anemoi-utils 0.4.3__py3-none-any.whl → 0.4.5__py3-none-any.whl

anemoi/utils/remote/__init__.py

@@ -0,0 +1,328 @@
+# (C) Copyright 2024 European Centre for Medium-Range Weather Forecasts.
+# This software is licensed under the terms of the Apache Licence Version 2.0
+# which can be obtained at http://www.apache.org/licenses/LICENSE-2.0.
+# In applying this licence, ECMWF does not waive the privileges and immunities
+# granted to it by virtue of its status as an intergovernmental organisation
+# nor does it submit to any jurisdiction.
+
+import concurrent.futures
+import logging
+import os
+import shutil
+from abc import abstractmethod
+
+import tqdm
+
+from ..humanize import bytes_to_human
+
+LOGGER = logging.getLogger(__name__)
+
+
+def _ignore(number_of_files, total_size, total_transferred, transfering):
+    pass
+
+
+class Loader:
+
+    def transfer_folder(self, *, source, target, overwrite=False, resume=False, verbosity=1, threads=1, progress=None):
+        assert verbosity == 1, verbosity
+
+        if progress is None:
+            progress = _ignore
+
+        # from boto3.s3.transfer import TransferConfig
+        # config = TransferConfig(use_threads=False)
+        config = None
+        with concurrent.futures.ThreadPoolExecutor(max_workers=threads) as executor:
+            try:
+                if verbosity > 0:
+                    LOGGER.info(f"{self.action} {source} to {target}")
+
+                total_size = 0
+                total_transferred = 0
+
+                futures = []
+                for name in self.list_source(source):
+
+                    futures.append(
+                        executor.submit(
+                            self.transfer_file,
+                            source=self.source_path(name, source),
+                            target=self.target_path(name, source, target),
+                            overwrite=overwrite,
+                            resume=resume,
+                            verbosity=verbosity - 1,
+                            config=config,
+                        )
+                    )
+                    total_size += self.source_size(name)
+
+                    if len(futures) % 10000 == 0:
+
+                        progress(len(futures), total_size, 0, False)
+
+                        if verbosity > 0:
+                            LOGGER.info(f"Preparing transfer, {len(futures):,} files... ({bytes_to_human(total_size)})")
+                        done, _ = concurrent.futures.wait(
+                            futures,
+                            timeout=0.001,
+                            return_when=concurrent.futures.FIRST_EXCEPTION,
+                        )
+                        # Trigger exceptions if any
+                        for future in done:
+                            future.result()
+
+                number_of_files = len(futures)
+                progress(number_of_files, total_size, 0, True)
+
+                if verbosity > 0:
+                    LOGGER.info(f"{self.action} {number_of_files:,} files ({bytes_to_human(total_size)})")
+                    with tqdm.tqdm(total=total_size, unit="B", unit_scale=True, unit_divisor=1024) as pbar:
+                        for future in concurrent.futures.as_completed(futures):
+                            size = future.result()
+                            pbar.update(size)
+                            total_transferred += size
+                            progress(number_of_files, total_size, total_transferred, True)
+                else:
+                    for future in concurrent.futures.as_completed(futures):
+                        size = future.result()
+                        total_transferred += size
+                        progress(number_of_files, total_size, total_transferred, True)
+
+            except Exception:
+                executor.shutdown(wait=False, cancel_futures=True)
+                raise
+
+    def transfer_file(self, source, target, overwrite, resume, verbosity, threads=1, progress=None, config=None):
+        try:
+            return self._transfer_file(source, target, overwrite, resume, verbosity, threads=threads, config=config)
+        except Exception as e:
+            LOGGER.exception(f"Error transferring {source} to {target}")
