Skip to content

fix(ci): run full-fragment merge_insert benchmarks on v2 only #8260

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Draft
wjones127 wants to merge 1 commit into
base: main
Choose a base branch
from
Draft

fix(ci): run full-fragment merge_insert benchmarks on v2 only #8260

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
39 changes: 31 additions & 8 deletions python/python/ci_benchmarks/benchmarks/test_merge_insert.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -367,6 +367,11 @@ def test_upsert_unindexed_baseline(benchmark, narrow: Target, num_rows: int) ->
ROW_FRACTIONS = [0.001, 0.01, 0.1, 1.0]
FRACTION_IDS = ["0.1pct", "1pct", "10pct", "100pct"]

# The projection sweeps only need the two ends of that range: one fraction where
# the updater interleaves and one where it rewrites whole column files.
PROJECTION_FRACTIONS = [0.01, 1.0]
PROJECTION_FRACTION_IDS = ["1pct", "100pct"]

# Isolates the two pressures a partial-column update can exert: field count
# (scalar columns, cheap per field) and byte volume (the vector column).
PROJECTIONS = {
Expand All @@ -383,6 +388,24 @@ def _wide_rounds(fraction: float) -> int:
return 1 if fraction == 1.0 else 3


def wide_plan_cases(fractions: Sequence[float], ids: Sequence[str]):
"""``(fraction, plan)`` cases, minus the ones v1 cannot execute.

At ``fraction == 1.0`` the source covers every row of the target, and the
legacy path builds its hash join on the source side: it asks for the whole
source at once -- ~1 GB once the 256-dim vector column is projected --
against a 150 MiB pool, and fails with "Resources exhausted". That is the
same limitation that keeps ``test_upsert_source_equals_target`` v2-only, so
the full-fragment shapes report a v2 number only. The 10pct fraction still
gives a v1-vs-v2 write-amplification comparison at fragment scale.
"""
return [
pytest.param(fraction, plan, id=f"{fraction_id}-{plan}")
for fraction, fraction_id in zip(fractions, ids)
for plan in (["v2_hash"] if fraction == 1.0 else PLANS)
]


def update_subset(
source: pa.Table, *, use_index: bool
) -> Callable[[lance.LanceDataset], ExecuteResult]:
Expand All @@ -399,8 +422,7 @@ def job(dataset: lance.LanceDataset) -> ExecuteResult:
return job


@pytest.mark.parametrize("plan", PLANS)
@pytest.mark.parametrize("fraction", ROW_FRACTIONS, ids=FRACTION_IDS)
@pytest.mark.parametrize("fraction,plan", wide_plan_cases(ROW_FRACTIONS, FRACTION_IDS))
def test_update_subset_row_fraction(
benchmark, wide: Target, fraction: float, plan: str
) -> None:
Expand All @@ -420,8 +442,9 @@ def test_update_subset_row_fraction(
)


@pytest.mark.parametrize("plan", PLANS)
@pytest.mark.parametrize("fraction", [0.01, 1.0], ids=["1pct", "100pct"])
@pytest.mark.parametrize(
"fraction,plan", wide_plan_cases(PROJECTION_FRACTIONS, PROJECTION_FRACTION_IDS)
)
@pytest.mark.parametrize("projection", list(PROJECTIONS), ids=list(PROJECTIONS))
def test_update_subset_projection(
benchmark, wide: Target, projection: str, fraction: float, plan: str
Expand All @@ -438,8 +461,9 @@ def test_update_subset_projection(
)


@pytest.mark.parametrize("plan", PLANS)
@pytest.mark.parametrize("fraction", [0.01, 1.0], ids=["1pct", "100pct"])
@pytest.mark.parametrize(
"fraction,plan", wide_plan_cases(PROJECTION_FRACTIONS, PROJECTION_FRACTION_IDS)
)
def test_upsert_wide_full_schema(
benchmark, wide: Target, fraction: float, plan: str
) -> None:
Expand Down Expand Up @@ -714,8 +738,7 @@ def test_io_mem_upsert_ratio(


@pytest.mark.io_memory_benchmark()
@pytest.mark.parametrize("plan", PLANS)
@pytest.mark.parametrize("fraction", ROW_FRACTIONS, ids=FRACTION_IDS)
@pytest.mark.parametrize("fraction,plan", wide_plan_cases(ROW_FRACTIONS, FRACTION_IDS))
def test_io_mem_update_subset_row_fraction(
io_mem_benchmark, wide: Target, fraction: float, plan: str
) -> None:
Expand Down
Loading