w3c / FileAPI

File API
https://w3c.github.io/FileAPI/
Other
105 stars 44 forks source link

Allow `createObjectURL` to retain filename from `File` (and similar) objects #203

Open jsoref opened 5 days ago

jsoref commented 5 days ago

What problem are you trying to solve?

We have a datastore that requires authentication to retrieve files (e.g. PDF, CSV, ZIP, PNG, or some Office suite file format) -- it has a file name, a content type, and data. We'd like to allow users to click a link (or button/...) and have the "file" corresponding to that data (which we currently retrieve in a Blob) open in their browser (or download depending on how the browser is feeling) and if at some point the browser decides to save the file (e.g. because the user asks to save a file, or because it doesn't have an internal handler) then we'd like it to favor our proposed filename instead of a blob id.

What solutions exist today?

  1. We can construct a Blob or a File w/ a Blob and call createObjectURL with it.
  2. We can create an anchor a with a#download with the href set the the url from createObjectURL such that users can click the anchor to get our suggested filename.
  3. We can add an iframe containing our blob url

Fwiw, Microsoft had an msSaveBlob which included a defaultName field...

How would you solve it?

Change the specification for createObjectURL so that for objects that have a name (e.g. File), browsers should prefer that name over the blob when saving.

Kaiido commented 5 days ago

Might be noted that Firefox is already able to derive this name in the case of an empty <a download> attribute, so the blob URL entry as currently specified might be enough to derive that name.

I'm not quite sure where it's specified where to get the file name for downloading a resource opened as a "document" though, and this Firefox fails it too.

annevk commented 1 day ago

I guess ideally Fetch would generate the appropriate headers so HTML doesn't have to know it's downloading a Blob object, but perhaps that should only happen for the download code path as people might rely on the headers a Blob object has today.

But yeah, this should just work.