|
| 1 | +"""This module contains storage classes related to Azure Blob Storage (ABS)""" |
| 2 | + |
| 3 | +from typing import Optional |
| 4 | +import warnings |
| 5 | + |
| 6 | +from numcodecs.compat import ensure_bytes |
| 7 | +from zarr.util import normalize_storage_path |
| 8 | +from zarr._storage.store import ( |
| 9 | + _get_metadata_suffix, |
| 10 | + data_root, |
| 11 | + meta_root, |
| 12 | + Store, |
| 13 | + StoreV3, |
| 14 | + V3_DEPRECATION_MESSAGE, |
| 15 | +) |
| 16 | +from zarr.types import DIMENSION_SEPARATOR |
| 17 | + |
| 18 | +__doctest_requires__ = { |
| 19 | + ("ABSStore", "ABSStore.*"): ["azure.storage.blob"], |
| 20 | +} |
| 21 | + |
| 22 | + |
| 23 | +class ABSStore(Store): |
| 24 | + """Storage class using Azure Blob Storage (ABS). |
| 25 | +
|
| 26 | + Parameters |
| 27 | + ---------- |
| 28 | + container : string |
| 29 | + The name of the ABS container to use. |
| 30 | +
|
| 31 | + .. deprecated:: |
| 32 | + Use ``client`` instead. |
| 33 | +
|
| 34 | + prefix : string |
| 35 | + Location of the "directory" to use as the root of the storage hierarchy |
| 36 | + within the container. |
| 37 | +
|
| 38 | + account_name : string |
| 39 | + The Azure blob storage account name. |
| 40 | +
|
| 41 | + .. deprecated:: 2.8.3 |
| 42 | + Use ``client`` instead. |
| 43 | +
|
| 44 | + account_key : string |
| 45 | + The Azure blob storage account access key. |
| 46 | +
|
| 47 | + .. deprecated:: 2.8.3 |
| 48 | + Use ``client`` instead. |
| 49 | +
|
| 50 | + blob_service_kwargs : dictionary |
| 51 | + Extra arguments to be passed into the azure blob client, for e.g. when |
| 52 | + using the emulator, pass in blob_service_kwargs={'is_emulated': True}. |
| 53 | +
|
| 54 | + .. deprecated:: 2.8.3 |
| 55 | + Use ``client`` instead. |
| 56 | +
|
| 57 | + dimension_separator : {'.', '/'}, optional |
| 58 | + Separator placed between the dimensions of a chunk. |
| 59 | +
|
| 60 | + client : azure.storage.blob.ContainerClient, optional |
| 61 | + And ``azure.storage.blob.ContainerClient`` to connect with. See |
| 62 | + `here <https://docs.microsoft.com/en-us/python/api/azure-storage-blob/azure.storage.blob.containerclient?view=azure-python>`_ # noqa |
| 63 | + for more. |
| 64 | +
|
| 65 | + .. versionadded:: 2.8.3 |
| 66 | +
|
| 67 | + Notes |
| 68 | + ----- |
| 69 | + In order to use this store, you must install the Microsoft Azure Storage SDK for Python, |
| 70 | + ``azure-storage-blob>=12.5.0``. |
| 71 | + """ # noqa: E501 |
| 72 | + |
| 73 | + def __init__( |
| 74 | + self, |
| 75 | + container=None, |
| 76 | + prefix="", |
| 77 | + account_name=None, |
| 78 | + account_key=None, |
| 79 | + blob_service_kwargs=None, |
| 80 | + dimension_separator: Optional[DIMENSION_SEPARATOR] = None, |
| 81 | + client=None, |
| 82 | + ): |
| 83 | + warnings.warn( |
| 84 | + V3_DEPRECATION_MESSAGE.format(store=self.__class__.__name__), |
| 85 | + FutureWarning, |
| 86 | + stacklevel=3, |
| 87 | + ) |
| 88 | + |
| 89 | + self._dimension_separator = dimension_separator |
| 90 | + self.prefix = normalize_storage_path(prefix) |
| 91 | + if client is None: |
| 92 | + # deprecated option, try to construct the client for them |
| 93 | + msg = ( |
| 94 | + "Providing 'container', 'account_name', 'account_key', and 'blob_service_kwargs'" |
| 95 | + "is deprecated. Provide and instance of 'azure.storage.blob.ContainerClient' " |
| 96 | + "'client' instead." |
| 97 | + ) |
| 98 | + warnings.warn(msg, FutureWarning, stacklevel=2) |
| 99 | + from azure.storage.blob import ContainerClient |
| 100 | + |
| 101 | + blob_service_kwargs = blob_service_kwargs or {} |
