fsspec 2024.9.0__tar.gz → 2024.10.0__tar.gz

{fsspec-2024.9.0 → fsspec-2024.10.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: fsspec
-Version: 2024.9.0
+Version: 2024.10.0
 Summary: File-system specification
 Project-URL: Changelog, https://filesystem-spec.readthedocs.io/en/latest/changelog.html
 Project-URL: Documentation, https://filesystem-spec.readthedocs.io/en/latest/

{fsspec-2024.9.0 → fsspec-2024.10.0}/docs/source/changelog.rst

@@ -1,6 +1,27 @@
 Changelog
 =========
 
+2024.10.0
+---------
+
+Fixes
+
+- Performance of memoryFS rm (#1725)
+- Performance of git FS info (#1712)
+- Avoid git hex for newer pygit (#1703)
+- tests fix for zip (#1700, 1691)
+- missing open_async for dirFS (#1698)
+- handle pathlib in zip (#1689)
+- skip tests needing kerchunk if not installed (#1689)
+- allow repeated kwargs in unchain (#1673)
+
+Other
+
+- Code style (#1704, 1706)
+- allow pyarrow in referenceFS parquet (#1692)
+- don't hardcode test port for parallel runs (#1690)
+
+
 2024.9.0
 --------
 

{fsspec-2024.9.0 → fsspec-2024.10.0}/fsspec/_version.py

@@ -12,5 +12,5 @@ __version__: str
 __version_tuple__: VERSION_TUPLE
 version_tuple: VERSION_TUPLE
 
-__version__ = version = '2024.9.0'
-__version_tuple__ = version_tuple = (2024, 9, 0)
+__version__ = version = '2024.10.0'
+__version_tuple__ = version_tuple = (2024, 10, 0)

{fsspec-2024.9.0 → fsspec-2024.10.0}/fsspec/asyn.py

@@ -344,6 +344,10 @@ class AsyncFileSystem(AbstractFileSystem):
     async def _cp_file(self, path1, path2, **kwargs):
         raise NotImplementedError
 
+    async def _mv_file(self, path1, path2):
+        await self._cp_file(path1, path2)
+        await self._rm_file(path1)
+
     async def _copy(
         self,
         path1,

{fsspec-2024.9.0 → fsspec-2024.10.0}/fsspec/core.py

@@ -346,7 +346,10 @@ def _un_chain(path, kwargs):
         kws = kwargs.pop(protocol, {})
         if bit is bits[0]:
             kws.update(kwargs)
-        kw = dict(**extra_kwargs, **kws)
+        kw = dict(
+            **{k: v for k, v in extra_kwargs.items() if k not in kws or v != kws[k]},
+            **kws,
+        )
         bit = cls._strip_protocol(bit)
         if (
             protocol in {"blockcache", "filecache", "simplecache"}
@@ -578,7 +581,7 @@ def expand_paths_if_needed(paths, mode, num, fs, name_function):
     paths = list(paths)
 
     if "w" in mode:  # read mode
-        if sum([1 for p in paths if "*" in p]) > 1:
+        if sum(1 for p in paths if "*" in p) > 1:
             raise ValueError(
                 "When writing data, only one filename mask can be specified."
             )

{fsspec-2024.9.0 → fsspec-2024.10.0}/fsspec/implementations/dirfs.py

@@ -370,3 +370,15 @@ class DirFileSystem(AsyncFileSystem):
             *args,
             **kwargs,
         )
+
+    async def open_async(
+        self,
+        path,
+        *args,
+        **kwargs,
+    ):
+        return await self.fs.open_async(
+            self._join(path),
+            *args,
+            **kwargs,
+        )

{fsspec-2024.9.0 → fsspec-2024.10.0}/fsspec/implementations/git.py

@@ -55,6 +55,8 @@ class GitFileSystem(AbstractFileSystem):
         tree = comm.tree
         for part in parts:
             if part and isinstance(tree, pygit2.Tree):
+                if part not in tree:
+                    raise FileNotFoundError(path)
                 tree = tree[part]
         return tree
 
@@ -69,46 +71,32 @@ class GitFileSystem(AbstractFileSystem):
             out["ref"], path = path.split("@", 1)
         return out
 
