Question

I’m wondering how to do this. I looked at the sdk documentation and have some examples, but am confused how the syntax generally goes.

If I want to delete a file, I assume I use deleteObject(path, key). However, what is the “key”?

Also how do you delete a directory? I can’t seem to find a method for doing that.

Answer 1

A “key” in S3 is similar to a file path:

http://bucket.s3.amazonaws.com/some/path/to/use

… is in a bucket named bucket and has a key of some/path/to/use.

It’s not actually a path though, because there are no folders. The S3 key is just the file name for a file in one big directory (the entire bucket). S3 keys can contain /, but it has no special meaning unless you set the delimiter argument with listing a bucket.

In other words, having an object named some/object doesn’t tell you anything about the object some (it might or might not exist — the two objects are not related).

However, you can request keys with a specific prefix, so I could say “give me all keys starting with some/path/to/ and it will return some/path/to/use. It looks like “listing a directory”, but it’s really just asking for files that start with a specific string of characters.

I could just as easily name things like this:

somepathtousea
somepathtouseb

And say “give me everything starting with somepathtouse” (and it would say somepathtousea and somepathtouseb).

Note: S3 URL’s come in several forms:

http://s3.amazonaws.com/bucket/key
http://bucket.s3.amazonaws.com/key
http://bucket/key (where bucket is a DNS CNAME record pointing to bucket.s3.amazonaws.com)

EDIT:

I looked at the JavaDocs and this is the function signature I see (for AmazonS3Client):

public void deleteObject(java.lang.String bucketName,
                         java.lang.String key)
                  throws AmazonClientException,
                         AmazonServiceException

EDIT again:

Folders do kind-of exist now, as zero-length objects with a content-type of application/x-directory and a key ending in /:

$ AWS_PROFILE=prod aws s3api head-object --bucket example-bucket --key example-directory/
{
    "AcceptRanges": "bytes",
    "LastModified": "Mon, 29 Apr 2019 14:59:36 GMT",
    "ContentLength": 0,
    "ETag": "\"d41d8cd98f00b204e9800998ecf8427e\"",
    "ContentType": "application/x-directory",
    "ServerSideEncryption": "AES256",
    "Metadata": {}
}

This is still just convention and there’s nothing stopping you from having files ending / or files inside of “folders” that don’t exist.