Source code for swh.indexer.metadata
# Copyright (C) 2017-2024 The Software Heritage developers
# See the AUTHORS file at the top-level directory of this distribution
# License: GNU General Public License version 3, or any later version
# See top-level LICENSE file for more information
from copy import deepcopy
import datetime
import hashlib
import logging
import re
import time
from typing import (
Any,
Callable,
Dict,
Iterable,
Iterator,
List,
Optional,
Tuple,
TypeVar,
cast,
)
import urllib.parse
from urllib.parse import urlparse
import pkg_resources
import sentry_sdk
from swh.core.config import merge_configs
from swh.core.utils import grouper
from swh.indexer.codemeta import merge_documents
from swh.indexer.indexer import (
BaseIndexer,
ContentIndexer,
DirectoryIndexer,
ObjectsDict,
OriginIndexer,
)
from swh.indexer.metadata_detector import detect_metadata
from swh.indexer.metadata_dictionary import EXTRINSIC_MAPPINGS, INTRINSIC_MAPPINGS
from swh.indexer.metadata_dictionary.base import DirectoryLsEntry
from swh.indexer.origin_head import get_head_swhid
from swh.indexer.storage import INDEXER_CFG_KEY
from swh.indexer.storage.model import (
ContentMetadataRow,
DirectoryIntrinsicMetadataRow,
OriginExtrinsicMetadataRow,
OriginIntrinsicMetadataRow,
)
from swh.model import hashutil
from swh.model.model import (
Directory,
MetadataAuthorityType,
Origin,
RawExtrinsicMetadata,
ReleaseTargetType,
Sha1Git,
)
from swh.model.swhids import CoreSWHID, ExtendedObjectType, ObjectType
from swh.objstorage.interface import CompositeObjId
REVISION_GET_BATCH_SIZE = 10
RELEASE_GET_BATCH_SIZE = 10
ORIGIN_GET_BATCH_SIZE = 10
T1 = TypeVar("T1")
T2 = TypeVar("T2")
logger = logging.getLogger(__name__)
[docs]
def call_with_batches(
f: Callable[[List[T1]], Iterable[T2]],
args: List[T1],
batch_size: int,
) -> Iterator[T2]:
"""Calls a function with batches of args, and concatenates the results."""
groups = grouper(args, batch_size)
for group in groups:
yield from f(list(group))
[docs]
class ExtrinsicMetadataIndexer(
BaseIndexer[Sha1Git, RawExtrinsicMetadata, OriginExtrinsicMetadataRow]
):
[docs]
def process_journal_objects(self, objects: ObjectsDict) -> Dict:
summary: Dict[str, Any] = {"status": "uneventful"}
try:
results = {}
for item in objects.get("raw_extrinsic_metadata", []):
remd = RawExtrinsicMetadata.from_dict(item)
sentry_sdk.set_tag("swh-indexer-remd-swhid", str(remd.swhid()))
for result in self.index(remd.id, data=remd):
results[result.id] = result
except Exception:
if not self.catch_exceptions:
raise
summary["status"] = "failed"
return summary
summary_persist = self.persist_index_computations(list(results.values()))
if summary_persist:
for value in summary_persist.values():
if value > 0:
summary["status"] = "eventful"
summary.update(summary_persist)
return summary
[docs]
def index(
self,
id: Sha1Git,
data: Optional[RawExtrinsicMetadata],
**kwargs,
) -> List[OriginExtrinsicMetadataRow]:
if data is None:
raise NotImplementedError(
"ExtrinsicMetadataIndexer.index() without RawExtrinsicMetadata data"
)
if data.target.object_type == ExtendedObjectType.ORIGIN:
origin_sha1 = data.target.object_id
elif data.origin is not None:
if (
data.fetcher.name == "swh-deposit"
and data.discovery_date
< datetime.datetime(
2024, 5, 16, 15, 20, 0, tzinfo=datetime.timezone.utc
)
):
# Workaround for deposits while swh.model.swhids did not unescape origin
# URLs while parsing <swh:object swhid="..." />, which was fixed in
# https://gitlab.softwareheritage.org/swh/devel/swh-model/-/merge_requests/348
# itself deployed shortly after 2024-05-16T15:18:07 by
# https://gitlab.softwareheritage.org/swh/infra/swh-apps/-/commit/70bd86aafcbc1787183e5d2cd52c392ae012e65e
if "%" in data.origin:
assert re.match(
"^https://cran.r-project.org/package%3D[a-zA-Z]+$", data.origin
), data
origin_url = urllib.parse.unquote_to_bytes(data.origin)
else:
origin_url = data.origin.encode()
