anemoi-utils 0.3.3__py3-none-any.whl → 0.3.5__py3-none-any.whl

anemoi/utils/_version.py

@@ -12,5 +12,5 @@ __version__: str
 __version_tuple__: VERSION_TUPLE
 version_tuple: VERSION_TUPLE
 
-__version__ = version = '0.3.3'
-__version_tuple__ = version_tuple = (0, 3, 3)
+__version__ = version = '0.3.5'
+__version_tuple__ = version_tuple = (0, 3, 5)

anemoi/utils/cli.py

@@ -64,7 +64,7 @@ def register_commands(here, package, select, fail=None):
     result = {}
     not_available = {}
 
-    for p in os.listdir(here):
+    for p in sorted(os.listdir(here)):
         full = os.path.join(here, p)
         if p.startswith("_"):
             continue

anemoi/utils/config.py

@@ -8,6 +8,7 @@
 
 import logging
 import os
+import threading
 
 try:
     import tomllib  # Only available since 3.11
@@ -60,16 +61,10 @@ class DotDict(dict):
 
 
 CONFIG = None
+CONFIG_LOCK = threading.Lock()
 
 
-def load_config():
-    """Load the configuration from `~/.anemoi.toml`.
-
-    Returns
-    -------
-    DotDict
-        The configuration
-    """
+def _load_config():
     global CONFIG
     if CONFIG is not None:
         return CONFIG
@@ -86,9 +81,20 @@ def load_config():
     return DotDict(CONFIG)
 
 
-def save_config():
-    """Save the configuration to `~/.anemoi.toml`."""
+def load_config():
+    """Load the configuration from `~/.anemoi.toml`.
+
+    Returns
+    -------
+    DotDict
+        The configuration
+    """
+    with CONFIG_LOCK:
+        return _load_config()
+
 
+def check_config_mode():
     conf = os.path.expanduser("~/.anemoi.toml")
-    with open(conf, "w") as f:
-        tomllib.dump(CONFIG, f)
+    mode = os.stat(conf).st_mode
+    if mode & 0o777 != 0o600:
+        raise SystemError(f"Configuration file {conf} is not secure. " "Please run `chmod 600 ~/.anemoi.toml`.")

anemoi/utils/s3.py

@@ -4,54 +4,460 @@
 # 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
+
+"""
+
+import concurrent
 import logging
 import os
-from contextlib import closing
+import threading
+from copy import deepcopy
 
-import boto3
 import tqdm
 
-LOG = logging.getLogger(__name__)
+from .config import check_config_mode
+from .config import load_config
+from .humanize import bytes
 
+LOGGER = logging.getLogger(__name__)
 
-def upload(source, target, overwrite=False, ignore_existing=False):
-    # https://boto3.amazonaws.com/v1/documentation/api/latest/guide/s3-uploading-files.html
-    assert target.startswith("s3://")
 
-    _, _, bucket, key = target.split("/", 3)
+# s3_clients are not thread-safe, so we need to create a new client for each thread
 
-    LOG.info(f"Uploading {source} to {target}")
-    s3_client = boto3.client("s3")
+thread_local = threading.local()
 
-    if not overwrite:
-        results = s3_client.list_objects(Bucket=bucket, Prefix=key)
-        if results.get("Contents"):
-            if ignore_existing:
-                LOG.info(f"{target} already exists, skipping")
-                return
-            else:
-                raise ValueError(f"{target} already exists, use --overwrite to replace")
 
