Files
zenml/venv/lib/python3.9/site-packages/databricks/sdk/dbutils.py
Christian Mantha 2ca0b9ef7c star
2026-03-02 19:10:52 -05:00

476 lines
16 KiB
Python

import base64
import json
import logging
import os
import threading
from collections import namedtuple
from dataclasses import dataclass
from typing import Any, Callable, Dict, List, Optional
from .core import ApiClient, Config, DatabricksError
from .mixins import compute as compute_ext
from .mixins import files as dbfs_ext
from .service import compute, workspace
_LOG = logging.getLogger("databricks.sdk")
class FileInfo(namedtuple("FileInfo", ["path", "name", "size", "modificationTime"])):
pass
class MountInfo(namedtuple("MountInfo", ["mountPoint", "source", "encryptionType"])):
pass
class SecretScope(namedtuple("SecretScope", ["name"])):
def getName(self):
return self.name
class SecretMetadata(namedtuple("SecretMetadata", ["key"])):
pass
class _FsUtil:
"""Manipulates the Databricks filesystem (DBFS)"""
def __init__(
self,
dbfs_ext: dbfs_ext.DbfsExt,
proxy_factory: Callable[[str], "_ProxyUtil"],
):
self._dbfs = dbfs_ext
self._proxy_factory = proxy_factory
def cp(self, from_: str, to: str, recurse: bool = False) -> bool:
"""Copies a file or directory, possibly across FileSystems"""
self._dbfs.copy(from_, to, recursive=recurse)
return True
def head(self, file: str, maxBytes: int = 65536) -> str:
"""Returns up to the first 'maxBytes' bytes of the given file as a String encoded in UTF-8"""
with self._dbfs.download(file) as f:
return f.read(maxBytes).decode("utf8")
def ls(self, dir: str) -> List[FileInfo]:
"""Lists the contents of a directory"""
return [
FileInfo(
f.path,
os.path.basename(f.path),
f.file_size,
f.modification_time,
)
for f in self._dbfs.list(dir)
]
def mkdirs(self, dir: str) -> bool:
"""Creates the given directory if it does not exist, also creating any necessary parent directories"""
self._dbfs.mkdirs(dir)
return True
def mv(self, from_: str, to: str, recurse: bool = False) -> bool:
"""Moves a file or directory, possibly across FileSystems"""
self._dbfs.move_(from_, to, recursive=recurse, overwrite=True)
return True
def put(self, file: str, contents: str, overwrite: bool = False) -> bool:
"""Writes the given String out to a file, encoded in UTF-8"""
with self._dbfs.open(file, write=True, overwrite=overwrite) as f:
f.write(contents.encode("utf8"))
return True
def rm(self, dir: str, recurse: bool = False) -> bool:
"""Removes a file or directory"""
self._dbfs.delete(dir, recursive=recurse)
return True
def mount(
self,
source: str,
mount_point: str,
encryption_type: str = None,
owner: str = None,
extra_configs: Dict[str, str] = None,
) -> bool:
"""Mounts the given source directory into DBFS at the given mount point"""
fs = self._proxy_factory("fs")
kwargs = {}
if encryption_type:
kwargs["encryption_type"] = encryption_type
if owner:
kwargs["owner"] = owner
if extra_configs:
kwargs["extra_configs"] = extra_configs
return fs.mount(source, mount_point, **kwargs)
def unmount(self, mount_point: str) -> bool:
"""Deletes a DBFS mount point"""
fs = self._proxy_factory("fs")
return fs.unmount(mount_point)
def updateMount(
self,
source: str,
mount_point: str,
encryption_type: str = None,
owner: str = None,
extra_configs: Dict[str, str] = None,
) -> bool:
"""Similar to mount(), but updates an existing mount point (if present) instead of creating a new one"""
fs = self._proxy_factory("fs")
kwargs = {}
if encryption_type:
kwargs["encryption_type"] = encryption_type
if owner:
kwargs["owner"] = owner
if extra_configs:
kwargs["extra_configs"] = extra_configs
return fs.updateMount(source, mount_point, **kwargs)
def mounts(self) -> List[MountInfo]:
"""Displays information about what is mounted within DBFS"""
result = []
fs = self._proxy_factory("fs")
for info in fs.mounts():
result.append(MountInfo(info[0], info[1], info[2]))
return result
def refreshMounts(self) -> bool:
"""Forces all machines in this cluster to refresh their mount cache,
ensuring they receive the most recent information"""
fs = self._proxy_factory("fs")
return fs.refreshMounts()
class _SecretsUtil:
"""Remote equivalent of secrets util"""
def __init__(self, secrets_api: workspace.SecretsAPI):
self._api = secrets_api # nolint
def getBytes(self, scope: str, key: str) -> bytes:
"""Gets the bytes representation of a secret value for the specified scope and key."""
query = {"scope": scope, "key": key}
raw = self._api._api.do("GET", "/api/2.0/secrets/get", query=query)
return base64.b64decode(raw["value"])
def get(self, scope: str, key: str) -> str:
"""Gets the string representation of a secret value for the specified secrets scope and key."""
val = self.getBytes(scope, key)
string_value = val.decode()
return string_value
def list(self, scope) -> List[SecretMetadata]:
"""Lists the metadata for secrets within the specified scope."""