+            LOGGER.error(e)
+            raise
+
+    @abstractmethod
+    def list_source(self, source):
+        raise NotImplementedError
+
+    @abstractmethod
+    def source_path(self, local_path, source):
+        raise NotImplementedError
+
+    @abstractmethod
+    def target_path(self, source_path, source, target):
+        raise NotImplementedError
+
+    @abstractmethod
+    def source_size(self, local_path):
+        raise NotImplementedError
+
+    @abstractmethod
+    def copy(self, source, target, **kwargs):
+        raise NotImplementedError
+
+    @abstractmethod
+    def get_temporary_target(self, target, pattern):
+        raise NotImplementedError
+
+    @abstractmethod
+    def rename_target(self, target, temporary_target):
+        raise NotImplementedError
+
+
+class BaseDownload(Loader):
+    action = "Downloading"
+
+    @abstractmethod
+    def copy(self, source, target, **kwargs):
+        raise NotImplementedError
+
+    def get_temporary_target(self, target, pattern):
+        dirname, basename = os.path.split(target)
+        return pattern.format(dirname=dirname, basename=basename)
+
+    def rename_target(self, target, new_target):
+        os.rename(target, new_target)
+
+    def delete_target(self, target):
+        if os.path.exists(target):
+            shutil.rmtree(target)
+
+
+class BaseUpload(Loader):
+    action = "Uploading"
+
+    def copy(self, source, target, **kwargs):
+        if os.path.isdir(source):
+            self.transfer_folder(source=source, target=target, **kwargs)
+        else:
+            self.transfer_file(source=source, target=target, **kwargs)
+
+    def list_source(self, source):
+        for root, _, files in os.walk(source):
+            for file in files:
+                yield os.path.join(root, file)
+
+    def source_path(self, local_path, source):
+        return local_path
+
+    def target_path(self, source_path, source, target):
+        relative_path = os.path.relpath(source_path, source)
+        path = os.path.join(target, relative_path)
+        return path
+
+    def source_size(self, local_path):
+        return os.path.getsize(local_path)
+
+
+class TransferMethodNotImplementedError(NotImplementedError):
+    pass
+
+
+class Transfer:
+    """This is the internal API and should not be used directly. Use the transfer function instead."""
+
+    TransferMethodNotImplementedError = TransferMethodNotImplementedError
+
+    def __init__(
+        self,
+        source,
+        target,
+        overwrite=False,
+        resume=False,
+        verbosity=1,
+        threads=1,
+        progress=None,
+        temporary_target=False,
+    ):
+        if target == ".":
+            target = os.path.basename(source)
+
+        temporary_target = {
+            False: "{dirname}/{basename}",
+            True: "{dirname}-downloading/{basename}",
+            "-tmp/*": "{dirname}-tmp/{basename}",
+            "*-tmp": "{dirname}/{basename}-tmp",
+            "tmp-*": "{dirname}/tmp-{basename}",
+        }.get(temporary_target, temporary_target)
+        assert isinstance(temporary_target, str), (type(temporary_target), temporary_target)
+
+        self.source = source
+        self.target = target
+        self.overwrite = overwrite
+        self.resume = resume
+        self.verbosity = verbosity
+        self.threads = threads
+        self.progress = progress
+        self.temporary_target = temporary_target
+
+        cls = _find_transfer_class(self.source, self.target)
+        self.loader = cls()
+
+    def run(self):
+
+        target = self.loader.get_temporary_target(self.target, self.temporary_target)
+        if target != self.target:
+            LOGGER.info(f"Using temporary target {target} to copy to {self.target}")
+
+        if self.overwrite:
+            # delete the target if it exists
+            LOGGER.info(f"Deleting {self.target}")
+            self.delete_target(target)
+
+            # carefully delete the temporary target if it exists
+            head, tail = os.path.split(self.target)
+            head_, tail_ = os.path.split(target)
+            if not head_.startswith(head) or tail not in tail_:
+                LOGGER.info(f"{target} is too different from {self.target} to delete it automatically.")
