给定S3路径以及有效的密钥和机密，如何更新对象缓存控制标头？

Question

I need to update the headers on files after they are uploaded to S3. I don't have control over the upload process (I'm using the FilePicker.io API which doesn't provide a way to specify the cache-control header as far as I now), they just magically appear in a bucket. I have the full s3 path to the objects and the key and secret for the bucket.

Using Go, what is the easiest way to add new headers to these objects? Seems like you need to do a PUT copy request but that requires request signing and it overwrites all of the existing headers. All I want to do is add a cache-control header, there has to be an easier way right?

Answer 1

The small program below should simply add a cache-control header to the the given bucket / key combo. The important bit is the s3.CopyOptions struct. The MetadataDirective can also be COPY - see http://docs.aws.amazon.com/AmazonS3/latest/API/RESTObjectCOPY.html for details. Also the source must be bucket/key since the source of course can be in another bucket.

package main

import (
    "fmt"
    "os"

    "github.com/goamz/goamz/aws"
    "github.com/goamz/goamz/s3"
    //// should work as well
    //"github.com/crowdmob/goamz/aws"
    //"github.com/crowdmob/goamz/s3"
)

func main() {
    // use as
    //  $ go run s3meta.go bucket key
    // will add a 1 hour Cache-Control header to
    // key in bucket
    auth := aws.Auth{
        AccessKey: os.Getenv("AWS_ACCESS_KEY_ID"),
        SecretKey: os.Getenv("AWS_SECRET_KEY_ID"),
    }

    bucketName, keyName := os.Args[1], os.Args[2]

    bucket := s3.New(auth, aws.USEast).Bucket(bucketName)
    opts := s3.CopyOptions{}
    opts.CacheControl = "maxage=3600"
    opts.MetadataDirective = "REPLACE"

    _, err := bucket.PutCopy(keyName, s3.PublicRead, opts, bucketName+"/"+keyName)
    if err != nil {
        panic(err)
    }

}

Trial run (bucket has since been deleted):

╭─brs at stengaard in ~/ using
╰─○ curl  -I https://s3.amazonaws.com/cf-templates-1r14by1vl75o0-us-east-1/success.png
HTTP/1.1 200 OK
x-amz-id-2: 49oTuRARMHlx32nqv34CMOjdTMBUCZIVzP8YKBS2Wz5h1w5KBG62u8nFru1UkIbJ
x-amz-request-id: C92E9952BFF31D77
Date: Mon, 30 Jun 2014 08:57:15 GMT
Last-Modified: Mon, 30 Jun 2014 08:50:45 GMT
ETag: "41b9951893f1bbff89e2b9c8a5b7ea18"
Accept-Ranges: bytes
Content-Type: image/png
Content-Length: 61585
Server: AmazonS3

╭─brs at stengaard in ~/ using
╰─○ go run s3meta.go cf-templates-1r14by1vl75o0-us-east-1 success.png
╭─brs at stengaard in ~/ using
╰─○ curl  -I https://s3.amazonaws.com/cf-templates-1r14by1vl75o0-us-east-1/success.png
HTTP/1.1 200 OK
x-amz-id-2: oiDeXjO1V4kquWo8UlNWBi/HAHoqfvlOSHVeXFZXv2yA4o0+Njcdshhu15PIiw7J
x-amz-request-id: 0BB1A397DE7EBE75
Date: Mon, 30 Jun 2014 09:00:17 GMT
Cache-Control: maxage=3600
Last-Modified: Mon, 30 Jun 2014 09:00:12 GMT
ETag: "41b9951893f1bbff89e2b9c8a5b7ea18"
Accept-Ranges: bytes
Content-Type: binary/octet-stream
Content-Length: 61585
Server: AmazonS3

Note that Content-Type changes as well since we have opts.MetadataDirective = "REPLACE". If this little thing is worth the hassle of updating headers out-of-band is really domain specific: How important is it to cache the uploaded files in the client? Is it to expensive to do the HEAD request to S3?