AWS S3, "simple storage service", is the classic AWS service. It was the first to launch, the first one I ever used and, seemingly, lies at the very heart of almost everything AWS does.

Given that S3 is essentially a filesystem, a logical thing is to be able to count the files in an S3 bucket. Illustrated below are three ways.

Method 1: aws s3 ls

S3 is fundamentally a filesystem and you can just call ls on it. Yep – ls in the cloud. blink

aws s3 ls s3://adl-ohi/ --recursive --summarize | grep "Total Objects:"
Total Objects: 444803

Method 2: aws s3api

And since S3 is a modern filesystem, it actually has an API that you can call. Yep – a json api. blink blink

aws s3api list-objects --bucket adl-ohi --output json --query "[length(Contents[])]"
[
    448444
]

Method 3: A Python Example

Naturally you can just run code to do all this. I started with an example from the Stack Overflow link below that was written for boto and upgraded it to boto3 (as still a Python novice, I feel pretty good about doing this successfully; I remember when Ruby went thru the same AWS v2 to v3 transition and it sucked there too). I also learned how to dynamically introspect methods from Python objects as part of this debugging cycle.

#!/usr/local/bin/python

import sys
import boto3

s3 = boto3.resource('s3')
s3bucket = s3.Bucket(sys.argv[1])
size = 0
totalCount = 0

for key in s3bucket.objects.all():
    totalCount += 1
    size += key.size

print('total size:')
print("%.3f GB" % (size*1.0/1024/1024/1024))
print('total count:')
print(totalCount)

which gives output like this:

python3 scratch/count_s3.py adl-ohi
total size:
0.298 GB
total count:
486468

Note: I have a live upload happening on another machine so the numbers do change and that's actually fine.

References