+            else:
+                self.delete_target(target)
+
+        self.loader.copy(
+            self.source,
+            target,
+            overwrite=self.overwrite,
+            resume=self.resume,
+            verbosity=self.verbosity,
+            threads=self.threads,
+            progress=self.progress,
+        )
+
+        self.rename_target(target, self.target)
+
+        return self
+
+    def rename_target(self, target, new_target):
+        if target != new_target:
+            LOGGER.info(f"Renaming temporary target {target} into {self.target}")
+            return self.loader.rename_target(target, new_target)
+
+    def delete_target(self, target):
+        return self.loader.delete_target(target)
+
+
+def _find_transfer_class(source, target):
+    from_ssh = source.startswith("ssh://")
+    into_ssh = target.startswith("ssh://")
+
+    from_s3 = source.startswith("s3://")
+    into_s3 = target.startswith("s3://")
+
+    from_local = not from_ssh and not from_s3
+    into_local = not into_ssh and not into_s3
+
+    # check that exactly one source type and one target type is specified
+    assert sum([into_ssh, into_local, into_s3]) == 1, (into_ssh, into_local, into_s3)
+    assert sum([from_ssh, from_local, from_s3]) == 1, (from_ssh, from_local, from_s3)
+
+    if from_local and into_ssh:  # local -> ssh
+        from .ssh import RsyncUpload
+
+        return RsyncUpload
+
+    if from_s3 and into_local:  # local <- S3
+        from .s3 import S3Download
+
+        return S3Download
+
+    if from_local and into_s3:  # local -> S3
+        from .s3 import S3Upload
+
+        return S3Upload
+
+    raise TransferMethodNotImplementedError(f"Transfer from {source} to {target} is not implemented")
+
+
+# this is the public API
+def transfer(*args, **kwargs) -> Loader:
+    """Parameters
+    ----------
+    source : str
+        A path to a local file or folder or a URL to a file or a folder on S3.
+        The url should start with 's3://'.
+    target : str
+        A path to a local file or folder or a URL to a file or a folder on S3 or a remote folder.
+        The url should start with 's3://' or 'ssh://'.
+    overwrite : bool, optional
+        If the data is alreay on in the target location it will be overwritten.
+        By default False
+    resume : bool, optional
+        If the data is alreay on S3 it will not be uploaded, unless the remote file has a different size
+        Ignored if the target is an SSH remote folder (ssh://).
+        By default False
+    verbosity : int, optional
+        The level of verbosity, by default 1
+    progress: callable, optional
+        A callable that will be called with the number of files, the total size of the files, the total size
+        transferred and a boolean indicating if the transfer has started. By default None
+    threads : int, optional
+        The number of threads to use when uploading a directory, by default 1
+    temporary_target : bool, optional
+        Experimental feature
+        If True and if the target location supports it, the data will be uploaded to a temporary location
+        then renamed to the final location. Supported by SSH and local targets, not supported by S3.
+        By default False
+    """
+    copier = Transfer(*args, **kwargs)
+    copier.run()
+    return copier

anemoi/utils/remote/s3.py

@@ -0,0 +1,386 @@
+# (C) Copyright 2024 European Centre for Medium-Range Weather Forecasts.
+# This software is licensed under the terms of the Apache Licence Version 2.0
+# which can be obtained at http://www.apache.org/licenses/LICENSE-2.0.
+# In applying this licence, ECMWF does not waive the privileges and immunities
+# granted to it by virtue of its status as an intergovernmental organisation
+# nor does it submit to any jurisdiction.
+
+"""This module provides functions to upload, download, list and delete files and folders on S3.