| 102 | + client = ContainerClient( |
| 103 | + f"https://{account_name}.blob.core.windows.net/", |
| 104 | + container, |
| 105 | + credential=account_key, |
| 106 | + **blob_service_kwargs, |
| 107 | + ) |
| 108 | + |
| 109 | + self.client = client |
| 110 | + self._container = container |
| 111 | + self._account_name = account_name |
| 112 | + self._account_key = account_key |
| 113 | + |
| 114 | + @staticmethod |
| 115 | + def _warn_deprecated(property_): |
| 116 | + msg = ( |
| 117 | + "The {} property is deprecated and will be removed in a future " |
| 118 | + "version. Get the property from 'ABSStore.client' instead." |
| 119 | + ) |
| 120 | + warnings.warn(msg.format(property_), FutureWarning, stacklevel=3) |
| 121 | + |
| 122 | + @property |
| 123 | + def container(self): |
| 124 | + self._warn_deprecated("container") |
| 125 | + return self._container |
| 126 | + |
| 127 | + @property |
| 128 | + def account_name(self): |
| 129 | + self._warn_deprecated("account_name") |
| 130 | + return self._account_name |
| 131 | + |
| 132 | + @property |
| 133 | + def account_key(self): |
| 134 | + self._warn_deprecated("account_key") |
| 135 | + return self._account_key |
| 136 | + |
| 137 | + def _append_path_to_prefix(self, path): |
| 138 | + if self.prefix == "": |
| 139 | + return normalize_storage_path(path) |
| 140 | + else: |
| 141 | + return "/".join([self.prefix, normalize_storage_path(path)]) |
| 142 | + |
| 143 | + @staticmethod |
| 144 | + def _strip_prefix_from_path(path, prefix): |
| 145 | + # normalized things will not have any leading or trailing slashes |
| 146 | + path_norm = normalize_storage_path(path) |
| 147 | + prefix_norm = normalize_storage_path(prefix) |
| 148 | + if prefix: |
| 149 | + return path_norm[(len(prefix_norm) + 1) :] |
| 150 | + else: |
| 151 | + return path_norm |
| 152 | + |
| 153 | + def __getitem__(self, key): |
| 154 | + from azure.core.exceptions import ResourceNotFoundError |
| 155 | + |
| 156 | + blob_name = self._append_path_to_prefix(key) |
| 157 | + try: |
| 158 | + return self.client.download_blob(blob_name).readall() |
| 159 | + except ResourceNotFoundError as e: |
| 160 | + raise KeyError(f"Blob {blob_name} not found") from e |
| 161 | + |
| 162 | + def __setitem__(self, key, value): |
| 163 | + value = ensure_bytes(value) |
| 164 | + blob_name = self._append_path_to_prefix(key) |
| 165 | + self.client.upload_blob(blob_name, value, overwrite=True) |
| 166 | + |
| 167 | + def __delitem__(self, key): |
| 168 | + from azure.core.exceptions import ResourceNotFoundError |
| 169 | + |
| 170 | + try: |
| 171 | + self.client.delete_blob(self._append_path_to_prefix(key)) |
| 172 | + except ResourceNotFoundError as e: |
| 173 | + raise KeyError(f"Blob {key} not found") from e |
| 174 | + |
| 175 | + def __eq__(self, other): |
| 176 | + return ( |
| 177 | + isinstance(other, ABSStore) |
| 178 | + and self.client == other.client |
| 179 | + and self.prefix == other.prefix |
| 180 | + ) |
| 181 | + |
| 182 | + def keys(self): |
| 183 | + return list(self.__iter__()) |
| 184 | + |
| 185 | + def __iter__(self): |
| 186 | + if self.prefix: |
| 187 | + list_blobs_prefix = self.prefix + "/" |
| 188 | + else: |
| 189 | + list_blobs_prefix = None |
| 190 | + for blob in self.client.list_blobs(list_blobs_prefix): |
| 191 | + yield self._strip_prefix_from_path(blob.name, self.prefix) |
| 192 | + |
| 193 | + def __len__(self): |
| 194 | + return len(self.keys()) |
| 195 | + |
| 196 | + def __contains__(self, key): |
| 197 | + blob_name = self._append_path_to_prefix(key) |
