Read a Microsoft Fabric Warehouse table — fabric_warehouse_read

Provides the table-oriented read counterpart to fabric_warehouse_write_table(). It resolves the Warehouse like the writer, safely quotes the schema, table, and projected columns, and delegates query execution and type conversion to fabric_sql_query(). Use that lower-level function for filters, ordering, joins, aggregations, or other T-SQL.

Usage

fabric_warehouse_read_table(
  warehouse,
  table,
  workspace = NULL,
  schema = "dbo",
  columns = NULL,
  limit = NULL,
  result = c("tibble", "arrow_stream"),
  backend = c("odbc", "adbc"),
  numeric_policy = c("auto", "exact", "driver"),
  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(),
  api_base = .fabric_api_base,
  verbose = TRUE,
  timeout = 30L,
  max_tries = 3L,
  retry_delay = 5,
  sql_token = NULL
)

Arguments

warehouse

A Warehouse object returned by fabric_warehouses() or fabric_item(), or its name or GUID when workspace is supplied.

table

Warehouse table name, or a record containing a name, table, or displayName field.

workspace

Workspace name, GUID, or discovery object containing warehouse. May be omitted when warehouse is a discovery object.

schema

Warehouse schema. Defaults to "dbo"; a table record can supply its schema when this argument is omitted.

columns

Optional unique column names to project.

limit

Optional non-negative maximum number of rows to return.

result

Return a "tibble" for ordinary R analysis, or a single-use "arrow_stream". ADBC streams retain native Arrow types. ODBC streams are converted from R data frames and cannot recover values lost by the driver. The 'adbi' driver may fetch the complete result before returning the stream, so this option does not guarantee bounded-memory retrieval. An Arrow stream owns its DBI result and connection until the stream is released; consume it promptly or release it explicitly with nanoarrow::nanoarrow_pointer_release()

backend

SQL connection backend, "odbc" or "adbc".

numeric_policy

"auto" (default) uses "driver" for ODBC and "exact" for ADBC. Automatic ODBC conversion warns once per R session about possible numeric precision loss. Set "driver" explicitly to accept the driver's conversions without this warning, or use "exact" to reject unsafe ODBC results before fetching

"exact" preserves ADBC decimals as character and BIGINT as bit64::integer64, using character for columns containing the minimum BIGINT. INT columns containing -2147483648 use exact doubles. Nested lists retain character decimals and 64-bit integers, and double 32-bit integers. Null struct parents require result = "arrow_stream"; exact tibble collection raises fabric_arrow_null_struct_error to preserve their distinction from valid structs with all-null fields. ODBC rejects DECIMAL, NUMERIC, INT and BIGINT columns before fetching: its conversion can round or truncate values or turn valid integer boundaries into missing values. Cast these columns to varchar in SQL or use ADBC. "driver" explicitly accepts the backend's conversions, including possible rounding and missing values, for either output format. This policy applies to this query helper; direct DBI calls on fabric_sql_connect() use the selected driver's conversion settings

tenant_id

Microsoft Entra tenant ID. Defaults to FABRICQUERYR_TENANT_ID

client_id

Microsoft Entra application/client 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()

api_base

Fabric REST API base used when a Warehouse name or GUID must be discovered.

verbose

Whether to report SQL connection progress.

timeout

Non-negative whole-number login/connect timeout in seconds; 0 lets the driver use an unlimited or driver-specific timeout

max_tries

Maximum attempts after temporary Fabric SQL failures

retry_delay

Initial delay in seconds before retrying. Later retries wait progressively longer, up to 60 seconds

sql_token

Optional separate Azure SQL token or token-provider function. Supply it when token is fixed rather than audience-aware.

Value

A tibble, or a single-use nanoarrow_array_stream when result = "arrow_stream".

Large results

Use backend = "adbc" with result = "arrow_stream" for a native Arrow result path that avoids conversion to an R data frame. The current result path through 'DBI' and 'adbi' may fetch the complete result before returning the stream, so use a selective query or limit when the result may exceed memory. The external ADBC mssql driver must be installed.

limit uses T-SQL TOP and does not define row order. Use fabric_sql_query() with an explicit ORDER BY when deterministic row selection matters.

References

Query a Fabric Warehouse

Fabric Warehouse connectivity

Examples

if (FALSE) { # \dontrun{
# Discover the Warehouse instead of copying its SQL connection details
workspace <- fabric_workspaces()[[1L]]
warehouse <- fabric_warehouses(workspace)[[1L]]

# Use 'DBI' metadata to discover an existing table in that Warehouse
con <- fabric_sql_connect(warehouse)
tables <- DBI::dbListTables(con)
DBI::dbDisconnect(con)
table <- tables[[1L]]

# Read a bounded selection into a tibble
orders <- fabric_warehouse_read_table(
  warehouse,
  table,
  backend = "adbc",
  limit = 1000
)

# Keep the result Arrow-native rather than converting it to a data frame
stream <- fabric_warehouse_read_table(
  warehouse,
  table,
  backend = "adbc",
  result = "arrow_stream"
)
reader <- arrow::as_record_batch_reader(stream)
} # }