---
title: STDIO Adapter | Juttle Language Reference
---

# STDIO Adapter

The `stdio` adapter allows reading points from stdin and writing points to stdout.

[TOC]

## read stdio

Supported formats are JSON array, JSON lines, column-aligned text, and CSV; see examples below.

```text
read stdio [-format <format>] [-timeField <fieldname>]
```

Parameter         |             Description          | Required?
----------------- | -------------------------------- | ---------:
`-format`         | Input file format, supports: `csv`, `json`, `jsonl`, or `grok` for text [parseable by grok](parsers/index.md) | No; defaults to `json`
`-timeField`      | Name of the field in the data which contains a valid timestamp  | No; defaults to `time`
`-pattern`        | When `-format 'grok'` you can specify the grok matching pattern here. More information on grok [here](parsers/grok.md)  | No
`-separator`  | When `-format 'csv'` is used, you can specify the separator between columns in a CSV file. | No: defaults to `,`
`-commentSymbol`    | When `-format 'csv'` is used, you can specify the comment character that prefixes comment lines in a CSV file. | No: defaults to `,`
`-ignoreEmptyLines`     | When `-format 'csv'` is used, you can skip empty lines in a CSV file. | No: defaults to `false`
`-allowIncompleteLines` | When `-format 'csv'` is used, you can allow for parsing of incomplete lines in a CSV file. | No: defaults to `false`

The data is assumed to contain valid timestamps in a field named `time` by default; a different name for the time field may be specified with `-timeField` option. If the data contains fields `time` and another named field specified with `-timeField`, the original contents of field `time` will be overwritten by the valid timestamp from `timeField`. 
Timeless data, which contains no timestamps, is acceptable; however certain operations which expect time to be present in the points, such as `reduce -every :interval:`, will execute with warnings or error out. Timeless data can be joined in the Juttle flowgraph with other data which contains timestamps; a common use case for reading timeless data from a file or another backend is to join it with streaming data for enrichment.

The stdio adapter does not support any kind of filtering (neither filter expressions, nor full text search). In order to filter the data from the `read stdio`, pipe to the [filter](../processors/filter.md) proc.

_Example: redirect a JSON file to your juttle program_

This is an example using the CLI since this is where the stdio adapter comes in handy:

```bash
cat myfile.json | juttle -e "read stdio | filter hostname = 'lemoncake' | view table"
```

_Example: count the processes holding the most open files_

This is an example using column-aligned text formatting that comes as the output of UNIX commands like lsof.

```bash
lsof | juttle -e "read stdio -format 'columns' | reduce count() by COMMAND | sort count -desc | head 10 | view table"
```

## write stdio

Supported formats are JSON array, JSON lines and CSV; see examples below.

```text
... | write stdio [-format <format>]
```

Parameter         |             Description          | Required?
----------------- | -------------------------------- | ---------:
`-format`         | Input input format: `csv` for [CSV](https://tools.ietf.org/html/rfc4180) data, `json` for [JSON](https://tools.ietf.org/html/rfc7159) data, or `jsonl` for [JSON lines](http://jsonlines.org/) data | No; defaults to `json`

_Example: writing csv data to stdout_

```juttle
emit -from :2015-01-01: -limit 2
| put name = 'write_me', value = count()
| write stdio -format 'csv'
```

The resulting output will look like so:

```csv
time,name,value
2015-01-01T00:00:00.000Z,write_me,1
2015-01-01T00:00:01.000Z,write_me,2
```