+The functions of this package expect that the AWS credentials are set up in the environment
+typicaly by setting the `AWS_ACCESS_KEY_ID` and `AWS_SECRET_ACCESS_KEY` environment variables or
+by creating a `~/.aws/credentials` file. It is also possible to set the `endpoint_url` in the same file
+to use a different S3 compatible service::
+
+    [default]
+    endpoint_url = https://some-storage.somewhere.world
+    aws_access_key_id = xxxxxxxxxxxxxxxxxxxxxxxx
+    aws_secret_access_key = xxxxxxxxxxxxxxxxxxxxxxxx
+
+Alternatively, the `endpoint_url`, and keys can be set in one of
+the `~/.config/anemoi/settings.toml`
+or `~/.config/anemoi/settings-secrets.toml` files.
+
+"""
+
+import logging
+import os
+import threading
+from copy import deepcopy
+from typing import Iterable
+
+import tqdm
+
+from ..config import load_config
+from ..humanize import bytes_to_human
+from . import BaseDownload
+from . import BaseUpload
+
+LOGGER = logging.getLogger(__name__)
+
+
+# s3_clients are not thread-safe, so we need to create a new client for each thread
+
+thread_local = threading.local()
+
+
+def s3_client(bucket, region=None):
+    import boto3
+    from botocore import UNSIGNED
+    from botocore.client import Config
+
+    if not hasattr(thread_local, "s3_clients"):
+        thread_local.s3_clients = {}
+
+    key = f"{bucket}-{region}"
+
+    boto3_config = dict(max_pool_connections=25)
+
+    if key in thread_local.s3_clients:
+        return thread_local.s3_clients[key]
+
+    boto3_config = dict(max_pool_connections=25)
+
+    if region:
+        # This is using AWS
+
+        options = {"region_name": region}
+
+        # Anonymous access
+        if not (
+            os.path.exists(os.path.expanduser("~/.aws/credentials"))
+            or ("AWS_ACCESS_KEY_ID" in os.environ and "AWS_SECRET_ACCESS_KEY" in os.environ)
+        ):
+            boto3_config["signature_version"] = UNSIGNED
+
+    else:
+
+        # We may be accessing a different S3 compatible service
+        # Use anemoi.config to get the configuration
+
+        options = {}
+        config = load_config(secrets=["aws_access_key_id", "aws_secret_access_key"])
+
+        cfg = config.get("object-storage", {})
+        for k, v in cfg.items():
+            if isinstance(v, (str, int, float, bool)):
+                options[k] = v
+
+        for k, v in cfg.get(bucket, {}).items():
+            if isinstance(v, (str, int, float, bool)):
+                options[k] = v
+
+        type = options.pop("type", "s3")
+        if type != "s3":
+            raise ValueError(f"Unsupported object storage type {type}")
+
+        if "config" in options:
+            boto3_config.update(options["config"])
+            del options["config"]
+            from botocore.client import Config
+
+    options["config"] = Config(**boto3_config)
+
+    thread_local.s3_clients[key] = boto3.client("s3", **options)
+
+    return thread_local.s3_clients[key]
+
+
+class S3Upload(BaseUpload):
+
+    def get_temporary_target(self, target, pattern):
+        return target
+
+    def rename_target(self, target, temporary_target):
+        pass
+
+    def delete_target(self, target):
+        pass
+        # delete(target)
+
+    def _transfer_file(self, source, target, overwrite, resume, verbosity, threads, config=None):
+
+        from botocore.exceptions import ClientError
+
+        assert target.startswith("s3://")
+
+        _, _, bucket, key = target.split("/", 3)
+        s3 = s3_client(bucket)
+
+        size = os.path.getsize(source)
+
+        if verbosity > 0:
+            LOGGER.info(f"{self.action} {source} to {target} ({bytes_to_human(size)})")
+
+        try:
+            results = s3.head_object(Bucket=bucket, Key=key)
+            remote_size = int(results["ContentLength"])
+        except ClientError as e:
+            if e.response["Error"]["Code"] != "404":
+                raise
+            remote_size = None
+
+        if remote_size is not None:
+            if remote_size != size:
+                LOGGER.warning(
+                    f"{target} already exists, but with different size, re-uploading (remote={remote_size}, local={size})"
+                )
+            elif resume:
+                # LOGGER.info(f"{target} already exists, skipping")
+                return size
+
+        if remote_size is not None and not overwrite and not resume:
+            raise ValueError(f"{target} already exists, use 'overwrite' to replace or 'resume' to skip")
+
+        if verbosity > 0:
