Work with files in Microsoft Fabric OneLake — fabric_onelake

List, inspect, download, upload, and delete ordinary files stored in OneLake These helpers are intended for files such as CSV, JSON, images, and model artifacts in a Fabric item's Files/ area

fabric_onelake_list() lists paths and follows all continuation tokens
fabric_onelake_metadata() returns file or directory properties
fabric_onelake_download() reads a file into memory or streams it to disk
fabric_onelake_upload() creates or replaces a file
fabric_onelake_delete() explicitly deletes a file or directory

Usage

fabric_onelake_list(
  workspace,
  item = NULL,
  path = "",
  recursive = FALSE,
  page_size = 5000L,
  begin_from = NULL,
  item_type = NULL,
  tenant_id = Sys.getenv("FABRICQUERYR_TENANT_ID"),
  client_id = Sys.getenv("FABRICQUERYR_CLIENT_ID", unset =
    "04b07795-8ddb-461a-bbee-02f9e1bf7b46"),
  token = NULL,
  auth_args = list(),
  dfs_base = "https://onelake.dfs.fabric.microsoft.com"
)

fabric_onelake_metadata(
  workspace,
  item = NULL,
  path = "",
  item_type = NULL,
  tenant_id = Sys.getenv("FABRICQUERYR_TENANT_ID"),
  client_id = Sys.getenv("FABRICQUERYR_CLIENT_ID", unset =
    "04b07795-8ddb-461a-bbee-02f9e1bf7b46"),
  token = NULL,
  auth_args = list(),
  dfs_base = "https://onelake.dfs.fabric.microsoft.com"
)

fabric_onelake_download(
  workspace,
  item = NULL,
  path = "",
  dest = NULL,
  range = NULL,
  overwrite = FALSE,
  if_match = NULL,
  item_type = NULL,
  tenant_id = Sys.getenv("FABRICQUERYR_TENANT_ID"),
  client_id = Sys.getenv("FABRICQUERYR_CLIENT_ID", unset =
    "04b07795-8ddb-461a-bbee-02f9e1bf7b46"),
  token = NULL,
  auth_args = list(),
  dfs_base = "https://onelake.dfs.fabric.microsoft.com"
)

fabric_onelake_upload(
  workspace,
  item = NULL,
  path = "",
  source,
  overwrite = FALSE,
  if_match = NULL,
  content_type = NULL,
  create_parents = TRUE,
  item_type = NULL,
  tenant_id = Sys.getenv("FABRICQUERYR_TENANT_ID"),
  client_id = Sys.getenv("FABRICQUERYR_CLIENT_ID", unset =
    "04b07795-8ddb-461a-bbee-02f9e1bf7b46"),
  token = NULL,
  auth_args = list(),
  dfs_base = "https://onelake.dfs.fabric.microsoft.com",
  allow_managed_tables = FALSE,
  chunk_size = getOption("fabricqueryr.onelake.chunk_size", 8 * 1024^2)
)

fabric_onelake_delete(
  workspace,
  item = NULL,
  path = "",
  recursive = FALSE,
  confirm = FALSE,
  if_match = NULL,
  item_type = NULL,
  tenant_id = Sys.getenv("FABRICQUERYR_TENANT_ID"),
  client_id = Sys.getenv("FABRICQUERYR_CLIENT_ID", unset =
    "04b07795-8ddb-461a-bbee-02f9e1bf7b46"),
  token = NULL,
  auth_args = list(),
  dfs_base = "https://onelake.dfs.fabric.microsoft.com",
  allow_managed_tables = FALSE
)

Arguments

workspace: Workspace name, ID, object from fabric_workspaces(), or a complete OneLake HTTPS/ABFSS path When item contains a workspace ID, a supplied workspace name must match its recorded workspaceDisplayName. If that name is unavailable, supply the workspace ID or a discovered workspace object instead.
item: Item name, GUID, or discovered Fabric item. Use NULL when workspace is a complete OneLake path. An item from fabric_lakehouses() is the least ambiguous input
path: Path relative to the item, usually beginning with Files/ or Tables/, for example "Files/incoming/data.csv". Use forward slashes A complete OneLake path already contains this value
recursive: For listing, whether to include all descendants. For deletion, whether a non-empty directory may be removed
page_size: Maximum paths requested from OneLake per API call, from 1 to 5000. Smaller values reduce each response size but require more requests
begin_from: Optional path at which to begin a listing. Use this to resume a long, alphabetically ordered scan. Non-recursive listings accept only a single path level
item_type: Optional Fabric item type appended to an item name unless that name already ends in the same suffix, for example "Lakehouse" Usually unnecessary for a discovered item or a name such as "Sales.Lakehouse"
tenant_id: Entra tenant ID. Defaults to FABRICQUERYR_TENANT_ID
client_id: Entra application ID. Defaults to FABRICQUERYR_CLIENT_ID, then the Azure CLI application ID
token: Optional access token or token-provider function. Leave NULL to let 'fabricQueryR' use its normal sign-in flow
auth_args: Additional sign-in options passed to AzureAuth::get_azure_token()
dfs_base: OneLake service address. Most users should keep the default; a workspace-specific address discovered from Fabric is used when available
dest: Optional local destination. When NULL, download returns a raw vector held in R memory. Supply a path to stream large files to disk. A destination download is staged before it replaces an existing file
range: Optional inclusive zero-based byte range. Supply one value for all bytes from that offset onward, or two values for start through end Leave NULL to download the entire file. A ranged request must receive a matching HTTP 206 Content-Range response whose body has the claimed size
overwrite: Whether an existing local or OneLake file may be replaced Existing files are protected by default
if_match: Optional file version (etag) returned by fabric_onelake_metadata(). The operation proceeds only if the file still has that version
source: Local file path or raw vector to upload. A path is streamed; a raw vector is already held in memory
content_type: Optional MIME type stored with an uploaded file, for example "text/csv"
create_parents: Logical. Create missing parent directories below the Fabric-managed first-level folder. Keep TRUE for normal uploads
allow_managed_tables: Whether to allow direct changes below Tables/ Keep FALSE for normal use: changing Delta files directly can corrupt a managed table. This guard checks only the supplied path and does not resolve shortcuts. A path below Files/ can reach a managed table through a shortcut; writes or deletion of descendants then change the shortcut target's data.
chunk_size: Upload chunk size in bytes. The default suits most files; larger values make fewer requests but use more memory
confirm: Safety switch that must be explicitly set to TRUE before deletion is attempted

