mesa/bin/ci/structured_logger.py
Guilherme Gallo eb1945c3d9 bin/ci: Add StructuredLogger to improve log handling
This commit introduces the StructuredLogger module, which provides a
robust and flexible logging utility supporting multiple data formats
(CSV, JSON, and YAML). By incorporating this module into our CI system,
we enhance our log management capabilities, making it easier to:

1. Monitor and analyze logs: The StructuredLogger is a dict-like data
   abstraction which autosaves into a structured data file, whenever it
   is updated. With this file, one can easily know specifics of the job
   execution without having to grep it in the traces logs or exploring
   the job artifacts. The autosave feature makes it useful even when the
   CI job fails unexpectedly, since the partial dict is always written
   back to the disk.
2. Maintain data integrity: The module includes context managers for
   file locking and editing log data, ensuring data integrity and
   preventing race conditions.
3. Support multiple formats: With built-in support for CSV, JSON, and
   YAML formats, this module caters to a wide range of use cases and
   user preferences.
4. Increase maintainability: The modular design of the StructuredLogger
   and its corresponding strategies simplifies maintenance and allows
   for seamless integration of additional formats in the future.

Signed-off-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22500>
2023-04-19 14:36:37 +00:00

295 lines
8.7 KiB
Python

"""
A structured logging utility supporting multiple data formats such as CSV, JSON,
and YAML.
The main purpose of this script, besides having relevant information available
in a condensed and deserialized.
This script defines a protocol for different file handling strategies and provides
implementations for CSV, JSON, and YAML formats. The main class, StructuredLogger,
allows for easy interaction with log data, enabling users to load, save, increment,
set, and append fields in the log. The script also includes context managers for
file locking and editing log data to ensure data integrity and avoid race conditions.
"""
import json
import os
from collections.abc import MutableMapping, MutableSequence
from contextlib import contextmanager
from datetime import datetime
from pathlib import Path
from typing import Any, Protocol
import fire
from filelock import FileLock
try:
import polars as pl
CSV_LIB_EXCEPTION = None
except ImportError as e:
CSV_LIB_EXCEPTION: ImportError = e
try:
from ruamel.yaml import YAML
YAML_LIB_EXCEPTION = None
except ImportError as e:
YAML_LIB_EXCEPTION: ImportError = e
class ContainerProxy:
"""
A proxy class that wraps a mutable container object (such as a dictionary or
a list) and calls a provided save_callback function whenever the container
or its contents
are changed.
"""
def __init__(self, container, save_callback):
self.container = container
self.save_callback = save_callback
def __getitem__(self, key):
value = self.container[key]
if isinstance(value, (MutableMapping, MutableSequence)):
return ContainerProxy(value, self.save_callback)
return value
def __setitem__(self, key, value):
self.container[key] = value
self.save_callback()
def __delitem__(self, key):
del self.container[key]
self.save_callback()
def __getattr__(self, name):
attr = getattr(self.container, name)
if callable(attr):
def wrapper(*args, **kwargs):
result = attr(*args, **kwargs)
self.save_callback()
return result
return wrapper
return attr
def __iter__(self):
return iter(self.container)
def __len__(self):
return len(self.container)
def __repr__(self):
return repr(self.container)
class AutoSaveDict(dict):
"""
A subclass of the built-in dict class with additional functionality to
automatically save changes to the dictionary. It maintains a timestamp of
the last modification and automatically wraps nested mutable containers
using ContainerProxy.
"""
timestamp_key = "_timestamp"
def __init__(self, *args, save_callback, register_timestamp=True, **kwargs):
self.save_callback = save_callback
self.__register_timestamp = register_timestamp
self.__heartbeat()
super().__init__(*args, **kwargs)
self.__wrap_dictionaries()
def __heartbeat(self):
if self.__register_timestamp:
self[AutoSaveDict.timestamp_key] = datetime.now().isoformat()
def __save(self):
self.__heartbeat()
self.save_callback()