+            with tqdm.tqdm(total=size, unit="B", unit_scale=True, unit_divisor=1024, leave=False) as pbar:
+                s3.upload_file(source, bucket, key, Callback=lambda x: pbar.update(x), Config=config)
+        else:
+            s3.upload_file(source, bucket, key, Config=config)
+
+        return size
+
+
+class S3Download(BaseDownload):
+
+    def copy(self, source, target, **kwargs):
+        assert source.startswith("s3://")
+
+        if source.endswith("/"):
+            self.transfer_folder(source=source, target=target, **kwargs)
+        else:
+            self.transfer_file(source=source, target=target, **kwargs)
+
+    def list_source(self, source):
+        yield from _list_objects(source)
+
+    def source_path(self, s3_object, source):
+        _, _, bucket, _ = source.split("/", 3)
+        return f"s3://{bucket}/{s3_object['Key']}"
+
+    def target_path(self, s3_object, source, target):
+        _, _, _, folder = source.split("/", 3)
+        local_path = os.path.join(target, os.path.relpath(s3_object["Key"], folder))
+        os.makedirs(os.path.dirname(local_path), exist_ok=True)
+        return local_path
+
+    def source_size(self, s3_object):
+        return s3_object["Size"]
+
+    def _transfer_file(self, source, target, overwrite, resume, verbosity, threads, config=None):
+        # from boto3.s3.transfer import TransferConfig
+
+        _, _, bucket, key = source.split("/", 3)
+        s3 = s3_client(bucket)
+
+        try:
+            response = s3.head_object(Bucket=bucket, Key=key)
+        except s3.exceptions.ClientError as e:
+            if e.response["Error"]["Code"] == "404":
+                raise ValueError(f"{source} does not exist ({bucket}, {key})")
+            raise
+
+        size = int(response["ContentLength"])
+
+        if verbosity > 0:
+            LOGGER.info(f"{self.action} {source} to {target} ({bytes_to_human(size)})")
+
+        if overwrite:
+            resume = False
+
+        if resume:
+            if os.path.exists(target):
+                local_size = os.path.getsize(target)
+                if local_size != size:
+                    LOGGER.warning(
+                        f"{target} already with different size, re-downloading (remote={size}, local={local_size})"
+                    )
+                else:
+                    # if verbosity > 0:
+                    #     LOGGER.info(f"{target} already exists, skipping")
+                    return size
+
+        if os.path.exists(target) and not overwrite:
+            raise ValueError(f"{target} already exists, use 'overwrite' to replace or 'resume' to skip")
+
+        if verbosity > 0:
+            with tqdm.tqdm(total=size, unit="B", unit_scale=True, unit_divisor=1024, leave=False) as pbar:
+                s3.download_file(bucket, key, target, Callback=lambda x: pbar.update(x), Config=config)
+        else:
+            s3.download_file(bucket, key, target, Config=config)
+
+        return size
+
+
+def _list_objects(target, batch=False):
+    _, _, bucket, prefix = target.split("/", 3)
+    s3 = s3_client(bucket)
+
+    paginator = s3.get_paginator("list_objects_v2")
+
+    for page in paginator.paginate(Bucket=bucket, Prefix=prefix):
+        if "Contents" in page:
+            objects = deepcopy(page["Contents"])
+            if batch:
+                yield objects
+            else:
+                yield from objects
+
+
+def _delete_folder(target) -> None:
+    _, _, bucket, _ = target.split("/", 3)
+    s3 = s3_client(bucket)
+
+    total = 0
+    for batch in _list_objects(target, batch=True):
+        LOGGER.info(f"Deleting {len(batch):,} objects from {target}")
+        s3.delete_objects(Bucket=bucket, Delete={"Objects": [{"Key": o["Key"]} for o in batch]})
+        total += len(batch)
+        LOGGER.info(f"Deleted {len(batch):,} objects (total={total:,})")
+
+
+def _delete_file(target) -> None:
+    from botocore.exceptions import ClientError
+
+    _, _, bucket, key = target.split("/", 3)
+    s3 = s3_client(bucket)
+
+    try:
+        s3.head_object(Bucket=bucket, Key=key)
+        exits = True
+    except ClientError as e:
+        if e.response["Error"]["Code"] != "404":
+            raise
+        exits = False
+
+    if not exits:
+        LOGGER.warning(f"{target} does not exist. Did you mean to delete a folder? Then add a trailing '/'")
+        return
+
+    LOGGER.info(f"Deleting {target}")
+    s3.delete_object(Bucket=bucket, Key=key)
+    LOGGER.info(f"{target} is deleted")
+
+
+def delete(target) -> None:
+    """Delete a file or a folder from S3.