Value

fabric_onelake_list() returns one row per path, including its item-relative path, file name, is_directory, content_length, etag, and modification/permission fields fabric_onelake_metadata() and fabric_onelake_upload() return a one-row tibble with the resolved path and available HTTP metadata fabric_onelake_download() returns a raw vector when dest = NULL, or invisibly returns the destination path after writing to disk fabric_onelake_delete() invisibly returns TRUE

Choosing a target

The easiest inputs are a workspace plus an item returned by fabric_lakehouses(). You can also use names, IDs, or a complete OneLake HTTPS/ABFSS path. When using an item name, include its type suffix, such as "Sales.Lakehouse", or supply item_type

A Lakehouse's Tables/ area is managed as Delta tables. Use fabric_onelake_read_delta_table() to read those tables, and use SQL, Spark, or another Delta-aware tool to change them. Uploading or deleting individual files below Tables/ can damage a table and is blocked by default

Permissions

The signed-in user or application needs access through a workspace role or the item's OneLake security roles, configured under Manage OneLake security. Uploading and deleting need write permission. Your Fabric administrator must also allow external apps to access OneLake. If a call returns HTTP 403 after sign-in succeeds, check both that tenant setting and the item's data permissions

Listing integrity

Directory listing validates every JSON page and path record before returning data. Malformed envelopes, invalid metadata values, and paths outside the requested item directory raise fabric_onelake_protocol_error; they are not silently converted to empty or partial results

Storage API version

Requests use OneLake's currently documented ADLS API version, 2021-06-08. For controlled compatibility testing with a later service version, set option fabricqueryr.onelake.api_version to another date in YYYY-MM-DD form

Safe file replacement

Existing files are protected unless overwrite = TRUE. Uploads and downloads are staged before replacing their destination, so an interrupted transfer does not normally leave a partial file. Local downloads are published with an atomic same-directory rename or hard link and fail closed when the filesystem cannot provide the required primitive. Use if_match when a OneLake file should be replaced only if it has not changed since you inspected it

A response failure during an upload's final rename can leave the server-side outcome unknown. In that case a fabric_onelake_commit_ambiguous error reports absolute target and staging URLs plus their relative paths. Automatic cleanup is not attempted, but a committed rename may already have consumed the staging path, so the condition reports its presence as unknown

References

Connect to OneLake with ADLS APIs

ADLS Gen2 List Paths

Create and manage OneLake security roles

OneLake security best practices

OneLake tenant settings

Examples

if (FALSE) { # \dontrun{
# Discover the OneLake target instead of typing workspace and item names
workspace <- fabric_workspaces()[[1L]]
lakehouse <- fabric_lakehouses(workspace)[[1L]]

# Create a small local CSV and upload it to the discovered Lakehouse
local_csv <- tempfile(fileext = ".csv")
write.csv(data.frame(id = 1:3), local_csv, row.names = FALSE)
fabric_onelake_upload(
  workspace,
  lakehouse,
  "Files/incoming/example.csv",
  source = local_csv
)

# List the folder and inspect metadata for the uploaded file
files <- fabric_onelake_list(
  workspace = workspace,
  item = lakehouse,
  path = "Files/incoming",
  recursive = TRUE
)
metadata <- fabric_onelake_metadata(
  workspace,
  lakehouse,
  "Files/incoming/example.csv"
)

# Download the first 100 bytes when only a file sample is needed
bytes <- fabric_onelake_download(
  workspace,
  lakehouse,
  "Files/incoming/example.csv",
  range = c(0, 99)
)

# Deletion is explicit and requires confirm = TRUE
fabric_onelake_delete(
  workspace,
  lakehouse,
  "Files/incoming/example.csv",
  confirm = TRUE
)
} # }