# HACK: As swh-search does not (yet?) support searching on directories
# and traversing back to origins, we index metadata on non-origins with
# an origin context as if they were on the origin itself.
origin_sha1 = hashlib.sha1(origin_url).digest()
else:
# other types are not supported yet
return []
if data.authority.type == MetadataAuthorityType.REGISTRY:
# metadata provided by a third-party; don't trust it
# (technically this could be handled below, but we check it here
# to return early; sparing a translation and origin lookup)
# TODO: add ways to define trusted authorities
return []
metadata_items = []
mappings: List[str] = []
for mapping_cls in EXTRINSIC_MAPPINGS.values():
if data.format in mapping_cls.extrinsic_metadata_formats():
mapping = mapping_cls()
metadata_item = mapping.translate(data.metadata)
if metadata_item is not None:
metadata_items.append(metadata_item)
mappings.append(mapping.name)
if not metadata_items:
# Don't have any mapping to parse it, ignore
return []
# TODO: batch requests to origin_get_by_sha1()
num_retries = 6
sleep_delay = 10
for _ in range(num_retries):
origins = self.storage.origin_get_by_sha1([origin_sha1])
try:
(origin,) = origins
if origin is not None:
break
except ValueError:
pass
# The origin does not exist. This may be due to some replication lag
# between the loader's DB/journal and the DB we are consuming from.
# Wait a bit and try again
logger.debug(
"Origin %s not found, sleeping for %ss.", data.target, sleep_delay
)
time.sleep(sleep_delay)
else:
# Does not exist, or replication lag > 60s.
raise ValueError(
f"Unknown origin swh:1:ori:{origin_sha1.hex()} for metadata target: "
f"{data.target}. Is the swh-storage database replication lag "
f"over {num_retries*sleep_delay}s?"
) from None
if urlparse(data.authority.url).netloc != urlparse(origin["url"]).netloc:
# metadata provided by a third-party; don't trust it
# TODO: add ways to define trusted authorities
return []
metadata = merge_documents(metadata_items)
return [
OriginExtrinsicMetadataRow(
id=origin["url"],
indexer_configuration_id=self.tool["id"],
from_remd_id=data.id,
mappings=mappings,
metadata=metadata,
)
]
[docs]
def persist_index_computations(
self, results: List[OriginExtrinsicMetadataRow]
) -> Dict[str, int]:
"""Persist the results in storage."""
return self.idx_storage.origin_extrinsic_metadata_add(results)
[docs]
class ContentMetadataIndexer(ContentIndexer[ContentMetadataRow]):
"""Content-level indexer
This indexer is in charge of:
- filtering out content already indexed in content_metadata
- reading content from objstorage with the content's id sha1
- computing metadata by given context
- using the metadata_dictionary as the 'swh-metadata-translator' tool
- store result in content_metadata table
"""
[docs]
def filter(self, ids: List[CompositeObjId]):
"""Filter out known sha1s and return only missing ones."""
yield from self.idx_storage.content_metadata_missing(
(
{
"id": id["sha1"],
"indexer_configuration_id": self.tool["id"],
}
for id in ids
)
)
[docs]
def index(
self,
id: CompositeObjId,
data: Optional[bytes] = None,
log_suffix="unknown directory",
**kwargs,
) -> List[ContentMetadataRow]:
"""Index sha1s' content and store result.
Args:
id: content's identifier
data: raw content in bytes
Returns:
dict: dictionary representing a content_metadata. If the
translation wasn't successful the metadata keys will
be returned as None
"""
assert "sha1" in id
assert data is not None
metadata = None
try:
mapping_name = self.tool["tool_configuration"]["context"]
log_suffix += ", content_id=%s" % hashutil.hash_to_hex(id["sha1"])
metadata = INTRINSIC_MAPPINGS[mapping_name](log_suffix).translate(data)
except Exception:
self.log.exception(
"Problem during metadata translation "
"for content %s" % hashutil.hash_to_hex(id["sha1"])
)
sentry_sdk.capture_exception()
if metadata is None:
return []
return [
ContentMetadataRow(
id=id["sha1"],
indexer_configuration_id=self.tool["id"],
metadata=metadata,
)
]
[docs]
def persist_index_computations(
self, results: List[ContentMetadataRow]
) -> Dict[str, int]:
"""Persist the results in storage."""