# transform from SDK dataclass to dbutils-compatible namedtuple
return [SecretMetadata(v.key) for v in self._api.list_secrets(scope)]
def listScopes(self) -> List[SecretScope]:
"""Lists the available scopes."""
# transform from SDK dataclass to dbutils-compatible namedtuple
return [SecretScope(v.name) for v in self._api.list_scopes()]
class _JobsUtil:
"""Remote equivalent of jobs util"""
class _TaskValuesUtil:
"""Remote equivalent of task values util"""
def get(
self,
taskKey: str,
key: str,
default: any = None,
debugValue: any = None,
) -> None:
"""
Returns `debugValue` if present, throws an error otherwise as this implementation is always run outside of a job run
"""
if debugValue is None:
raise TypeError(
"Must pass debugValue when calling get outside of a job context. debugValue cannot be None."
)
return debugValue
def set(self, key: str, value: any) -> None:
"""
Sets a task value on the current task run
"""
def __init__(self) -> None:
self.taskValues = self._TaskValuesUtil()
class RemoteDbUtils:
def __init__(self, config: "Config" = None):
# Create a shallow copy of the config to allow the use of a custom
# user-agent while avoiding modifying the original config.
self._config = Config() if not config else config.copy()
self._config.with_user_agent_extra("dbutils", "remote")
self._client = ApiClient(self._config)
self._clusters = compute_ext.ClustersExt(self._client)
self._commands = compute.CommandExecutionAPI(self._client)
self._lock = threading.Lock()
self._ctx = None
self.fs = _FsUtil(dbfs_ext.DbfsExt(self._client), self.__getattr__)
self.secrets = _SecretsUtil(workspace.SecretsAPI(self._client))
self.jobs = _JobsUtil()
self._widgets = None
# When we import widget_impl, the init file checks whether user has the
# correct dependencies required for running on notebook or not (ipywidgets etc).
# We only want these checks (and the subsequent errors and warnings), to
# happen when the user actually uses widgets.
@property
def widgets(self):
if self._widgets is None:
from ._widgets import widget_impl
self._widgets = widget_impl()
return self._widgets
@property
def _cluster_id(self) -> str:
cluster_id = self._config.cluster_id
if not cluster_id:
message = "cluster_id is required in the configuration"
raise ValueError(self._config.wrap_debug_info(message))
return cluster_id
def _running_command_context(self) -> compute.ContextStatusResponse:
if self._ctx:
return self._ctx
with self._lock:
if self._ctx:
return self._ctx
self._clusters.ensure_cluster_is_running(self._cluster_id)
self._ctx = self._commands.create(cluster_id=self._cluster_id, language=compute.Language.PYTHON).result()
return self._ctx
def __getattr__(self, util) -> "_ProxyUtil":
return _ProxyUtil(
command_execution=self._commands,
context_factory=self._running_command_context,
cluster_id=self._cluster_id,
name=util,
)
@dataclass
class OverrideResult:
result: Any
def get_local_notebook_path():
value = os.getenv("DATABRICKS_SOURCE_FILE")
if value is None:
raise ValueError(
"Getting the current notebook path is only supported when running a notebook using the `Databricks Connect: Run as File` or `Databricks Connect: Debug as File` commands in the Databricks extension for VS Code. To bypass this error, set environment variable `DATABRICKS_SOURCE_FILE` to the desired notebook path."
)
return value
def not_supported_method_err_msg(methodName):
return f"Method '{methodName}' is not supported in the SDK version of DBUtils"
class _OverrideProxyUtil:
@classmethod
def new(cls, path: str):
if path in cls.not_supported_override_paths:
raise ValueError(cls.not_supported_override_paths[path])
if len(cls.__get_matching_overrides(path)) > 0:
return _OverrideProxyUtil(path)
return None
def __init__(self, name: str):
self._name = name
# These are the paths that we want to override and not send to remote dbutils. NOTE, for each of these paths, no prefixes
# are sent to remote either. This could lead to unintentional breakage.
# Our current proxy implementation (which sends everything to remote dbutils) uses `{util}.{method}(*args, **kwargs)` ONLY.
# This means, it is completely safe to override paths starting with `{util}.{attribute}.<other_parts>`, since none of the prefixes
# are being proxied to remote dbutils currently.
proxy_override_paths = {
"notebook.entry_point.getDbutils().notebook().getContext().notebookPath().get()": get_local_notebook_path,
}
# These paths work the same as 'proxy_override_paths' but instead of using a local implementation we raise an exception.
not_supported_override_paths = {
# The object returned by 'credentials.getServiceCredentialProvider()' can't be serialized to JSON.
# Without this override, the command would fail with an error 'TypeError: Object of type Session is not JSON serializable'.
