to operate on arbitrary UploadedFile, instead of only currently attached (#get) one.

with inter...

The user can now pass :filename to `Shrine.data_uri`:

io = Shrine.data_uri("data:,content", f...

Currently `Shrine.data_uri(io)` is returning a DataFile object, where
the #content_type attribut...

Now it's possible to reuse extracting MIME type from the #content_type
attribute via

Shrine.m...

The #content_type attribute of the input IO object, which equals to the
"Content-Type" request h...

This naming makes much more sense, as it makes it explicit how the MIME
type gets extracted. Als...

The whole point of the :default determine_mime_type plugin analyzer is
to allow users to confirm...

This way all scanning is done together in one sequence, instead of
sprinkled over the method.

Both S3#download and S3#open, in addition to the #get_object request,
are implicitly doing a #he...

Require dependencies prior to initialising.

The change was made in https://github.com/shrinerb/shrine/pull/304.

This section is now copied almost word-for-word from
https://docs.aws.amazon.com/AmazonS3/latest...

Before the change, it fails with a misleading error message mentioning
a missing `name` option....

This conveys much better that we're using the endpoint of our app. It's
also shorter and doesn't...

Uppy 0.28 now passes cookies to presign requests to the Companion
endpoint, so now we can remove...

RDoc doesn't support indented code blocks.

Some HTTP libraries rely on this information to return the response body
in correct encoding. On...

Rack's multipart parser will return filename in binary encoding if it
has any non-ascii characte...

It appears that the file command can also fail with a successful exit
status, in which case the...

Rack 2.0.x requires MRI >= 2.2.0.

... to download endpoint plugin docs

Now it's possible to do this:

s3.url("foo", host: "https://your-s3-host.com/prefix")
#=> "h...

This change was introduced in Shrine 2.11.0, but it causes a regression
when using an endpoint t...

That way the behaviour doesn't change, you only get the additional
filename.

fetch() doesn't properly URI-encode some characters (such as "%"), so
when Rack receives the req...

We don't need the response we get when we set :success_action_status to
"201", we already have t...

No need to do it, it's just an example, showing setup is out of the
scope.

Some people have asked for the ability to skip temporary storage and
upload directly to permanen...

This makes it possible to do things like overriding metadata, specifying
upload location, or pas...

This commit adds the :metadata option to Shrine#upload for overriding
extracted metadata on uplo...

* explain direct uploads in the README better
* show example of direct S3 uploads
* show how to ...

Add more advice on various strategies for metadata extraction.

Markdown parsers on GitHub and in RDoc don't handle hard-wrapped link
specifiers, so we make sur...

changelog entry

The libvips project repositories have been moved from the "jcupitt" user
to the "libvips" organi...

When wanting to serve private content through CloudFront, the S3 object
URLs need to be signed s...

Making uploads public consists of two things: setting "public-read" ACL
on upload and generating...

This is not necessary when doing direct uploads, because in that case
files are not actually upl...

aws-sdk-s3 1.14.0 added Aws::S3::Object#upload_stream, which uses
parallelized multipart upload ...

This method does NOT, ever, "uses S3's multi delete capabilities."

Date.now() returns the current time in milliseconds, but this isn't
strong enough to be a unique...

Aws::S3::Object#upload_file uploads multipart parts concurrently, so we
don't know in which orde...

The "Aws S3" Uppy plugin recommends a different CORS configuration than
the one we copied from R...

We also update the CORS configuration so that it works both for regular
and multipart uploads to...

aws-sdk-s3 uses threads when doing multipart uploads, so the order of
parts can be different.

UploadedFile#open now always opens the file, meaning it will close an
already open file and reop...

Aws::S3::Object#upload_file does multipart upload in threads, so on
JRuby this test will fail if...