Avoid serializing cache for file objects

fsspec/implementations/http.py

@@ -692,25 +692,6 @@ async def async_fetch_range(self, start, end):
 
     _fetch_range = sync_wrapper(async_fetch_range)
 
-    def __reduce__(self):
-        return (
-            reopen,
-            (
-                self.fs,
-                self.url,
-                self.mode,
-                self.blocksize,
-                self.cache.name if self.cache else "none",
-                self.size,
-            ),
-        )
-
-
-def reopen(fs, url, mode, blocksize, cache_type, size=None):
-    return fs.open(
-        url, mode=mode, block_size=blocksize, cache_type=cache_type, size=size
-    )
-
 
 magic_check = re.compile("([*[])")
 
@@ -760,9 +741,6 @@ def close(self):
         asyncio.run_coroutine_threadsafe(self._close(), self.loop)
         super().close()
 
-    def __reduce__(self):
-        return reopen, (self.fs, self.url, self.mode, self.blocksize, self.cache.name)
-
 
 class AsyncStreamFile(AbstractAsyncStreamedFile):
     def __init__(

fsspec/spec.py

@@ -2050,6 +2050,22 @@ def writable(self):
         """Whether opened for writing"""
         return self.mode in {"wb", "ab"} and not self.closed
 
+    def __reduce__(self):
+        if self.mode != "rb":
+            raise RuntimeError("Pickling a writeable file is not supported")
+
+        return reopen, (
+            self.fs,
+            self.path,
+            self.mode,
+            self.blocksize,
+            self.loc,
+            self.size,
+            self.autocommit,
+            self.cache.name if self.cache else "none",
+            self.kwargs,
+        )
+
     def __del__(self):
         if not self.closed:
             self.close()
@@ -2064,3 +2080,18 @@ def __enter__(self):
 
     def __exit__(self, *args):
         self.close()
+
+
+def reopen(fs, path, mode, blocksize, loc, size, autocommit, cache_type, kwargs):
+    file = fs.open(
+        path,
+        mode=mode,
+        block_size=blocksize,
+        autocommit=autocommit,
+        cache_type=cache_type,
+        size=size,
+        **kwargs,
+    )
+    if loc > 0:
+        file.seek(loc)
+    return file