srerickson / chaparral Goto Github PK

With ko, multi-platform images are easier to do and faster.

figure out how to get file size information into the object cache.

• clean up commit service tests
• need more tests for main package's run.Run()

Currently, download links for files won't work well in web browsers because an authorization header (bearer token) needs to be included with the download request. To make it easier use chaparral with browser-based front-ends, I need to add a way to generate signed urls for restricted files.

Something like this: GetSigneDownloadURL(resource, expire) returns a URL with a (single-use?) token as a query parameter.

This commit request fails:

{
  "storageRootId": "working",
  "objectId": "demo-object-03",
  "message": "this is a test",
  "state": {
    "file.zip": "21dd1dff33fcd1ece989869bb4677c2be7407b0e4c36d7daddf89608c43e2135"
  },
  "user": {
    "name": "Seth Erickson"
  },
  "digestAlgorithm": "sha256",
  "contentSources": [
    {"object":{
      "storageRootId": "restricted",
      "objectId": "45545/1937/07/02"}}
  ]
}

This is the erorr response:

{
  "code": "invalid_argument",
  "message": "commit error with possible object corruption: transferring new object contents: during copy: blob (key \"restricted/534/39a/9d0/45545%2f1937%2f07%2f02/v1/content/xml.zip -> working/76f/a6d/ac9/demo-object-03/v1/content/nytimes.zip\") (code=NotFound): NoSuchKey: The specified key does not exist.\n\tstatus code: 404, request id: FRN11A4363RMW577, host id: ZfrWeQhrnIEPhhIRPK8YYv0quJ/9uvP2kmpYzOlsW/tm0WHQaRitpYsMLh0AUsjOlUovEMAfSIWAb7/MPtvjrw=="
}

The bug most likely stems from the fact that object id used as a content source includes '/'. For comparison, this commit request works as expected:

{
  "storageRootId": "working",
  "objectId": "demo-object-04",
  "message": "this is a test",
  "state": {
    "file.zip": "53a6da718ab3da9364769d3622a2cc7dc48b2f61b6a5db3f9ed8b77969f90bfb"
  },
  "user": {
    "name": "Seth Erickson"
  },
  "digestAlgorithm": "sha256",
  "contentSources": [
    {"object":{
      "storageRootId": "restricted",
      "objectId": "index"}}
  ]
}

Delete requests should be able to clean up partial/incomplete objects. It should check that there is an object declaration file, but not require additional validation.

The storage group abstraction was meant to allow collections of storage roots with the same backend to be managed together, along with the uploader for the backend. Instead it's become another identifier that has to get added to every request -- the storage root identifier on its own should be sufficient. Storage roots and uploaders can share the same backend behind the scenes, but this doesn't need to be something the API user should be concerned with.

The spec field has the wrong value (spec: "https://ocfl.io/1.1/spec/#inventory")

This should initialize a storage root and start the server:

docker run --rm  -v $(pwd):/data -p 8080:8080 srerickson/chaparral 

The server container image should be configured with functional defaults, such as:

• Backend set to /data
• initialize an OCFL storage root in /data
• listen on :8080

Minimally, add an API endpoint for paging through all the object ids in a storage root. Also consider one that triggers a reindex of a storage root.

the CHAPARRAL_BACKEND environment variable set in the Docker container is preventing downstream images from setting the backend using a config file. (fig privileges environment variables of config files values)

Add the ability to configure roles and access permissions.