在Boto3中，如果您正在检查文件夹(前缀)或使用list_objects的文件。您可以使用响应字典中的“Contents”是否存在来检查对象是否存在。这是另一种避免try/except捕获的方法，就像@EvilPuppetMaster建议的那样

import boto3
client = boto3.client('s3')
results = client.list_objects(Bucket='my-bucket', Prefix='dootdoot.jpg')
return 'Contents' in results

FWIW，这里是我正在使用的非常简单的函数

import boto3

def get_resource(config: dict={}):
    """Loads the s3 resource.

    Expects AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY to be in the environment
    or in a config dictionary.
    Looks in the environment first."""

    s3 = boto3.resource('s3',
                        aws_access_key_id=os.environ.get(
                            "AWS_ACCESS_KEY_ID", config.get("AWS_ACCESS_KEY_ID")),
                        aws_secret_access_key=os.environ.get("AWS_SECRET_ACCESS_KEY", config.get("AWS_SECRET_ACCESS_KEY")))
    return s3


def get_bucket(s3, s3_uri: str):
    """Get the bucket from the resource.
    A thin wrapper, use with caution.

    Example usage:

    >> bucket = get_bucket(get_resource(), s3_uri_prod)"""
    return s3.Bucket(s3_uri)


def isfile_s3(bucket, key: str) -> bool:
    """Returns T/F whether the file exists."""
    objs = list(bucket.objects.filter(Prefix=key))
    return len(objs) == 1 and objs[0].key == key


def isdir_s3(bucket, key: str) -> bool:
    """Returns T/F whether the directory exists."""
    objs = list(bucket.objects.filter(Prefix=key))
    return len(objs) > 1

您可以使用S3Fs，它本质上是boto3的包装器，它公开了典型的文件系统风格操作:

import s3fs
s3 = s3fs.S3FileSystem()
s3.exists('myfile.txt')

我不太喜欢在控制流中使用异常。这是在boto3中工作的另一种方法:

import boto3

s3 = boto3.resource('s3')
bucket = s3.Bucket('my-bucket')
key = 'dootdoot.jpg'
objs = list(bucket.objects.filter(Prefix=key))
if any([w.key == path_s3 for w in objs]):
    print("Exists!")
else:
    print("Doesn't exist")

我发现的最简单的方法(可能也是最有效的)是:

import boto3
from botocore.errorfactory import ClientError

s3 = boto3.client('s3')
try:
    s3.head_object(Bucket='bucket_name', Key='file_path')
except ClientError:
    # Not found
    pass

我注意到，为了使用botocore.exceptions. clienterror捕获异常，我们需要安装botocore。botocore占用36M的磁盘空间。如果我们使用aws lambda函数，这尤其会产生影响。如果我们只是使用异常，那么我们可以跳过使用额外的库!

我正在验证文件扩展名为'.csv' 如果桶不存在，这将不会抛出异常! 如果桶存在但对象不存在，则不会抛出异常! 如果桶为空，则抛出异常! 如果桶没有权限，就会抛出异常!

代码看起来像这样。请分享你的想法:

import boto3
import traceback

def download4mS3(s3bucket, s3Path, localPath):
    s3 = boto3.resource('s3')

    print('Looking for the csv data file ending with .csv in bucket: ' + s3bucket + ' path: ' + s3Path)
    if s3Path.endswith('.csv') and s3Path != '':
        try:
            s3.Bucket(s3bucket).download_file(s3Path, localPath)
        except Exception as e:
            print(e)
            print(traceback.format_exc())
            if e.response['Error']['Code'] == "404":
                print("Downloading the file from: [", s3Path, "] failed")
                exit(12)
            else:
                raise
        print("Downloading the file from: [", s3Path, "] succeeded")
    else:
        print("csv file not found in in : [", s3Path, "]")
        exit(12)