import logging
from collections.abc import Callable, Mapping
from pathlib import Path
from typing import Any, Unpack
import pandas as pd
from rics.misc import get_by_full_name, tname
from rics.types import AnyPath
from ..logging import generate_task_id
from ..offline.types import PlaceholderTranslations
from ..translator_typing import AbstractFetcherParams
from ..types import IdType
from ._abstract_fetcher import AbstractFetcher
from .types import FetchInstruction
PandasReadFunction = Callable[[AnyPath], pd.DataFrame]
FormatFn = Callable[[str], str]
[docs]
class PandasFetcher(AbstractFetcher[str, IdType]):
"""Fetcher implementation using :class:`pandas.DataFrame` as the data format.
Fetch data from serialized frames. How this is done is determined by the `read_function`. This is typically a Pandas
function such as :func:`pandas.read_csv` or :func:`pandas.read_parquet`, but any function that accepts a string or
path `source` as the first argument and returns a data frame can be used.
.. hint::
When using **remote file systems**, :attr:`~.AbstractFetcher.sources` are resolved using
`AbstractFileSystem.glob()`_. If resolution fails, consider overriding the :meth:`find_sources`-method.
Args:
read_function: A Pandas `read`-function. String are resolved by :func:`~rics.misc.get_by_full_name`, using
``default_module=pandas`` for convenience. Will attempt to derive from `read_path_format` if ``None``
(requires a string path).
read_path_format: A string on the form ``protocol://path/to/sources/{}.<ext>``, or a callable to apply
to a source before passing them to `read_function`.
read_function_kwargs: Additional keyword arguments for `read_function`.
**kwargs: See :class:`.AbstractFetcher`.
See Also:
The official `Pandas IO documentation <https://pandas.pydata.org/pandas-docs/stable/user_guide/io.html>`_
.. _AbstractFileSystem.glob(): https://filesystem-spec.readthedocs.io/en/latest/api.html?highlight=glob#fsspec.spec.AbstractFileSystem.glob
"""
def __init__(
self,
read_function: PandasReadFunction | str | None = None,
read_path_format: str | FormatFn = "data/{}.csv",
read_function_kwargs: Mapping[str, Any] | None = None,
**kwargs: Unpack[AbstractFetcherParams[str, IdType]],
) -> None:
super().__init__(**kwargs)
self._read = self._derive_read_function(read_function, read_path_format)
self._format_source: FormatFn = read_path_format if callable(read_path_format) else read_path_format.format
self._kwargs = read_function_kwargs or {}
self._source_paths: dict[str, str] = {}
[docs]
def read(self, source_path: AnyPath) -> pd.DataFrame:
"""Read a ``DataFrame`` from a source path.
Args:
source_path: Path to serialized ``DataFrame``.
Returns:
A deserialized ``DataFrame``.
"""
return self._read(source_path, **self._kwargs).convert_dtypes()
[docs]
def find_sources(self, task_id: int | None = None) -> dict[str, str]:
"""Resolve sources and their associated paths.
Args:
task_id: Used for logging.
Sources are resolved in three steps:
1. Create glob pattern by calling :meth:`format_source` with ``source='*'``.
2. Glob files using `AbstractFileSystem.glob()`_ (requires ``fsspec``) or :meth:`Path.glob() <pathlib.Path.glob>`.
3. Strip the directory and file suffix from the globbed paths to create source names.
.. _AbstractFileSystem.glob(): https://filesystem-spec.readthedocs.io/en/latest/api.html?highlight=glob#fsspec.spec.AbstractFileSystem.glob
Returns:
A dict ``{source: path}``.
"""
if task_id is None:
task_id = generate_task_id()
pattern = self.format_source("*")
try:
sources = self._find_sources_fsspec(pattern)
except ModuleNotFoundError as e:
self.logger.debug(f"Falling back to 'pathlib.Path': {e!r}")
sources = self._find_sources_pathlib(Path(pattern).expanduser().resolve())
source_paths: dict[str, str] = {source: self.format_source(source) for source in sources}
extra = {"task_id": task_id, "pattern": pattern, "source_paths": source_paths}
if source_paths:
self.logger.debug(f"Path {pattern=} matched {len(source_paths)} files: {source_paths}", extra=extra)
return source_paths
else:
self.logger.warning(f"Path {pattern=} did not match any files.", extra=extra)
return {}
def _find_sources_fsspec(self, pattern: str) -> list[str]:
from fsspec.core import url_to_fs # type: ignore # noqa: PLC0415
fs, pattern = url_to_fs(pattern, **self._kwargs.get("storage_options", {}))
return [self._path_to_source(path, pattern) for path in fs.glob(pattern)]
@classmethod
def _find_sources_pathlib(cls, pattern: Path) -> list[str]:
return [cls._path_to_source(str(path), str(pattern)) for path in pattern.parent.glob(pattern.name)]
@classmethod
def _path_to_source(cls, path: str, pattern: str) -> str:
glob_index = pattern.index("*")
prefix = pattern[:glob_index]
suffix = pattern[glob_index + 1 :]
return path.removeprefix(prefix).removesuffix(suffix)
def _initialize_sources(self, task_id: int) -> dict[str, list[str]]:
self._source_paths = self.find_sources(task_id)
return {source: self.read(path).columns.tolist() for source, path in self._source_paths.items()}
[docs]
def fetch_translations(self, instr: FetchInstruction[str, IdType]) -> PlaceholderTranslations[str]:
return PlaceholderTranslations.make(
instr.source,
self.read(self._source_paths[instr.source]),
)
def __repr__(self) -> str:
read_path_format = self.format_source("{}")
return f"{tname(self)}(read_function={tname(self._read)}, {read_path_format=})"
def _derive_read_function(
self,
read_function: PandasReadFunction | str | None,
read_path_format: str | FormatFn,
) -> PandasReadFunction:
if callable(read_function):
return read_function
if read_function is None:
if not isinstance(read_path_format, str):
msg = f"Cannot derive `read_function` from {read_path_format=} of type={type(read_function)}."
raise ValueError(msg)
func_to_suffixes: dict[PandasReadFunction, list[str]] = {
pd.read_csv: [".csv"],
pd.read_pickle: [".pickle", ".pkl"],
pd.read_feather: [".feather", ".ftr"],
pd.read_json: [".json"],
pd.read_parquet: [".parquet", ".parq"],
}
for func, suffixes in func_to_suffixes.items():
for suffix in suffixes:
if suffix in read_path_format: # Using endswith would break paths such as .csv.zip
if self.logger.isEnabledFor(logging.INFO):
read_function = f"pandas.{func.__name__}"
self.logger.debug(
f"Derived {read_function=} based on {suffix=} found in {read_path_format=}.",
extra={"suffix": suffix, "read_function": read_function},
)
return func
suffixes = [suffix for suffixes in func_to_suffixes.values() for suffix in suffixes]
msg = f"Cannot derive `read_function` from {read_path_format=}; does not contain any known {suffixes=}."
raise ValueError(msg)
func = get_by_full_name(read_function, pd)
if not callable(func):
msg = f"Bad {read_function=}; type {type(func).__name__} is not callable."
raise TypeError(msg)
return func # type: ignore[no-any-return]