-    size = os.path.getsize(source)
-    with closing(tqdm.tqdm(total=size, unit="B", unit_scale=True)) as t:
-        s3_client.upload_file(source, bucket, key, Callback=lambda x: t.update(x))
+def s3_client(bucket):
+    import boto3
+
+    config = load_config()
+    if "object-storage" in config:
+        check_config_mode()
+
+    if not hasattr(thread_local, "s3_clients"):
+        thread_local.s3_clients = {}
+
+    if bucket not in thread_local.s3_clients:
+
+        options = {}
+        options.update(config.get("object-storage", {}))
+        options.update(config.get("object-storage", {}).get(bucket, {}))
+
+        type = options.pop("type", "s3")
+        if type != "s3":
+            raise ValueError(f"Unsupported object storage type {type}")
+
+        if "config" in options:
+            from botocore.client import Config
+
+            options["config"] = Config(**options["config"])
+            del options["config"]
+
+        thread_local.s3_clients[bucket] = boto3.client("s3", **options)
+
+    return thread_local.s3_clients[bucket]
+
+
+class Transfer:
+
+    def transfer_folder(self, *, source, target, overwrite=False, resume=False, verbosity=1, threads=1):
+        assert verbosity == 1, verbosity
+
+        # 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 = 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 += self.source_size(name)
+
+                    if len(futures) % 10000 == 0:
+                        if verbosity > 0:
+                            LOGGER.info(f"Preparing transfer, {len(futures):,} files... ({bytes(total)})")
+                        done, _ = concurrent.futures.wait(
+                            futures,
+                            timeout=0.001,
+                            return_when=concurrent.futures.FIRST_EXCEPTION,
+                        )
+                        # Trigger exceptions if any
+                        for future in done:
+                            future.result()
+
+                if verbosity > 0:
+                    LOGGER.info(f"{self.action} {len(futures):,} files ({bytes(total)})")
+                    with tqdm.tqdm(total=total, unit="B", unit_scale=True, unit_divisor=1024) as pbar:
+                        for future in futures:
+                            pbar.update(future.result())
+                else:
+                    for future in futures:
+                        future.result()
+
+            except Exception:
+                executor.shutdown(wait=False, cancel_futures=True)
+                raise
+
+
+class Upload(Transfer):
+    action = "Uploading"
+
+    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)
+        s3_path = os.path.join(target, relative_path)
+        return s3_path
+
+    def source_size(self, local_path):
+        return os.path.getsize(local_path)
+
+    def transfer_file(self, source, target, overwrite, resume, verbosity, 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(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 Download(Transfer):
+    action = "Downloading"
+
+    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, 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:
+            print(e.response["Error"]["Code"], e.response["Error"]["Message"], bucket, key)
+            if e.response["Error"]["Code"] == "404":
+                raise ValueError(f"{source} does not exist ({bucket}, {key})")
+            raise
 
-    LOG.info(f"{target} is ready")
+        size = int(response["ContentLength"])
 
+        if verbosity > 0:
+            LOGGER.info(f"Downloading {source} to {target} ({bytes(size)})")
 
-def download(source, target, overwrite=False):
+        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={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 upload(source, target, *, overwrite=False, resume=False, verbosity=1, threads=1):
+    """Upload a file or a folder to S3.
+
+    Parameters
+    ----------
+    source : str
+        A path to a file or a folder to upload.
+    target : str
+        A URL to a file or a folder on S3. The url should start with 's3://'.
+    overwrite : bool, optional
+        If the data is alreay on S3 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, by default False
+    threads : int, optional
+        The number of threads to use when uploading a directory, by default 1
+    """
+
+    uploader = Upload()
+    if os.path.isdir(source):
+        uploader.transfer_folder(
+            source=source, target=target, overwrite=overwrite, resume=resume, verbosity=verbosity, threads=threads
+        )
+    else:
+        uploader.transfer_file(source=source, target=target, overwrite=overwrite, resume=resume, verbosity=verbosity)
+
+
+def download(source, target, *, overwrite=False, resume=False, verbosity=1, threads=1):
+    """Download a file or a folder from S3.
+
+    Parameters
+    ----------
+    source : 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.
+    target : str
+        The local path where the file or folder will be downloaded.
+    overwrite : bool, optional
+        If false, files which have already been download will be skipped, unless their size
+        does not match their size on S3 , by default False
+    resume : bool, optional
+        If the data is alreay on local it will not be downloaded, unless the remote file
+        has a different size, by default False
+    threads : int, optional
+        The number of threads to use when downloading a directory, by default 1
+    """
     assert source.startswith("s3://")
 
