DashIngest — Databricks Library

Part of the Dashlibs suite — Databricks libraries built for business users.

ADF-style data ingestion: pick a source kind, fill a few plain fields — no hand-written abfss:// URIs or JDBC connection strings — and run.

Installation

%pip install dash-ingest

Quick Start

import dashingest
dashingest.launch()   # Opens interactive UI in your Databricks notebook

Or drive it directly from code:

from dashingest import ADLSSource, IngestTarget, run_ingestion

source = ADLSSource(storage_account="myacct", container="raw", path="sales/2024.csv")
target = IngestTarget(table="main.bronze.sales", write_mode="merge", merge_keys=["order_id"])
result = run_ingestion(source, target)
result.display()

Sources

File format (csv/json/parquet/excel/avro/orc/text) is inferred from the path's extension if not set explicitly — most ingestions need zero format options.

File format readers

Each format has its own options dataclass with real per-format defaults — not a generic options dict. Excel gets the most coverage, since vanilla Spark has no native Excel reader and a raw file path alone doesn't tell it which sheet to read, where the header starts, or whether the workbook is password-protected:

from dashingest import ExcelReaderOptions, VolumeSource

source = VolumeSource(
    catalog="main", schema_name="bronze", volume="landing",
    path="regional_sales.xlsx",
    reader_options=ExcelReaderOptions(
        sheet_name="Q1 Actuals",
        header_row=2,              # skips two title/banner rows above the header
        workbook_password="secret",  # optional
    ),
)

Set sheet_names=["Jan", "Feb", "Mar"] instead of sheet_name to read and stack several same-shaped sheets into one DataFrame — the common "one tab per month" spreadsheet layout.

CsvReaderOptions (delimiter, quote/escape chars, encoding, null markers, date/timestamp formats, parse mode), JsonReaderOptions, ParquetReaderOptions/OrcReaderOptions (schema merging), and TextReaderOptions are also available — pass any of them via reader_options= on a source.

Write modes

append · overwrite · merge (upsert into Delta by merge_keys, with schema evolution where the runtime supports it).

Test Connection & Preview

Both the UI and the API let you check a source before committing to a full run — the same pattern ADF's linked-service "Test Connection" and dataset "Data preview" use:

from dashingest import test_connection, preview

test_connection(source).display()   # reachability/credentials check, no data read
preview(source, limit=10)            # pandas DataFrame of the first N rows

test_connection runs a lightweight check per source kind: SELECT 1 for databases, an HTTP request for REST APIs, a filesystem existence check for Volumes/ADLS/S3/DBFS (no dbutils needed — it uses Spark's Hadoop filesystem API directly, so it works the same way across all of them).

Advanced database & REST options

DatabaseSource supports SSL, JDBC fetch size, parallel reads (split a large table by partition_column across num_partitions), and a raw connection_properties escape hatch:

from dashingest import DatabaseSource

source = DatabaseSource(
    engine="postgresql", host="db.internal", database="analytics",
    table="events", user="svc", password="...",
    ssl=True, num_partitions=8, partition_column="id",
    lower_bound=0, upper_bound=10_000_000,
)

RestApiSource supports auth (bearer / api_key / basic) and pagination (page_param or cursor-based, up to max_pages):

from dashingest import RestApiSource

source = RestApiSource(
    url="https://api.example.com/records",
    auth_type="bearer", bearer_token="...",
    pagination="cursor", cursor_json_path="meta.next_cursor", max_pages=50,
)

Part of Dashlibs

License

Apache 2.0