+
+    Parameters
+    ----------
+    target : str
+        The URL of a file or a folder on S3. The url should start with 's3://'. If the URL ends with a '/' it is
+        assumed to be a folder, otherwise it is assumed to be a file.
+    """
+
+    assert target.startswith("s3://")
+
+    if target.endswith("/"):
+        _delete_folder(target)
+    else:
+        _delete_file(target)
+
+
+def list_folder(folder) -> Iterable:
+    """List the sub folders in a folder on S3.
+
+    Parameters
+    ----------
+    folder : str
+        The URL of a folder on S3. The url should start with 's3://'.
+
+    Returns
+    -------
+    list
+        A list of the subfolders names in the folder.
+    """
+
+    assert folder.startswith("s3://")
+    if not folder.endswith("/"):
+        folder += "/"
+
+    _, _, bucket, prefix = folder.split("/", 3)
+
+    s3 = s3_client(bucket)
+    paginator = s3.get_paginator("list_objects_v2")
+
+    for page in paginator.paginate(Bucket=bucket, Prefix=prefix, Delimiter="/"):
+        if "CommonPrefixes" in page:
+            yield from [folder + _["Prefix"] for _ in page.get("CommonPrefixes")]
+
+
+def object_info(target) -> dict:
+    """Get information about an object on S3.
+
+    Parameters
+    ----------
+    target : str
+        The URL of a file or a folder on S3. The url should start with 's3://'.
+
+    Returns
+    -------
+    dict
+        A dictionary with information about the object.
+    """
+
+    _, _, bucket, key = target.split("/", 3)
+    s3 = s3_client(bucket)
+
+    try:
+        return s3.head_object(Bucket=bucket, Key=key)
+    except s3.exceptions.ClientError as e:
+        if e.response["Error"]["Code"] == "404":
+            raise ValueError(f"{target} does not exist")
+        raise
+
+
+def object_acl(target) -> dict:
+    """Get information about an object's ACL on S3.
+
+    Parameters
+    ----------
+    target : str
+        The URL of a file or a folder on S3. The url should start with 's3://'.
+
+    Returns
+    -------
+    dict
+        A dictionary with information about the object's ACL.
+    """
+
+    _, _, bucket, key = target.split("/", 3)
+    s3 = s3_client()
+
+    return s3.get_object_acl(Bucket=bucket, Key=key)
+
+
+def download(source, target, *args, **kwargs):
+    from . import transfer
+
+    assert source.startswith("s3://"), f"source {source} should start with 's3://'"
+    return transfer(source, target, *args, **kwargs)
+
+
+def upload(source, target, *args, **kwargs):
+    from . import transfer
+
+    assert target.startswith("s3://"), f"target {target} should start with 's3://'"
+    return transfer(source, target, *args, **kwargs)