Writing a Shell Function

On This Page

This guide uses practical examples to guide you through the process of writing serverless shell functions.

Overview

The shell runtime allows function developers to fork a process on every received event. Developers can choose to provide an executable script or run any executable binary in the Docker image. This guide walks you through both scenarios.

Handle events with a bash script

This example guides you through the steps for deploying a shell script that reverses the event’s body. To implement this, you call rev and pass stdin as input; (the event body will appear to shell functions as stdin).

Create a /tmp/nuclio-shell-script/reverser.sh file with the following code:

#!/bin/sh

# @nuclio.configure
#
# function.yaml:
#   apiVersion: "nuclio.io/v1"
#   kind: "Function"
#   spec:
#     runtime: "shell"
#     handler: "reverser.sh"
#

rev /dev/stdin

The function configuration needs to include the following:

  1. runtime — set to shell.
  2. handler — set to the name of the executable file. In this example, the file is reverser.sh.

Run the following command to deploy the function with the nuctl Nuclio CLI:

Note

if you’re not running on top of Kubernetes, pass the --platform local option to nuctl.

nuctl deploy -p /tmp/nuclio-shell-script/reverser.sh rev

And now, use the nuctl CLI to invoke the function:

nuctl invoke rev -m POST -b reverse-me

> Response headers:
Date = Sun, 03 Dec 2017 12:53:51 GMT
Content-Type = text/plain; charset=utf-8
Content-Length = 10
Server = nuclio

> Response body:
em-esrever

Handle events with any executable binary

Because the shell runtime simply forks a process, you can leverage it to run any executable binary in the Docker image. This means that you don’t need to provide any code to the shell runtime, only a function configuration. In this example, you install the ImageMagick utility and call its convert executable on each event. You then send the function images and use convert to reduce the image by 50% in the response. You do this by invoking the nuctl CLI as follows:

nuctl deploy -p /dev/null convert \
    --runtime shell \
    --build-command "apk --update --no-cache add imagemagick" \
    --handler convert \
    --runtime-attrs '{"arguments": "- -resize 50% fd:1"}'

Following is an explanation of the options used in the command:

  • -p /dev/null — because you don’t need to pass a path, you just instruct nuctl to read from /dev/null.
  • --build-command "apk --update --no-cache add imagemagick" — instruct the builder to install ImageMagick on the build through apk.
  • --handler convert — the handler must be set to the name or path of the executable. In this example, convert is in the environment PATH so there’s no need for a full path.
  • --runtime-attrs '{"arguments": "- -resize 50% fd:1"}' — through runtime specific attributes, you specify the arguments for the executable. In this example, - instructs the runtime to read from stdin, and the rest of the arguments specify how to convert the received image.

Because invoke can’t (yet) send images, use HTTPie to create a thumbnail file; replace the <function ip:port> placeholder with your function-URL information:

http https://blog.golang.org/gopher/header.jpg | http <function ip:port> > thumb.jpg

Overriding the arguments per request

The shell runtime allows events to override the default arguments through the use of a header. This means that you can supply x-nuclio-arguments as a header and provide any convert arguments that you wish, per event. Thus, you can create, for example, a smaller thumbnail file by using the following invocation command; replace the <function ip:port> placeholder with your function-URL information:

http https://blog.golang.org/gopher/header.jpg | http <function ip:port> x-nuclio-arguments:"- -resize 20% fd:1" > thumb.jpg 

See also