# We override it to show a better error message
"credentials.getServiceCredentialsProvider": not_supported_method_err_msg(
"credentials.getServiceCredentialsProvider"
),
}
@classmethod
def __get_matching_overrides(cls, path: str):
return [x for x in cls.proxy_override_paths.keys() if x.startswith(path)]
def __run_override(self, path: str) -> Optional[OverrideResult]:
overrides = self.__get_matching_overrides(path)
if len(overrides) == 1 and overrides[0] == path:
return OverrideResult(self.proxy_override_paths[overrides[0]]())
if len(overrides) > 0:
return OverrideResult(_OverrideProxyUtil(name=path))
return None
def __call__(self, *args, **kwds) -> Any:
if len(args) != 0 or len(kwds) != 0:
raise TypeError(
f"Arguments are not supported for overridden method {self._name}. Invoke as: {self._name}()"
)
callable_path = f"{self._name}()"
result = self.__run_override(callable_path)
if result:
return result.result
raise TypeError(f"{self._name} is not callable")
def __getattr__(self, method: str) -> Any:
result = self.__run_override(f"{self._name}.{method}")
if result:
return result.result
raise AttributeError(f"module {self._name} has no attribute {method}")
class _ProxyUtil:
"""Enables temporary workaround to call remote in-REPL dbutils without having to re-implement them"""
def __init__(
self,
*,
command_execution: compute.CommandExecutionAPI,
context_factory: Callable[[], compute.ContextStatusResponse],
cluster_id: str,
name: str,
):
self._commands = command_execution
self._cluster_id = cluster_id
self._context_factory = context_factory
self._name = name
def __call__(self):
raise NotImplementedError(f"dbutils.{self._name} is not callable")
def __getattr__(self, method: str) -> "_ProxyCall | _ProxyUtil | _OverrideProxyUtil":
override = _OverrideProxyUtil.new(f"{self._name}.{method}")
if override:
return override
return _ProxyCall(
command_execution=self._commands,
cluster_id=self._cluster_id,
context_factory=self._context_factory,
util=self._name,
method=method,
)
import html
import re
class _ProxyCall:
def __init__(
self,
*,
command_execution: compute.CommandExecutionAPI,
context_factory: Callable[[], compute.ContextStatusResponse],
cluster_id: str,
util: str,
method: str,
):
self._commands = command_execution
self._cluster_id = cluster_id
self._context_factory = context_factory
self._util = util
self._method = method
_out_re = re.compile(r"Out\[[\d\s]+]:\s")
_tag_re = re.compile(r"<[^>]*>")
_exception_re = re.compile(r".*Exception:\s+(.*)")
_execution_error_re = re.compile(r"ExecutionError: ([\s\S]*)\n(StatusCode=[0-9]*)\n(StatusDescription=.*)\n")
_error_message_re = re.compile(r"ErrorMessage=(.+)\n")
_ascii_escape_re = re.compile(r"(\x9B|\x1B\[)[0-?]*[ -/]*[@-~]")
def _is_failed(self, results: compute.Results) -> bool:
return results.result_type == compute.ResultType.ERROR
def _text(self, results: compute.Results) -> str:
if results.result_type != compute.ResultType.TEXT:
return ""
return self._out_re.sub("", str(results.data))
def _raise_if_failed(self, results: compute.Results):
if not self._is_failed(results):
return
raise DatabricksError(self._error_from_results(results))
def _error_from_results(self, results: compute.Results):
if not self._is_failed(results):
return
if results.cause:
_LOG.debug(f'{self._ascii_escape_re.sub("", results.cause)}')
summary = self._tag_re.sub("", results.summary)
summary = html.unescape(summary)
exception_matches = self._exception_re.findall(summary)
if len(exception_matches) == 1:
summary = exception_matches[0].replace("; nested exception is:", "")
summary = summary.rstrip(" ")
return summary
execution_error_matches = self._execution_error_re.findall(results.cause)
if len(execution_error_matches) == 1:
return "\n".join(execution_error_matches[0])
error_message_matches = self._error_message_re.findall(results.cause)
if len(error_message_matches) == 1:
return error_message_matches[0]
return summary
def __call__(self, *args, **kwargs):
raw = json.dumps((args, kwargs))
code = f"""
import json
(args, kwargs) = json.loads('{raw}')
result = dbutils.{self._util}.{self._method}(*args, **kwargs)
dbutils.notebook.exit(json.dumps(result))
"""
ctx = self._context_factory()
result = self._commands.execute(
cluster_id=self._cluster_id,
language=compute.Language.PYTHON,
context_id=ctx.id,
command=code,
).result()
if result.status == compute.CommandStatus.FINISHED:
self._raise_if_failed(result.results)
raw = result.results.data
return json.loads(raw)
else:
raise Exception(result.results.summary)