fs: make constructor closer to fsspec (#5911)

* fs: don't use repo for jobs/hash_jobs

* fs: local: exists: use fsspec's LocalFileSystem

* fs: make constructor comply with fsspec

dvc/objects/db/__init__.py

@@ -3,22 +3,22 @@ from collections import defaultdict
 from dvc.scheme import Schemes
 
 
-def get_odb(fs):
+def get_odb(fs, **config):
     from .base import ObjectDB
     from .gdrive import GDriveObjectDB
     from .local import LocalObjectDB
     from .ssh import SSHObjectDB
 
     if fs.scheme == Schemes.LOCAL:
-        return LocalObjectDB(fs)
+        return LocalObjectDB(fs, **config)
 
     if fs.scheme == Schemes.SSH:
-        return SSHObjectDB(fs)
+        return SSHObjectDB(fs, **config)
 
     if fs.scheme == Schemes.GDRIVE:
-        return GDriveObjectDB(fs)
+        return GDriveObjectDB(fs, **config)
 
-    return ObjectDB(fs)
+    return ObjectDB(fs, **config)
 
 
 def _get_odb(repo, settings):
@@ -27,8 +27,8 @@ def _get_odb(repo, settings):
     if not settings:
         return None
 
-    fs = get_cloud_fs(repo, **settings)
-    return get_odb(fs)
+    cls, config = get_cloud_fs(repo, **settings)
+    return get_odb(cls(**config), state=repo.state, **config)
 
 
 class ODBManager:

dvc/objects/db/base.py

@@ -18,15 +18,15 @@ class ObjectDB:
     DEFAULT_CACHE_TYPES = ["copy"]
     CACHE_MODE: Optional[int] = None
 
-    def __init__(self, fs):
-        self.fs = fs
-        self.repo = fs.repo
+    def __init__(self, fs, **config):
+        from dvc.state import StateNoop
 
-        self.verify = fs.config.get("verify", self.DEFAULT_VERIFY)
-        self.cache_types = fs.config.get("type") or copy(
-            self.DEFAULT_CACHE_TYPES
-        )
+        self.fs = fs
+        self.state = config.get("state", StateNoop())
+        self.verify = config.get("verify", self.DEFAULT_VERIFY)
+        self.cache_types = config.get("type") or copy(self.DEFAULT_CACHE_TYPES)
         self.cache_type_confirmed = False
+        self.slow_link_warning = config.get("slow_link_warning", True)
 
     def move(self, from_info, to_info):
         self.fs.move(from_info, to_info)
@@ -80,7 +80,7 @@ class ObjectDB:
                 raise
 
         self.protect(cache_info)
-        self.fs.repo.state.save(cache_info, self.fs, hash_info)
+        self.state.save(cache_info, self.fs, hash_info)
 
         callback = kwargs.get("download_callback")
         if callback:

dvc/objects/db/local.py

@@ -22,11 +22,11 @@ class LocalObjectDB(ObjectDB):
     CACHE_MODE = 0o444
     UNPACKED_DIR_SUFFIX = ".unpacked"
 
-    def __init__(self, fs):
-        super().__init__(fs)
-        self.cache_dir = fs.config.get("url")
+    def __init__(self, fs, **config):
+        super().__init__(fs, **config)
+        self.cache_dir = config.get("url")
 
-        shared = fs.config.get("shared")
+        shared = config.get("shared")
         if shared:
             self._file_mode = 0o664
             self._dir_mode = 0o2775

dvc/objects/file.py

@@ -49,7 +49,7 @@ class HashFile:
                 return None
 
         actual = get_file_hash(
-            self.path_info, self.fs, self.hash_info.name, odb.repo.state
+            self.path_info, self.fs, self.hash_info.name, odb.state
         )
 
         logger.trace(

dvc/objects/stage.py

@@ -186,7 +186,7 @@ def stage(odb, path_info, fs, name, upload=False, **kwargs):
             errno.ENOENT, os.strerror(errno.ENOENT), path_info
         )
 
-    state = odb.repo.state
+    state = odb.state
     # pylint: disable=assignment-from-none
     hash_info = state.get(path_info, fs)
 

dvc/objects/tree.py

@@ -40,7 +40,7 @@ class Tree(HashFile):
         from dvc.path_info import PathInfo
         from dvc.utils import tmp_fname
 
-        memfs = MemoryFileSystem(None, {})
+        memfs = MemoryFileSystem()
         path_info = PathInfo(tmp_fname(""))
         with memfs.open(path_info, "wb") as fobj:
             fobj.write(self.as_bytes())