup.

A consistent change requires more than 20% median paired movement, at least 1 ms between the median runtimes, and at least 80% of pairs exceeding the relative threshold in the same direction. A slowdown without enough pair agreement is reported as inconsistent.

The displayed change is the median of paired before-and-after ratios. It is not recalculated from the two displayed median runtimes.

Both revisions use profiling-enabled builds on the same agent and database, with alternating order and discarded warmups. Results are diagnostic and do not represent production-wheel latency.

Raw samples and logs are attached to the ADO run as profiler-* artifacts.

[Copilot]

🔵 Needs a closer look

Reported integration results use published-wheel bindings; SQL Server validation against the current native build still needs human confirmation.

No open issues 🎉

What changed in this PR

Addresses #863 by refreshing the synchronous cursor’s rowcount when nextset() advances, aligning it with execute() behavior.

Changes:

• Reports affected-row counts for non-query results and resets rowcount to -1 for result sets.
• Adds regression tests for mixed-statement batches and loops with differing insert counts.
• Documents the fix in the changelog; asynchronous cursors remain unchanged.

File	Description
tests/​test_004_cursor.py	Tests per-result row counts in mixed batches and loops.
mssql_python/​cursor.py	Refreshes rowcount after advancing to a new result.
CHANGELOG.md	Records the fix for #863.

_{Review effort: Balanced}

[LP Bispo (phelipelui)]

@microsoft-github-policy-service agree

[Copilot]

🟢 Approval recommended

No active review issues are recorded for the synchronous cursor’s per-result row-count update.

No open issues 🎉

_{Review effort: Balanced}

[azure-pipelines]

Azure Pipelines:
Successfully started running 1 pipeline(s).

[Jahnvi Thakkar (jahnvi480)]

LGTM!

[Jahnvi Thakkar (jahnvi480)]

/azp run

[azure-pipelines]

Azure Pipelines:
Successfully started running 1 pipeline(s).

[Copilot]

🟢 Approval recommended

No active review issues were identified for the synchronous cursor row-count update.

No open issues 🎉

_{Review effort: Balanced}