+    @staticmethod
+    def _object_to_info(obj, path=None):
+        # obj.name and obj.filemode are None for the root tree!
+        is_dir = isinstance(obj, pygit2.Tree)
+        return {
+            "type": "directory" if is_dir else "file",
+            "name": (
+                "/".join([path, obj.name or ""]).lstrip("/") if path else obj.name
+            ),
+            "hex": str(obj.id),
+            "mode": "100644" if obj.filemode is None else f"{obj.filemode:o}",
+            "size": 0 if is_dir else obj.size,
+        }
+
     def ls(self, path, detail=True, ref=None, **kwargs):
-        path = self._strip_protocol(path)
-        tree = self._path_to_object(path, ref)
-        if isinstance(tree, pygit2.Tree):
-            out = []
-            for obj in tree:
-                if isinstance(obj, pygit2.Tree):
-                    out.append(
-                        {
-                            "type": "directory",
-                            "name": "/".join([path, obj.name]).lstrip("/"),
-                            "hex": obj.hex,
-                            "mode": f"{obj.filemode:o}",
-                            "size": 0,
-                        }
-                    )
-                else:
-                    out.append(
-                        {
-                            "type": "file",
-                            "name": "/".join([path, obj.name]).lstrip("/"),
-                            "hex": obj.hex,
-                            "mode": f"{obj.filemode:o}",
-                            "size": obj.size,
-                        }
-                    )
-        else:
-            obj = tree
-            out = [
-                {
-                    "type": "file",
-                    "name": obj.name,
-                    "hex": obj.hex,
-                    "mode": f"{obj.filemode:o}",
-                    "size": obj.size,
-                }
-            ]
-        if detail:
-            return out
-        return [o["name"] for o in out]
+        tree = self._path_to_object(self._strip_protocol(path), ref)
+        return [
+            GitFileSystem._object_to_info(obj, path)
+            if detail
+            else GitFileSystem._object_to_info(obj, path)["name"]
+            for obj in (tree if isinstance(tree, pygit2.Tree) else [tree])
+        ]
+
+    def info(self, path, ref=None, **kwargs):
+        tree = self._path_to_object(self._strip_protocol(path), ref)
+        return GitFileSystem._object_to_info(tree, path)
 
     def ukey(self, path, ref=None):
         return self.info(path, ref=ref)["hex"]

{fsspec-2024.9.0 → fsspec-2024.10.0}/fsspec/implementations/http.py

@@ -358,9 +358,10 @@ class HTTPFileSystem(AsyncFileSystem):
         kw = self.kwargs.copy()
         kw["asynchronous"] = self.asynchronous
         kw.update(kwargs)
-        size = size or self.info(path, **kwargs)["size"]
+        info = {}
+        size = size or info.update(self.info(path, **kwargs)) or info["size"]
         session = sync(self.loop, self.set_session)
-        if block_size and size:
+        if block_size and size and info.get("partial", True):
             return HTTPFile(
                 self,
                 path,
@@ -520,9 +521,9 @@ class HTTPFileSystem(AsyncFileSystem):
 
 class HTTPFile(AbstractBufferedFile):
     """
-    A file-like object pointing to a remove HTTP(S) resource
+    A file-like object pointing to a remote HTTP(S) resource
 
-    Supports only reading, with read-ahead of a predermined block-size.
+    Supports only reading, with read-ahead of a predetermined block-size.
 
     In the case that the server does not supply the filesize, only reading of
     the complete file in one go is supported.
@@ -835,10 +836,6 @@ async def _file_info(url, session, size_policy="head", **kwargs):
     async with r:
         r.raise_for_status()
 
-        # TODO:
-        #  recognise lack of 'Accept-Ranges',
-        #                 or 'Accept-Ranges': 'none' (not 'bytes')
-        #  to mean streaming only, no random access => return None
         if "Content-Length" in r.headers:
             # Some servers may choose to ignore Accept-Encoding and return
             # compressed content, in which case the returned size is unreliable.
@@ -853,6 +850,11 @@ async def _file_info(url, session, size_policy="head", **kwargs):
         if "Content-Type" in r.headers:
             info["mimetype"] = r.headers["Content-Type"].partition(";")[0]
 
+        if r.headers.get("Accept-Ranges") == "none":
+            # Some servers may explicitly discourage partial content requests, but
+            # the lack of "Accept-Ranges" does not always indicate they would fail
+            info["partial"] = False
+
         info["url"] = str(r.url)
 
         for checksum_field in ["ETag", "Content-MD5", "Digest"]:

{fsspec-2024.9.0 → fsspec-2024.10.0}/fsspec/implementations/memory.py

@@ -248,6 +248,10 @@ class MemoryFileSystem(AbstractFileSystem):
         except KeyError as e:
             raise FileNotFoundError(path) from e
 
+    def isfile(self, path):
+        path = self._strip_protocol(path)
+        return path in self.store
+
     def rm(self, path, recursive=False, maxdepth=None):
         if isinstance(path, str):
             path = self._strip_protocol(path)
@@ -255,14 +259,14 @@ class MemoryFileSystem(AbstractFileSystem):
             path = [self._strip_protocol(p) for p in path]
         paths = self.expand_path(path, recursive=recursive, maxdepth=maxdepth)
         for p in reversed(paths):
+            if self.isfile(p):
+                self.rm_file(p)
             # If the expanded path doesn't exist, it is only because the expanded
             # path was a directory that does not exist in self.pseudo_dirs. This
             # is possible if you directly create files without making the
             # directories first.
-            if not self.exists(p):
+            elif not self.exists(p):
                 continue
-            if self.isfile(p):
-                self.rm_file(p)
             else:
                 self.rmdir(p)
 