return self.idx_storage.content_metadata_add(results)
DEFAULT_CONFIG: Dict[str, Any] = {
"tools": {
"name": "swh.indexer.metadata",
"version": pkg_resources.get_distribution("swh.indexer").version,
"configuration": {},
},
}
[docs]
class DirectoryMetadataIndexer(DirectoryIndexer[DirectoryIntrinsicMetadataRow]):
"""Directory-level indexer
This indexer is in charge of:
- filtering directories already indexed in directory_intrinsic_metadata table
with defined computation tool
- retrieve all entry_files in directory
- use metadata_detector for file_names containing metadata
- compute metadata translation if necessary and possible (depends on tool)
- send sha1s to content indexing if possible
- store the results for directory
"""
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
self.config = merge_configs(DEFAULT_CONFIG, self.config)
[docs]
def filter(self, sha1_gits):
"""Filter out known sha1s and return only missing ones."""
yield from self.idx_storage.directory_intrinsic_metadata_missing(
(
{
"id": sha1_git,
"indexer_configuration_id": self.tool["id"],
}
for sha1_git in sha1_gits
)
)
[docs]
def index(
self, id: Sha1Git, data: Optional[Directory] = None, **kwargs
) -> List[DirectoryIntrinsicMetadataRow]:
"""Index directory by processing it and organizing result.
use metadata_detector to iterate on filenames, passes them to the content
indexers, then merges (if more than one)
Args:
id: sha1_git of the directory
data: should always be None
Returns:
dict: dictionary representing a directory_intrinsic_metadata, with
keys:
- id: directory's identifier (sha1_git)
- indexer_configuration_id (bytes): tool used
- metadata: dict of retrieved metadata
"""
dir_: List[DirectoryLsEntry]
assert data is None, "Unexpected directory object"
dir_ = cast(
List[DirectoryLsEntry],
list(self.storage.directory_ls(id, recursive=False)),
)
try:
if [entry["type"] for entry in dir_] == ["dir"]:
# If the root is just a single directory, recurse into it
# eg. PyPI packages, GNU tarballs
subdir = dir_[0]["target"]
dir_ = cast(
List[DirectoryLsEntry],
list(self.storage.directory_ls(subdir, recursive=False)),
)
files = [entry for entry in dir_ if entry["type"] == "file"]
(mappings, metadata) = self.translate_directory_intrinsic_metadata(
files,
log_suffix="directory=%s" % hashutil.hash_to_hex(id),
)
except Exception as e:
self.log.exception("Problem when indexing dir: %r", e)
sentry_sdk.capture_exception()
return []
return [
DirectoryIntrinsicMetadataRow(
id=id,
indexer_configuration_id=self.tool["id"],
mappings=mappings,
metadata=metadata,
)
]
[docs]
def persist_index_computations(
self, results: List[DirectoryIntrinsicMetadataRow]
) -> Dict[str, int]:
"""Persist the results in storage."""
# TODO: add functions in storage to keep data in
# directory_intrinsic_metadata
return self.idx_storage.directory_intrinsic_metadata_add(results)
[docs]
def translate_directory_intrinsic_metadata(
self, files: List[DirectoryLsEntry], log_suffix: str
) -> Tuple[List[Any], Any]:
"""
Determine plan of action to translate metadata in the given root directory
Args:
files: list of file entries, as returned by
:meth:`swh.storage.interface.StorageInterface.directory_ls`
Returns:
(List[str], dict): list of mappings used and dict with
translated metadata according to the CodeMeta vocabulary
"""
metadata = []
# TODO: iterate on each context, on each file
# -> get raw_contents
# -> translate each content
config = {
k: self.config[k]
for k in [INDEXER_CFG_KEY, "objstorage", "storage", "tools"]
}
all_detected_files = detect_metadata(files)
used_mappings = [
INTRINSIC_MAPPINGS[context].name for context in all_detected_files
]
for mapping_name, detected_files in all_detected_files.items():
cfg = deepcopy(config)
cfg["tools"]["configuration"]["context"] = mapping_name
c_metadata_indexer = ContentMetadataIndexer(config=cfg)
# sha1s that are in content_metadata table
sha1s_in_storage = []
metadata_generator = self.idx_storage.content_metadata_get(
[f["sha1"] for f in detected_files]
)
for c in metadata_generator:
# extracting metadata
sha1 = c.id
sha1s_in_storage.append(sha1)
local_metadata = c.metadata
# local metadata is aggregated
if local_metadata:
