perf: use yaml.CSafeLoader/CDumper for 4x faster virtual index merge
ci/woodpecker/pr/pre-commit Pipeline was successful
ci/woodpecker/pr/test Pipeline was successful
ci/woodpecker/pr/build Pipeline was successful

Replace yaml.safe_load / yaml.Dumper with the C extension equivalents
(yaml.CSafeLoader, yaml.CDumper) for all YAML parsing and serialisation
in the virtual repo merge path. A module-level try/except falls back to
the pure-Python SafeLoader/Dumper when libyaml is not available.

Measured on 19-member helm-all (real upstream data):
  Before (SafeLoader + Dumper): merge=38,877ms
  After  (CSafeLoader + CDumper): merge=9,625ms  (4.0x faster)

Local microbenchmark (500 charts × 10 versions × 19 members):
  Before: 40.8s  After: 6.1s  (6.7x faster)
This commit is contained in:
2026-05-02 01:50:21 +10:00
parent 7b6c69b70f
commit b7488fce10
2 changed files with 39 additions and 2 deletions
+9 -2
View File
@@ -11,8 +11,15 @@ from fastapi import HTTPException, Request, Response
logger = logging.getLogger(__name__) logger = logging.getLogger(__name__)
try:
_YamlLoader = yaml.CSafeLoader
_YamlDumperBase = yaml.CDumper
except AttributeError:
_YamlLoader = yaml.SafeLoader
_YamlDumperBase = yaml.Dumper
class _HelmDumper(yaml.Dumper):
class _HelmDumper(_YamlDumperBase):
"""YAML dumper that serializes datetime/date objects back to ISO 8601 strings. """YAML dumper that serializes datetime/date objects back to ISO 8601 strings.
yaml.safe_load converts timestamp-shaped YAML scalars (e.g. chart `created` yaml.safe_load converts timestamp-shaped YAML scalars (e.g. chart `created`
@@ -109,7 +116,7 @@ def _merge_helm_indexes(raw_indexes: list[bytes], member_names: list[str], membe
base_url = member_cfg.get("base_url", "").rstrip("/") base_url = member_cfg.get("base_url", "").rstrip("/")
try: try:
index = yaml.safe_load(raw_data) index = yaml.load(raw_data, Loader=_YamlLoader)
except Exception as e: except Exception as e:
logger.warning(f"Virtual: failed to parse index.yaml from member '{member_name}': {e}") logger.warning(f"Virtual: failed to parse index.yaml from member '{member_name}': {e}")
continue continue
+30
View File
@@ -14,6 +14,8 @@ from artifactapi.artifact.virtual import (
_merge_helm_indexes, _merge_helm_indexes,
_rewrite_urls, _rewrite_urls,
_VirtualHandler, _VirtualHandler,
_YamlDumperBase,
_YamlLoader,
) )
# --------------------------------------------------------------------------- # ---------------------------------------------------------------------------
@@ -82,6 +84,34 @@ _CFG_A = {"base_url": "https://helm.releases.hashicorp.com", "cache": {"mutable_
_CFG_B = {"base_url": "https://charts.example.com", "cache": {"mutable_ttl": 1800}} _CFG_B = {"base_url": "https://charts.example.com", "cache": {"mutable_ttl": 1800}}
# ---------------------------------------------------------------------------
# _YamlLoader / _YamlDumperBase — C extension selection
# ---------------------------------------------------------------------------
class TestYamlExtensionSelection:
def test_loader_is_a_class(self):
assert isinstance(_YamlLoader, type)
def test_dumper_base_is_a_class(self):
assert isinstance(_YamlDumperBase, type)
def test_helm_dumper_uses_selected_base(self):
assert issubclass(_HelmDumper, _YamlDumperBase)
def test_c_extensions_used_when_available(self):
try:
assert _YamlLoader is yaml.CSafeLoader
assert _YamlDumperBase is yaml.CDumper
except AttributeError:
assert _YamlLoader is yaml.SafeLoader
assert _YamlDumperBase is yaml.Dumper
def test_loader_can_parse_yaml(self):
result = yaml.load(b"key: value", Loader=_YamlLoader)
assert result == {"key": "value"}
# --------------------------------------------------------------------------- # ---------------------------------------------------------------------------
# _HelmDumper — datetime/date YAML serialization # _HelmDumper — datetime/date YAML serialization
# --------------------------------------------------------------------------- # ---------------------------------------------------------------------------