This page explains how to parse files when you prepare data in the Wrangler
workspace of the Cloud Data Fusion Studio. Wrangler lets you parse a file before
loading it into the Wrangler workspace:

- Wrangler infers data types and maps each column to the inferred data type in the same way file source plugins do in the Pipeline Studio.
- When schema inference isn't possible, you can import the schema for a file format, such as JSON.
- The recipe doesn't include the parse directive, which reduces transformation logic during pipeline runs.
- When you create a pipeline from Wrangler, the source plugin includes all the same parsing properties and values that you set in Wrangler.

## Create a file connection

To parse a file before loading it into Wrangler, you must use a file connection,
such as File, Cloud Storage, or Amazon S3.

1. [Go to the Wrangler workspace in Cloud Data Fusion](https://docs.cloud.google.com/data-fusion/docs/concepts/wrangler-overview#navigate-to-wrangler).
2. Click the **Select data** expander arrow to view the available connections.
3. Add a connection for File, Cloud Storage, or S3. For more information, see [Create and manage connections](https://docs.cloud.google.com/data-fusion/docs/how-to/managing-connections).
4. To open the parsing options dialog, go to the **Select data** panel and click the name of the file.
5. In the **Parsing options** dialog, enter the following information:

   1. In the **Format** field, choose the file format of the data being
      read---for example, **csv** . For more information, see [Supported
      formats](https://docs.cloud.google.com/data-fusion/docs/how-to/wrangler-parse-files#supported-formats).

      - If you choose the delimiter format, in the **Delimiter** field that appears, enter the delimiter information.
      - If you choose CSV, TSV, or delimiter format, an **Enable quoted
        values** field appears. If your data is wrapped in quotation marks, select **True** . This setting trims quotation marks from the parsed output. For example, the following input, `1, "a, b, c"`, parses into two fields. The first field has the value: `1`. The second field has the value: `a, b, c`. The newline delimiter cannot be within quotes.
      - If you chose text, CSV, TSV, or delimiter format, a **Use first row
        as header** field appears. To use the first line of each file as a column header, select **True**.
   2. In the **File encoding** field, choose the file encoding type of the
      source file---for example, **UTF-8**.

   3. Optional: to import the schema or override the inferred schema for the
      file, click **Import Schema**. You import the schema for formats, such
      as JSON and some Avro files, where schema inference isn't possible. The
      schema must be in the Avro format.

   4. Click **Confirm**. The parsed file appears in the Wrangler workspace.

## Supported formats

The following formats are supported for file parsing:

- Avro
- Blob (the blob format requires a schema that contains a field named `body` of type `bytes`)
- CSV
- Delimited
- JSON
- Parquet
- Text (the text format requires a schema that contains a field named `body` of type `string`)
- TSV

## What's next

- Learn more about [Wrangler directives](https://docs.cloud.google.com/data-fusion/docs/concepts/wrangler-overview#apply_directives).