def __wrap_dictionaries(self):
for key, value in self.items():
if isinstance(value, MutableMapping) and not isinstance(
value, AutoSaveDict
):
self[key] = AutoSaveDict(
value, save_callback=self.save_callback, register_timestamp=False
)
def __setitem__(self, key, value):
if isinstance(value, MutableMapping) and not isinstance(value, AutoSaveDict):
value = AutoSaveDict(
value, save_callback=self.save_callback, register_timestamp=False
)
super().__setitem__(key, value)
if self.__register_timestamp and key == AutoSaveDict.timestamp_key:
return
self.__save()
def __getitem__(self, key):
value = super().__getitem__(key)
if isinstance(value, (MutableMapping, MutableSequence)):
return ContainerProxy(value, self.__save)
return value
def __delitem__(self, key):
super().__delitem__(key)
self.__save()
def pop(self, *args, **kwargs):
result = super().pop(*args, **kwargs)
self.__save()
return result
def update(self, *args, **kwargs):
super().update(*args, **kwargs)
self.__wrap_dictionaries()
self.__save()
class StructuredLoggerStrategy(Protocol):
def load_data(self, file_path: Path) -> dict:
pass
def save_data(self, file_path: Path, data: dict) -> None:
pass
class CSVStrategy:
def __init__(self) -> None:
if CSV_LIB_EXCEPTION:
raise RuntimeError(
"Can't parse CSV files. Missing library"
) from CSV_LIB_EXCEPTION
def load_data(self, file_path: Path) -> dict:
dicts: list[dict[str, Any]] = pl.read_csv(
file_path, try_parse_dates=True
).to_dicts()
data = {}
for d in dicts:
for k, v in d.items():
if k != AutoSaveDict.timestamp_key and k in data:
if isinstance(data[k], list):
data[k].append(v)
continue
data[k] = [data[k], v]
else:
data[k] = v
return data
def save_data(self, file_path: Path, data: dict) -> None:
pl.DataFrame(data).write_csv(file_path)
class JSONStrategy:
def load_data(self, file_path: Path) -> dict:
return json.loads(file_path.read_text())
def save_data(self, file_path: Path, data: dict) -> None:
with open(file_path, "w") as f:
json.dump(data, f, indent=2)
class YAMLStrategy:
def __init__(self):
if YAML_LIB_EXCEPTION:
raise RuntimeError(
"Can't parse YAML files. Missing library"
) from YAML_LIB_EXCEPTION
self.yaml = YAML()
self.yaml.indent(sequence=4, offset=2)
self.yaml.default_flow_style = False
self.yaml.representer.add_representer(AutoSaveDict, self.represent_dict)
@classmethod
def represent_dict(cls, dumper, data):
return dumper.represent_mapping("tag:yaml.org,2002:map", data)
def load_data(self, file_path: Path) -> dict:
return self.yaml.load(file_path.read_text())
def save_data(self, file_path: Path, data: dict) -> None:
with open(file_path, "w") as f:
self.yaml.dump(data, f)
class StructuredLogger:
def __init__(
self, file_name: str, strategy: StructuredLoggerStrategy = None, truncate=False
):
self.file_name: str = file_name
self.file_path = Path(self.file_name)
self._data: AutoSaveDict = AutoSaveDict(save_callback=self.save_data)
if strategy is None:
self.strategy: StructuredLoggerStrategy = self.guess_strategy_from_file(
self.file_path
)
else:
self.strategy = strategy
if not self.file_path.exists():
Path.mkdir(self.file_path.parent, exist_ok=True)
self.save_data()
return
if truncate:
with self.get_lock():
os.truncate(self.file_path, 0)
self.save_data()
def load_data(self):
self._data = self.strategy.load_data(self.file_path)
def save_data(self):
self.strategy.save_data(self.file_path, self._data)
@property
def data(self) -> AutoSaveDict:
return self._data
@contextmanager
def get_lock(self):
with FileLock(f"{self.file_path}.lock", timeout=10):
yield
@contextmanager
def edit_context(self):
"""
Context manager that ensures proper loading and saving of log data when
performing multiple modifications.
"""
with self.get_lock():
try:
self.load_data()
yield
finally:
self.save_data()
@staticmethod
def guess_strategy_from_file(file_path: Path) -> StructuredLoggerStrategy:
file_extension = file_path.suffix.lower().lstrip(".")
return StructuredLogger.get_strategy(file_extension)
@staticmethod
def get_strategy(strategy_name: str) -> StructuredLoggerStrategy:
strategies = {
"csv": CSVStrategy,
"json": JSONStrategy,
"yaml": YAMLStrategy,
"yml": YAMLStrategy,
}
try:
return strategies[strategy_name]()
except KeyError as e:
raise ValueError(f"Unknown strategy for: {strategy_name}") from e
if __name__ == "__main__":
fire.Fire(StructuredLogger)