Skip to content

Keep the collector out of the linear-scan test's clock #1429

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

Closed
MaxGhenis wants to merge 1 commit into from
Closed

Keep the collector out of the linear-scan test's clock #1429

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
130 changes: 90 additions & 40 deletions packages/microcosm-build/tests/engine_free/shared/test_spec_source_citations.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -59,6 +59,7 @@
from __future__ import annotations

import ast
import gc
import importlib
import inspect
import json
Expand Down Expand Up @@ -3250,21 +3251,73 @@ def test_a_line_link_of_this_repository_is_refused_however_markup_escapes_it(
}


def _assert_scans_quickly(text: str, *scans: Callable[[str], object]) -> None:
"""Fail when ``scans`` of ``text`` take 2 s of CPU time between them.

The collector stays off while the clock runs. On CI, one full collection of
a worker's heap took 1.1 to 1.2 s, which says nothing about a scan, and a
slow runner took twice as long as a fast one over the same scans. So each
text gets its own 2 s, and only the scans count against it.
"""
collecting = gc.isenabled()
gc.disable()
try:
started = time.process_time()
for scan in scans:
scan(text)
spent = time.process_time() - started
finally:
if collecting:
gc.enable()
assert spent < 2, text[:24]


def test_the_scan_clock_stops_the_collector_and_restores_it() -> None:
states = []
_assert_scans_quickly("x", lambda _: states.append(gc.isenabled()))
assert states == [False]
assert gc.isenabled()

def fail(_: str) -> None:
raise ValueError("scan failed")

with pytest.raises(ValueError, match="scan failed"):
_assert_scans_quickly("x", fail)
assert gc.isenabled()

gc.disable()
try:
_assert_scans_quickly("x", str)
assert not gc.isenabled()
finally:
gc.enable()


@pytest.mark.parametrize(("spent", "fails"), [(1.99, False), (2.0, True)])
def test_the_scan_clock_refuses_2_s(
monkeypatch: pytest.MonkeyPatch, spent: float, fails: bool
) -> None:
monkeypatch.setattr(time, "process_time", iter((0.0, spent)).__next__)
if fails:
with pytest.raises(AssertionError, match="^long text"):
_assert_scans_quickly("long text", str)
else:
_assert_scans_quickly("long text", str)


def test_a_long_string_is_scanned_in_linear_time() -> None:
# Each of these once took time quadratic in its length: a run of path
# characters was rescanned from each of its characters, and a user was
# looked for before a GitHub host from each character of a run that can
# be in one. At these lengths the quadratic scans took seconds each.
started = time.process_time()
for text in (
":" * 60_000,
"//" * 30_000,
"a" * 60_000,
"github.com/" * 6_000,
"https://a/" + "%74" * 20_000,
):
line_citations(text)
unrecognised_line_locators(text)
_assert_scans_quickly(text, line_citations, unrecognised_line_locators)
# So did a string holding thousands of citations: each one's clause,
# commits and commit-like tokens were found again from the whole string.
for text in (
Expand All @@ -3273,10 +3326,12 @@ def test_a_long_string_is_scanned_in_linear_time() -> None:
"a1b2c3d; " + "x.py:1; " * 8_000,
"x.py:1 y.py foo L1 " * 4_000,
):
misspelled_commit_pins(text)
unpinned_line_citations(text)
unrecognised_line_locators(text)
assert time.process_time() - started < 2
_assert_scans_quickly(
text,
misspelled_commit_pins,
unpinned_line_citations,
unrecognised_line_locators,
)
# Each of these took 4 to 28 seconds: a list of line numbers that failed
# was retried once for each way to share its whitespace out (exponential
# in its length), a failed locator once for each way to share a run of
Expand All @@ -3299,9 +3354,7 @@ def test_a_long_string_is_scanned_in_linear_time() -> None:
"L1 and " * 8_000,
"https://" + "".join(map(chr, range(0x4000, 0x4000 + 8_000))) + "/a.py#L12",
):
line_citations(text)
unrecognised_line_locators(text)
assert time.process_time() - started < 2, text[:24]
_assert_scans_quickly(text, line_citations, unrecognised_line_locators)
# Since round 9: ranges that each end where the next starts, a list read
# again from its last number, and runs of what a markup escape starts with.
for text in (
Expand All @@ -3312,28 +3365,22 @@ def test_a_long_string_is_scanned_in_linear_time() -> None:
"\\" * 100_000,
"&a" * 50_000,
):
line_citations(text)
unrecognised_line_locators(text)
assert time.process_time() - started < 2, text[:24]
# Since round 10, in 2 s of their own: runs of what an attribute's value,
# a destination in angle brackets, a long legacy-looking name and a URL
# read through Unicode spaces start with.
started = time.process_time()
_assert_scans_quickly(text, line_citations, unrecognised_line_locators)
# Since round 10: runs of what an attribute's value, a destination in angle
# brackets, a long legacy-looking name and a URL read through Unicode spaces
# start with.
for text in (
'="https:' * 20_000,
"='" + "x" * 100_000,
"](<https:" * 20_000,
"](<https:" + "\\x" * 50_000,
("&amp" + "x" * 28) * 3_000,
"https:" + "\u00a0" * 100_000,
"https://a\u00a0" * 20_000,
"https:" + " " * 100_000,
"https://a " * 20_000,
):
line_citations(text)
unrecognised_line_locators(text)
assert time.process_time() - started < 2, text[:24]
# Since round 11, in 2 s of their own: runs of what a dot segment, a URL
# read through its query and a destination after whitespace start with.
started = time.process_time()
_assert_scans_quickly(text, line_citations, unrecognised_line_locators)
# Since round 11: runs of what a dot segment, a URL read through its query
# and a destination after whitespace start with.
for text in (
"/.." * 30_000,
"=%2e." * 20_000,
Expand All @@ -3345,16 +3392,17 @@ def test_a_long_string_is_scanned_in_linear_time() -> None:
"](" + " " * 100_000 + "\n<",
"](\t \t<https:" * 15_000,
):
line_citations(text)
unrecognised_line_locators(text)
dot_segment_line_citations(text)
assert time.process_time() - started < 2, text[:24]
# Since round 12, in 2 s of their own: runs of what this repository's path,
# a line fragment, a run of slashes or ``.`` segments, a dot segment beside
# a ``%2f``, and the run of path characters before a name start with; and
# dot segments a browser resolves, after a ``?`` too, escaped or beside a
# backslash, which pop 10,000 segments back to the owner.
started = time.process_time()
_assert_scans_quickly(
text,
line_citations,
unrecognised_line_locators,
dot_segment_line_citations,
)
# Since round 12: runs of what this repository's path, a line fragment, a
# run of slashes or ``.`` segments, a dot segment beside a ``%2f``, and the
# run of path characters before a name start with; and dot segments a
# browser resolves, after a ``?`` too, escaped or beside a backslash, which
# pop 10,000 segments back to the owner.
for text in (
"policyengine/microcosm" * 2_500 + "#L1",
"PolicyEngine/microcosm/#L1" * 2_000,
Expand All @@ -3372,11 +3420,13 @@ def test_a_long_string_is_scanned_in_linear_time() -> None:
"%2e%2E/" * 9_000 + "#L1",
"\\.." * 20_000 + "#L1",
):
line_citations(text)
in_repository_line_citations(text)
dot_segment_line_citations(text)
line_links_of_this_repository(text)
assert time.process_time() - started < 2, text[:24]
_assert_scans_quickly(
text,
line_citations,
in_repository_line_citations,
dot_segment_line_citations,
line_links_of_this_repository,
)


#: Pieces of text around source names and line locators.
Expand Down
Loading