Download a folder from S3 using Boto3

Question

Using Boto3 Python SDK, I was able to download files using the method bucket.download_file()

Is there a way to download an entire folder?

Answer 1

quick and dirty but it works:

import boto3
import os 

def downloadDirectoryFroms3(bucketName, remoteDirectoryName):
    s3_resource = boto3.resource('s3')
    bucket = s3_resource.Bucket(bucketName) 
    for obj in bucket.objects.filter(Prefix = remoteDirectoryName):
        if not os.path.exists(os.path.dirname(obj.key)):
            os.makedirs(os.path.dirname(obj.key))
        bucket.download_file(obj.key, obj.key) # save to same path

Assuming you want to download the directory foo/bar from s3 then the for-loop will iterate all the files whose path starts with the Prefix=foo/bar.

Answer 2

A slightly less dirty modification of the accepted answer by Konstantinos Katsantonis:

import boto3
s3 = boto3.resource('s3') # assumes credentials & configuration are handled outside python in .aws directory or environment variables

def download_s3_folder(bucket_name, s3_folder, local_dir=None):
    """
    Download the contents of a folder directory
    Args:
        bucket_name: the name of the s3 bucket
        s3_folder: the folder path in the s3 bucket
        local_dir: a relative or absolute directory path in the local file system
    """
    bucket = s3.Bucket(bucket_name)
    for obj in bucket.objects.filter(Prefix=s3_folder):
        target = obj.key if local_dir is None \
            else os.path.join(local_dir, os.path.relpath(obj.key, s3_folder))
        if not os.path.exists(os.path.dirname(target)):
            os.makedirs(os.path.dirname(target))
        if obj.key[-1] == '/':
            continue
        bucket.download_file(obj.key, target)

This downloads nested subdirectories, too. I was able to download a directory with over 3000 files in it. You'll find other solutions at Boto3 to download all files from a S3 Bucket, but I don't know if they're any better.

Answer 3

You could also use cloudpathlib which, for S3, wraps boto3. For your use case, it's pretty simple:

from cloudpathlib import CloudPath

cp = CloudPath("s3://bucket/folder/folder2/")
cp.download_to("local_folder")

Answer 4

Using boto3 you can set aws credentials and download dataset from S3

import boto3
import os 

# set aws credentials 
s3r = boto3.resource('s3', aws_access_key_id='xxxxxxxxxxxxxxxxx',
    aws_secret_access_key='xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx')
bucket = s3r.Bucket('bucket_name')

# downloading folder 
prefix = 'dirname'
for object in bucket.objects.filter(Prefix = 'dirname'):
    if object.key == prefix:
        os.makedirs(os.path.dirname(object.key), exist_ok=True)
        continue;
    bucket.download_file(object.key, object.key)

If you cannot find ur access_key and secret_access_key, refer to this page
I hope it will helps.
thank you.

Answer 5

Another approach building on the answer from @bjc that leverages the built in Path library and parses the s3 uri for you:

import boto3
from pathlib import Path
from urllib.parse import urlparse

def download_s3_folder(s3_uri, local_dir=None):
    """
    Download the contents of a folder directory
    Args:
        s3_uri: the s3 uri to the top level of the files you wish to download
        local_dir: a relative or absolute directory path in the local file system
    """
    s3 = boto3.resource("s3")
    bucket = s3.Bucket(urlparse(s3_uri).hostname)
    s3_path = urlparse(s3_uri).path.lstrip('/')
    if local_dir is not None:
        local_dir = Path(local_dir)
    for obj in bucket.objects.filter(Prefix=s3_path):
        target = obj.key if local_dir is None else local_dir / Path(obj.key).relative_to(s3_path)
        target.parent.mkdir(parents=True, exist_ok=True)
        if obj.key[-1] == '/':
            continue
        bucket.download_file(obj.key, str(target))

Answer 6

You can call the awscli cp command from python to download an entire folder

 import os
 import subprocess

 remote_folder_name = 's3://my-bucket/my-dir'
 local_path = '.'
 if not os.path.exists(local_path):
     os.makedirs(local_path)
 subprocess.run(['aws', 's3', 'cp', remote_folder_name, local_path, '--recursive'])

Some notes regarding this solution:

1. You should install awscli (pip install awscli) and configure it. more info here
2. If you don't want to override existing files if they weren't changed, you can use sync instead of cp subprocess.run(['aws', 's3', 'sync', remote_folder_name, local_path])
3. Tested on python 3.6. on earlier versions of python you might need to replace subprocess.run with subprocess.call or os.system
4. The cli command that is executed by this code is aws s3 cp s3://my-bucket/my-dir . --recursive