-    _, _, bucket, key = source.split("/", 3)
+    downloader = Download()
+
+    if source.endswith("/"):
+        downloader.transfer_folder(
+            source=source,
+            target=target,
+            overwrite=overwrite,
+            resume=resume,
+            verbosity=verbosity,
+            threads=threads,
+        )
+    else:
+        downloader.transfer_file(
+            source=source,
+            target=target,
+            overwrite=overwrite,
+            resume=resume,
+            verbosity=verbosity,
+        )
+
+
+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):
+    _, _, 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):
+    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
 
-    s3 = boto3.client("s3")
-    response = s3.head_object(Bucket=bucket, Key=key)
-    size = response["ContentLength"]
+    LOGGER.info(f"Deleting {target}")
+    print(s3.delete_object(Bucket=bucket, Key=key))
+    LOGGER.info(f"{target} is deleted")
 
-    if not overwrite:
-        if os.path.exists(source) and os.path.getsize(source) == size:
-            LOG.info(f"{source} already exists, skipping")
-            return
 
-    with closing(tqdm.tqdm(total=size, unit="B", unit_scale=True)) as t:
-        s3.download_file(bucket, key, target, Callback=lambda x: t.update(x))
+def delete(target):
+    """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):
+    """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):
+    """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):
+    """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)

{anemoi_utils-0.3.3.dist-info → anemoi_utils-0.3.5.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: anemoi-utils
-Version: 0.3.3
+Version: 0.3.5
 Summary: A package to hold various functions to support training of ML models on ECMWF data.
 Author-email: "European Centre for Medium-Range Weather Forecasts (ECMWF)" <software.support@ecmwf.int>
 License: Apache License