metadata.append(local_metadata)
sha1s_filtered = [
item for item in detected_files if item["sha1"] not in sha1s_in_storage
]
if sha1s_filtered:
# content indexing
try:
_, results = c_metadata_indexer.run(
sha1s_filtered,
log_suffix=log_suffix,
)
# on the fly possibility:
for result in results:
local_metadata = result.metadata
metadata.append(local_metadata)
except Exception:
self.log.exception("Exception while indexing metadata on contents")
sentry_sdk.capture_exception()
metadata = merge_documents(metadata)
return (used_mappings, metadata)
[docs]
class OriginMetadataIndexer(
OriginIndexer[Tuple[OriginIntrinsicMetadataRow, DirectoryIntrinsicMetadataRow]]
):
USE_TOOLS = False
def __init__(self, config=None, **kwargs) -> None:
super().__init__(config=config, **kwargs)
self.directory_metadata_indexer = DirectoryMetadataIndexer(config=config)
[docs]
def index_list(
self,
origins: List[Origin],
*,
check_origin_known: bool = True,
**kwargs,
) -> List[Tuple[OriginIntrinsicMetadataRow, DirectoryIntrinsicMetadataRow]]:
head_rev_ids = []
head_rel_ids = []
origin_heads: Dict[Origin, CoreSWHID] = {}
# Filter out origins not in the storage
if check_origin_known:
known_origins = list(
call_with_batches(
self.storage.origin_get,
[origin.url for origin in origins],
ORIGIN_GET_BATCH_SIZE,
)
)
else:
known_origins = list(origins)
for origin in known_origins:
if origin is None:
continue
head_swhid = get_head_swhid(self.storage, origin.url)
if head_swhid:
origin_heads[origin] = head_swhid
if head_swhid.object_type == ObjectType.REVISION:
head_rev_ids.append(head_swhid.object_id)
elif head_swhid.object_type == ObjectType.RELEASE:
head_rel_ids.append(head_swhid.object_id)
else:
assert False, head_swhid
head_revs = dict(
zip(
head_rev_ids,
call_with_batches(
self.storage.revision_get, head_rev_ids, REVISION_GET_BATCH_SIZE
),
)
)
head_rels = dict(
zip(
head_rel_ids,
call_with_batches(
self.storage.release_get, head_rel_ids, RELEASE_GET_BATCH_SIZE
),
)
)
results = []
for origin, head_swhid in origin_heads.items():
sentry_sdk.set_tag("swh-indexer-origin-url", origin.url)
sentry_sdk.set_tag("swh-indexer-origin-head-swhid", str(head_swhid))
if head_swhid.object_type == ObjectType.REVISION:
rev = head_revs[head_swhid.object_id]
if not rev:
self.log.warning(
"Missing head object %s of origin %r", head_swhid, origin.url
)
continue
directory_id = rev.directory
elif head_swhid.object_type == ObjectType.RELEASE:
rel = head_rels[head_swhid.object_id]
if not rel:
self.log.warning(
"Missing head object %s of origin %r", head_swhid, origin.url
)
continue
if rel.target_type != ReleaseTargetType.DIRECTORY:
# TODO
self.log.warning(
"Head release %s of %r has unexpected target type %s",
head_swhid,
origin.url,
rel.target_type,
)
continue
assert rel.target, rel
directory_id = rel.target
else:
assert False, head_swhid
for dir_metadata in self.directory_metadata_indexer.index(directory_id):
# There is at most one dir_metadata
orig_metadata = OriginIntrinsicMetadataRow(
from_directory=dir_metadata.id,
id=origin.url,
metadata=dir_metadata.metadata,
mappings=dir_metadata.mappings,
indexer_configuration_id=dir_metadata.indexer_configuration_id,
)
results.append((orig_metadata, dir_metadata))
return results
[docs]
def persist_index_computations(
self,
results: List[Tuple[OriginIntrinsicMetadataRow, DirectoryIntrinsicMetadataRow]],
) -> Dict[str, int]:
# Deduplicate directories
dir_metadata: Dict[bytes, DirectoryIntrinsicMetadataRow] = {}
orig_metadata: Dict[str, OriginIntrinsicMetadataRow] = {}
summary: Dict = {}
for orig_item, dir_item in results:
assert dir_item.metadata == orig_item.metadata
if dir_item.metadata and not (dir_item.metadata.keys() <= {"@context"}):
# Only store non-empty metadata sets
if dir_item.id not in dir_metadata:
dir_metadata[dir_item.id] = dir_item
if orig_item.id not in orig_metadata:
orig_metadata[orig_item.id] = orig_item
if dir_metadata:
summary_dir = self.idx_storage.directory_intrinsic_metadata_add(
list(dir_metadata.values())
)
summary.update(summary_dir)
if orig_metadata:
summary_ori = self.idx_storage.origin_intrinsic_metadata_add(
list(orig_metadata.values())
)
summary.update(summary_ori)
return summary