Skip to content

DOWNLOAD_FILTERING.md

Latest commit

 

History

History
120 lines (83 loc) · 4.24 KB

DOWNLOAD_FILTERING.md

File metadata and controls

120 lines (83 loc) · 4.24 KB

Dataset Download Filtering

← Back to README

Overview

This document describes how the download UI filters datasets based on user permissions. We only show datasets where the user can download all files, providing a cleaner user experience.

How It Works

Dataset Search Flow

  1. Get user's datasets via /api/mydata/retrieve (includes own drafts, unpublished)
  2. Search public datasets via /api/search
  3. Filter by downloadability - only show datasets where user can download all files

Why Filter?

When a dataset contains restricted or embargoed files that a user cannot access, showing it in the download list would lead to failed transfers. By filtering these datasets out, we:

  • Provide a cleaner UI showing only actionable items
  • Avoid confusing error messages during transfer attempts
  • Give users a clear picture of what they can actually download

Permission Check Logic

CanUserDownloadAllFiles Function

This function determines if a user can download all files in a dataset:

func CanUserDownloadAllFiles(ctx context.Context, persistentId, token, user string, 
    hasRestricted, hasEmbargoed bool) (bool, error)

Returns true if:

  1. Dataset has no restrictions - no restricted or embargoed files, OR
  2. User has EditDataset permission - dataset owners/curators can access all files, OR
  3. User has CanViewUnpublishedDataset permission - preview URL / private URL reviewers have full download access

Dataverse API Used

GET /api/v1/datasets/:persistentId/userPermissions?persistentId={pid}

Returns:

{
    "status": "OK",
    "data": {
        "canViewUnpublishedDataset": true,
        "canEditDataset": true,
        "canPublishDataset": true,
        "canManageDatasetPermissions": true,
        "canDeleteDatasetDraft": true
    }
}

The canEditDataset permission indicates the user is an owner or curator with full access to all files.

Performance Optimization

The permission check is only made when necessary:

Scenario API Calls
Public dataset (no restrictions) 0 extra calls
Restricted dataset 1 call to userPermissions

This ensures minimal overhead for most datasets while still providing accurate filtering.

User Experience

Scenario Dataset Shown in Download List?
Public dataset ✅ Yes
Restricted dataset, user is owner ✅ Yes
Restricted dataset, user is NOT owner ❌ No (cannot download all files)
Embargoed dataset, user is owner ✅ Yes
Embargoed dataset, user is NOT owner ❌ No (cannot download all files)
Any restricted/embargoed dataset, preview URL user ✅ Yes

Users who cannot download all files in a dataset will simply not see that dataset in the download interface, avoiding any confusion or failed transfer attempts.

Code References

Go Files

  • image/app/common/downloadable.go - Main download handler with filtering
  • image/app/dataverse/dataverse_read.go - CanUserDownloadAllFiles and GetDatasetUserPermissions
  • image/app/dataverse/embargo.go - GetDatasetNodesWithAccessInfo checks for restrictions

Key Functions

  • CanUserDownloadAllFiles() - Determines if user can download all files
  • GetDatasetUserPermissions() - Fetches user's permissions on a dataset
  • GetDatasetNodesWithAccessInfo() - Returns file tree with restriction flags

Additional Notes

  1. Users with granted access to specific files but not EditDataset will still be filtered out

    • These users can download via direct HTTP instead of Globus transfer
    • This provides a consistent user experience where Globus shows only fully downloadable datasets

  2. Preview URL users have full access

    • Users arriving via a Dataverse private URL (CanViewUnpublishedDataset) can download all files, including restricted/embargoed ones
    • This matches Dataverse's built-in private URL behaviour for reviewers

  3. Fine-grained access not currently supported

    • We use EditDataset (and CanViewUnpublishedDataset) as proxies for "can access all files"
    • A more precise check would require per-file permission checks (expensive)
    • For most use cases, dataset owners are the primary users of bulk Globus downloads