| 198 | + return self.client.get_blob_client(blob_name).exists() |
| 199 | + |
| 200 | + def listdir(self, path=None): |
| 201 | + dir_path = normalize_storage_path(self._append_path_to_prefix(path)) |
| 202 | + if dir_path: |
| 203 | + dir_path += "/" |
| 204 | + items = [ |
| 205 | + self._strip_prefix_from_path(blob.name, dir_path) |
| 206 | + for blob in self.client.walk_blobs(name_starts_with=dir_path, delimiter="/") |
| 207 | + ] |
| 208 | + return items |
| 209 | + |
| 210 | + def rmdir(self, path=None): |
| 211 | + dir_path = normalize_storage_path(self._append_path_to_prefix(path)) |
| 212 | + if dir_path: |
| 213 | + dir_path += "/" |
| 214 | + for blob in self.client.list_blobs(name_starts_with=dir_path): |
| 215 | + self.client.delete_blob(blob) |
| 216 | + |
| 217 | + def getsize(self, path=None): |
| 218 | + store_path = normalize_storage_path(path) |
| 219 | + fs_path = self._append_path_to_prefix(store_path) |
| 220 | + if fs_path: |
| 221 | + blob_client = self.client.get_blob_client(fs_path) |
| 222 | + else: |
| 223 | + blob_client = None |
| 224 | + |
| 225 | + if blob_client and blob_client.exists(): |
| 226 | + return blob_client.get_blob_properties().size |
| 227 | + else: |
| 228 | + size = 0 |
| 229 | + if fs_path == "": |
| 230 | + fs_path = None |
| 231 | + elif not fs_path.endswith("/"): |
| 232 | + fs_path += "/" |
| 233 | + for blob in self.client.walk_blobs(name_starts_with=fs_path, delimiter="/"): |
| 234 | + blob_client = self.client.get_blob_client(blob.name) |
| 235 | + if blob_client.exists(): |
| 236 | + size += blob_client.get_blob_properties().size |
| 237 | + return size |
| 238 | + |
| 239 | + def clear(self): |
| 240 | + self.rmdir() |
| 241 | + |
| 242 | + |
| 243 | +class ABSStoreV3(ABSStore, StoreV3): |
| 244 | + def list(self): |
| 245 | + return list(self.keys()) |
| 246 | + |
| 247 | + def __eq__(self, other): |
| 248 | + return ( |
| 249 | + isinstance(other, ABSStoreV3) |
| 250 | + and self.client == other.client |
| 251 | + and self.prefix == other.prefix |
| 252 | + ) |
| 253 | + |
| 254 | + def __setitem__(self, key, value): |
| 255 | + self._validate_key(key) |
| 256 | + super().__setitem__(key, value) |
| 257 | + |
| 258 | + def rmdir(self, path=None): |
| 259 | + if not path: |
| 260 | + # Currently allowing clear to delete everything as in v2 |
| 261 | + |
| 262 | + # If we disallow an empty path then we will need to modify |
| 263 | + # TestABSStoreV3 to have the create_store method use a prefix. |
| 264 | + ABSStore.rmdir(self, "") |
| 265 | + return |
| 266 | + |
| 267 | + meta_dir = meta_root + path |
| 268 | + meta_dir = meta_dir.rstrip("/") |
| 269 | + ABSStore.rmdir(self, meta_dir) |
| 270 | + |
| 271 | + # remove data folder |
| 272 | + data_dir = data_root + path |
| 273 | + data_dir = data_dir.rstrip("/") |
| 274 | + ABSStore.rmdir(self, data_dir) |
| 275 | + |
| 276 | + # remove metadata files |
| 277 | + sfx = _get_metadata_suffix(self) |
| 278 | + array_meta_file = meta_dir + ".array" + sfx |
| 279 | + if array_meta_file in self: |
| 280 | + del self[array_meta_file] |
| 281 | + group_meta_file = meta_dir + ".group" + sfx |
| 282 | + if group_meta_file in self: |
| 283 | + del self[group_meta_file] |
| 284 | + |
| 285 | + # TODO: adapt the v2 getsize method to work for v3 |
| 286 | + # For now, calling the generic keys-based _getsize |
| 287 | + def getsize(self, path=None): |
| 288 | + from zarr.storage import _getsize # avoid circular import |
| 289 | + |
| 290 | + return _getsize(self, path) |
| 291 | + |
| 292 | + |
| 293 | +ABSStoreV3.__doc__ = ABSStore.__doc__ |
0 commit comments