{fsspec-2024.9.0 → fsspec-2024.10.0}/fsspec/implementations/reference.py

@@ -7,7 +7,7 @@ import math
 import os
 from itertools import chain
 from functools import lru_cache
-from typing import TYPE_CHECKING
+from typing import TYPE_CHECKING, Literal
 
 import fsspec.core
 
@@ -104,7 +104,13 @@ class LazyReferenceMapper(collections.abc.MutableMapping):
         return pd
 
     def __init__(
-        self, root, fs=None, out_root=None, cache_size=128, categorical_threshold=10
+        self,
+        root,
+        fs=None,
+        out_root=None,
+        cache_size=128,
+        categorical_threshold=10,
+        engine: Literal["fastparquet", "pyarrow"] = "fastparquet",
     ):
         """
 
@@ -126,16 +132,25 @@ class LazyReferenceMapper(collections.abc.MutableMapping):
             Encode urls as pandas.Categorical to reduce memory footprint if the ratio
             of the number of unique urls to total number of refs for each variable
             is greater than or equal to this number. (default 10)
+        engine: Literal["fastparquet","pyarrow"]
+            Engine choice for reading parquet files. (default is "fastparquet")
         """
+
         self.root = root
         self.chunk_sizes = {}
         self.out_root = out_root or self.root
         self.cat_thresh = categorical_threshold
+        self.engine = engine
         self.cache_size = cache_size
         self.url = self.root + "/{field}/refs.{record}.parq"
         # TODO: derive fs from `root`
         self.fs = fsspec.filesystem("file") if fs is None else fs
 
+        from importlib.util import find_spec
+
+        if self.engine == "pyarrow" and find_spec("pyarrow") is None:
+            raise ImportError("engine choice `pyarrow` is not installed.")
+
     def __getattr__(self, item):
         if item in ("_items", "record_size", "zmetadata"):
             self.setup()
@@ -158,7 +173,7 @@ class LazyReferenceMapper(collections.abc.MutableMapping):
             """cached parquet file loader"""
             path = self.url.format(field=field, record=record)
             data = io.BytesIO(self.fs.cat_file(path))
-            df = self.pd.read_parquet(data, engine="fastparquet")
+            df = self.pd.read_parquet(data, engine=self.engine)
             refs = {c: df[c].to_numpy() for c in df.columns}
             return refs
 
@@ -463,18 +478,28 @@ class LazyReferenceMapper(collections.abc.MutableMapping):
 
         fn = f"{base_url or self.out_root}/{field}/refs.{record}.parq"
         self.fs.mkdirs(f"{base_url or self.out_root}/{field}", exist_ok=True)
+
+        if self.engine == "pyarrow":
+            df_backend_kwargs = {"write_statistics": False}
+        elif self.engine == "fastparquet":
+            df_backend_kwargs = {
+                "stats": False,
+                "object_encoding": object_encoding,
+                "has_nulls": has_nulls,
+            }
+        else:
+            raise NotImplementedError(f"{self.engine} not supported")
+
         df.to_parquet(
             fn,
-            engine="fastparquet",
+            engine=self.engine,
             storage_options=storage_options
             or getattr(self.fs, "storage_options", None),
             compression="zstd",
             index=False,
-            stats=False,
-            object_encoding=object_encoding,
-            has_nulls=has_nulls,
-            # **kwargs,
+            **df_backend_kwargs,
         )
+
         partition.clear()
         self._items.pop((field, record))
 
@@ -486,6 +511,7 @@ class LazyReferenceMapper(collections.abc.MutableMapping):
         base_url: str
             Location of the output
         """
+
         # write what we have so far and clear sub chunks
         for thing in list(self._items):
             if isinstance(thing, tuple):

{fsspec-2024.9.0 → fsspec-2024.10.0}/fsspec/implementations/zip.py

@@ -1,3 +1,4 @@
+import os
 import zipfile
 
 import fsspec
@@ -48,7 +49,7 @@ class ZipFileSystem(AbstractArchiveFileSystem):
         if mode not in set("rwa"):
             raise ValueError(f"mode '{mode}' no understood")
         self.mode = mode
-        if isinstance(fo, str):
+        if isinstance(fo, (str, os.PathLike)):
             if mode == "a":
                 m = "r+b"
             else:

{fsspec-2024.9.0 → fsspec-2024.10.0}/fsspec/spec.py

@@ -428,11 +428,9 @@ class AbstractFileSystem(metaclass=_Cached):
         except (FileNotFoundError, OSError) as e:
             if on_error == "raise":
                 raise
-            elif callable(on_error):
+            if callable(on_error):
                 on_error(e)
-            if detail:
-                return path, {}, {}
-            return path, [], []
+            return
 
         for info in listing:
             # each info name must be at least [path]/part , but here
@@ -650,7 +648,7 @@ class AbstractFileSystem(metaclass=_Cached):
         Returns a single dictionary, with exactly the same information as ``ls``
         would with ``detail=True``.
 
-        The default implementation should calls ls and could be overridden by a
+        The default implementation calls ls and could be overridden by a
         shortcut. kwargs are passed on to ```ls()``.
 
         Some file systems might not be able to measure the file's size, in