{anemoi_utils-0.3.3.dist-info → anemoi_utils-0.3.5.dist-info}/RECORD

@@ -1,24 +1,24 @@
 anemoi/utils/__init__.py,sha256=zZZpbKIoGWwdCOuo6YSruLR7C0GzvzI1Wzhyqaa0K7M,456
 anemoi/utils/__main__.py,sha256=cLA2PidDTOUHaDGzd0_E5iioKYNe-PSTv567Y2fuwQk,723
-anemoi/utils/_version.py,sha256=FKnJIExgNrZG2xJ0y_dGNBpxGbGBYylvfat-jHhLUuM,411
+anemoi/utils/_version.py,sha256=3eLsZPTwWh0zKBhR6n3c4iAL9geCCdJGMojIL6dF0IA,411
 anemoi/utils/caching.py,sha256=HrC9aFHlcCTaM2Z5u0ivGIXz7eFu35UQQhUuwwuG2pk,1743
 anemoi/utils/checkpoints.py,sha256=1_3mg4B-ykTVfIvIUEv7IxGyREx_ZcilVbB3U-V6O6I,5165
-anemoi/utils/cli.py,sha256=d3TT9WSm6TDkbaJ9fL74wWG91Y27a9Uh17fPM4SomTs,3300
-anemoi/utils/config.py,sha256=XEesqODvkuE3ZA7dnEnZ-ooBRtU6ecPmkfP65FtialA,2147
+anemoi/utils/cli.py,sha256=w6YVYfJV-50Zm9FrO0KNrrIWDdgj5hPjxJvgAh391NY,3308
+anemoi/utils/config.py,sha256=WD3EZ2Ylt-HLvJnOPXKEGMepR2VVNYbJKC10fPJu45w,2333
 anemoi/utils/dates.py,sha256=Ot9OTY1uFvHxW1EU4DPv3oUqmzvkXTwKuwhlfVlY788,8426
 anemoi/utils/grib.py,sha256=gVfo4KYQv31iRyoqRDwk5tiqZDUgOIvhag_kO0qjYD0,3067
 anemoi/utils/humanize.py,sha256=LD6dGnqChxA5j3tMhSybsAGRQzi33d_qS9pUoUHubkc,10330
 anemoi/utils/provenance.py,sha256=v54L9jF1JgYcclOhg3iojRl1v3ajbiWz_oc289xTgO4,9574
-anemoi/utils/s3.py,sha256=HG3YulSwBvutTOZe0vY9oDkXExtUCcr26U9WObzl2I4,2024
+anemoi/utils/s3.py,sha256=OEZGm85gzWfZ29OuCx-A2dczC00XKM_SYnkGv9vgs14,15204
 anemoi/utils/text.py,sha256=4Zlc4r9dzRjkKL9xqp2vuQsoJY15bJ3y_Xv3YW_XsmU,8510
 anemoi/utils/timer.py,sha256=JKOgFkpJxmVRn57DEBolmTGwr25P-ePTWASBd8CLeqM,972
 anemoi/utils/commands/__init__.py,sha256=qAybFZPBBQs0dyx7dZ3X5JsLpE90pwrqt1vSV7cqEIw,706
 anemoi/utils/commands/checkpoint.py,sha256=SEnAizU3WklqMXUjmIh4eNrgBVwmheKG9gEBS90zwYU,1741
 anemoi/utils/mars/__init__.py,sha256=RAeY8gJ7ZvsPlcIvrQ4fy9xVHs3SphTAPw_XJDtNIKo,1750
 anemoi/utils/mars/mars.yaml,sha256=R0dujp75lLA4wCWhPeOQnzJ45WZAYLT8gpx509cBFlc,66
-anemoi_utils-0.3.3.dist-info/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
-anemoi_utils-0.3.3.dist-info/METADATA,sha256=3MtgyTflfoV8qhyRDDJfrcDNPz92sPtcULTFUJB4iB8,15513
-anemoi_utils-0.3.3.dist-info/WHEEL,sha256=GJ7t_kWBFywbagK5eo9IoUwLW6oyOeTKmQ-9iHFVNxQ,92
-anemoi_utils-0.3.3.dist-info/entry_points.txt,sha256=LENOkn88xzFQo-V59AKoA_F_cfYQTJYtrNTtf37YgHY,60
-anemoi_utils-0.3.3.dist-info/top_level.txt,sha256=DYn8VPs-fNwr7fNH9XIBqeXIwiYYd2E2k5-dUFFqUz0,7
-anemoi_utils-0.3.3.dist-info/RECORD,,
+anemoi_utils-0.3.5.dist-info/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
+anemoi_utils-0.3.5.dist-info/METADATA,sha256=N9sPN7e-Zb6y0sxhw6N124iHNOoWWh7Gc2RWjIAD0UY,15513
+anemoi_utils-0.3.5.dist-info/WHEEL,sha256=cpQTJ5IWu9CdaPViMhC9YzF8gZuS5-vlfoFihTBC86A,91
+anemoi_utils-0.3.5.dist-info/entry_points.txt,sha256=LENOkn88xzFQo-V59AKoA_F_cfYQTJYtrNTtf37YgHY,60
+anemoi_utils-0.3.5.dist-info/top_level.txt,sha256=DYn8VPs-fNwr7fNH9XIBqeXIwiYYd2E2k5-dUFFqUz0,7
+anemoi_utils-0.3.5.dist-info/RECORD,,

{anemoi_utils-0.3.3.dist-info → anemoi_utils-0.3.5.dist-info}/WHEEL

@@ -1,5 +1,5 @@
 Wheel-Version: 1.0
-Generator: bdist_wheel (0.43.0)
+Generator: setuptools (70.1.0)
 Root-Is-Purelib: true
 Tag: py3-none-any