Answer 7

The above solutions are good, and rely on S3 Resource.
The following solution, achieves the same goal, but with applying s3_client.
You might find it useful for your end (I've tested it, and it works well).

import boto3
from os import path, makedirs
from botocore.exceptions import ClientError
from boto3.exceptions import S3TransferFailedError

def download_s3_folder(s3_folder, local_dir, aws_access_key_id, aws_secret_access_key, aws_bucket, debug_en):
    """ Download the contents of a folder directory into a local area """

    success = True

    print('[INFO] Downloading %s from bucket %s...' % (s3_folder, aws_bucket))

    def get_all_s3_objects(s3, **base_kwargs):
        continuation_token = None
        while True:
            list_kwargs = dict(MaxKeys=1000, **base_kwargs)
            if continuation_token:
                list_kwargs['ContinuationToken'] = continuation_token
            response = s3.list_objects_v2(**list_kwargs)
            yield from response.get('Contents', [])
            if not response.get('IsTruncated'):
                break
            continuation_token = response.get('NextContinuationToken')

    s3_client = boto3.client('s3',
                             aws_access_key_id=aws_access_key_id,
                             aws_secret_access_key=aws_secret_access_key)

    all_s3_objects_gen = get_all_s3_objects(s3_client, Bucket=aws_bucket)

    for obj in all_s3_objects_gen:
        source = obj['Key']
        if source.startswith(s3_folder):
            destination = path.join(local_dir, source)
            if not path.exists(path.dirname(destination)):
                makedirs(path.dirname(destination))
            try:
                s3_client.download_file(aws_bucket, source, destination)
            except (ClientError, S3TransferFailedError) as e:
                print('[ERROR] Could not download file "%s": %s' % (source, e))
                success = False
            if debug_en:
                print('[DEBUG] Downloading: %s --> %s' % (source, destination))

    return success

Answer 8

I had some problems with this version. Modified variable destination and included variable for filter to files type.

from sre_constants import SUCCESS
import boto3 
from os import path, makedirs 
from botocore.exceptions import ClientError
from boto3.exceptions import S3TransferFailedError

def download_s3_folder(s3_folder, local_dir, aws_access_key_id, aws_secret_access_key, aws_bucket, debug_en, datatype):
""""" Download the contents of a folder directory into a local area """""

success = True
# Start do processo de copia
print('[INFO] Downloading %s from bucket %s...' % (s3_folder, aws_bucket))

# Metodo que lista todos os objetos do Bucket. 
def get_all_s3_objects(s3, **base_kwargs):
    continuation_token = None
    while True:
        list_kwargs = dict(MaxKeys=1000, **base_kwargs)
        if continuation_token:
            list_kwargs['ContinuationToken'] = continuation_token
        response = s3.list_objects_v2(**list_kwargs)
        yield from response.get('Contents', [])
        if not response.get('IsTruncated'):
            break
        continuation_token = response.get('NextContinuationToken')

s3_client = boto3.client('s3',
                         aws_access_key_id=aws_access_key_id,
                         aws_secret_access_key=aws_secret_access_key)

all_s3_objects_gen = get_all_s3_objects(s3_client, Bucket=aws_bucket)

# Loop into os objetos do S3,
for obj in all_s3_objects_gen:
    source = obj['Key']
    if source.startswith(s3_folder):
        # Transform path to using fo SO
        destination = path.join(local_dir,source).replace('/','\\')
        
        if not path.exists(path.dirname(destination)):
            makedirs(path.dirname(destination))
        try:
            #print('copy')
            if destination.endswith(datatype):
                #print(destination)
                print('Copia do arquivo "%s" Sucesso' % (destination))
                s3_client.download_file(aws_bucket, source, destination)
        except (ClientError, S3TransferFailedError) as e:
            print('[ERROR] Could not download file "%s": %s' % (source, e))
            success = False
        if debug_en:
            print(f"[DEBUG] Downloading: {source} --> {destination}")

return success

Answer 9

Here is my approach inspired by konstantinos-katsantonis and bjc answers.

import os
import boto3
from pathlib import Path

def download_s3_dir(bucketName, remote_dir, local_dir):
    assert remote_dir.endswith('/')
    assert local_dir.endswith('/')
    s3_resource = boto3.resource('s3')
    bucket = s3_resource.Bucket(bucketName) 
    objs = bucket.objects.filter(Prefix=remote_dir)
    sorted_objs = sorted(objs, key=attrgetter("key"))
    for obj in sorted_objs:
        path = Path(os.path.dirname(local_dir + obj.key))
        path.mkdir(parents=True, exist_ok=True)
        if not obj.key.endswith("/"):
            bucket.download_file(obj.key, str(path) + "/" + os.path.split(obj.key)[1])