fitter

I'm planning of make fitter a pluggable service based on serverless architecture

The fitter is an on-demand image server for generating dynamic images. It provides some various actions and modes to generate the dynamic images with desired parameters. As on-demand means, fitter is an on-demand server to generate the various images for your mobile app and website whenever those need it. It can be used on following scenarios.

• Backend on-demand image server
• Local image transforming tool
• Image transforming and cloud backup tool

It can save the efforts of create the images before really needed it. So it can also save the storage space either.

(Current) Architecture

Features

• Provides various actions for generating a desired image such as resize, flip and rotate
• Provides various mode for getting an image (information)
• Supports caching for url path of generated image. It supports followings now
  • In-memory (use pure python dictionary)
  • Redis (as front cache server)
• Supports flexible backend storage modules
  • Able to specify the source storage and store (as destination) storage
  • Able to either use local file system or AWS S3 as backend storage
• Provides a customizable configuration for fitter server
• (Not yet) Able to upload an image for generating image (optional)

How to use

On your machine (local or server)

1. git clone [email protected]:mingrammer/fitter.git (or use https)

2. pip install -r requirements.txt

   Note: Install the required packages in an isolated enviroment if you use virtualenv

3. Run cp fitter.yaml.sample fitter.yaml and modify & set the fitter configuration with yours. (see the configuration)

4. python3 run.py

5. Enjoy with fitter API! (see the usage)

Use docker container

Preparing. I'll support it ASAP

Configuration

You must set the configuration fitter.yaml to use fitter first.

There could be several use cases fit to you. I'll introduce some of them here.

Without caching, and use only local file system as storages (for both source and store)

If you only specify the store field, the source storage will be same as this config

port: 6001

# You must specify the storage config
storage:
  # You should specify the storage type one of followings
  # 1. fs (refers to file system)
  # 2. s3 (refers to AWS S3)
  #
  # If you want to use a single storage for both source and store, you should set the 'store' only
  store:
    type: fs
    location: /users/mingrammer/desktop # get the image on from this location as root

With redis cache server, and use the local file system and AWS S3 as source and store storage, respectively

# Fitter server port, Default is 6001
port: 6001

# Optional
cache_store:
   # If you want to use it, should specify the cache server type one of followings
   # 1. redis
   # 2. in-memory
   type: redis
   host: localhost
   port: 6379
   db: 0

storage:
  source:
    type: fs
    location: /users/mingrammer/desktop
  store:
    type: s3
    aws_access_key_id: <you-aws-access-key-id>
    aws_secret_access_key: <your-aws-secret-access-key>
    bucket_name: my-storage
    bucket_region: ap-northeast-2
    location: /store/image/upload # cached image will be stored in location + '/cache'

Same as 2, but also use the options

port: 6001

cache_store:
   type: redis
   host: localhost
   port: 6379
   db: 0

storage:
  source:
    type: fs
    location: /users/mingrammer/desktop
  store:
    type: s3
    aws_access_key_id: <you-aws-access-key-id>
    aws_secret_access_key: <your-aws-secret-access-key>
    bucket_name: my-storage
    bucket_region: ap-northeast-2
    location: /store/image/upload

options:
  enable_upload: true # allows image uploading
  # TODO: Will supports followings
  # shard
  # prefix
  # allowed_size
  # allowed_format
  # thumbnail_size

Usage

API Usage

http://<host>:<port>/<action>?mode=<mode>&path=<path>&format=<img_format>&width=<width>&height=<height>&upscale=<upscale>&quality=<quality>&direction=<direction>&degree=<degree>

ex) Resize the image to 300x400 in jpg with get action

http://localhost:6001/get?mode=resize&path=/wallpaper/lion.png&format=jpg&width=300&height=400

ex) Make a thumbnail of 600 width while preserving aspect ratio with show action

http://localhost:6001/show?mode=thumbnail&path=/image/profile.png&width=600&height=0

ex) Rotate the image by 90 degree with redirect action

http://localhost:6001/redirect?mode=rotate&degree=90

API Description

• action

  • /show: Get the image information asynchronously. It is usefule on like <img> tag. It will returns following json

    Note: Can not use this asynchronously yet. I'll make it ASAP

    { 
    	'filename': 'image',
      	'path': 'path/to/image',
      	'url': 'http://<your-server-or-s3-server>/path/to/image'
    }
    

  • /get: Get the image information synchronously

    { 
    	'filename': 'image',
      	'path': 'path/to/image',
      	'url': 'http://<your-server-or-s3-server>/path/to/image'
    }
    

  • /redirect: Redirect to the generated image url

• GET parameters

  • mode: The operation mode. One of followings
    • thumbnail: make a thumbnail with desired width/height
    • resize: resize the image with desired width/height
    • flip: flip or flop the image with desired direction
    • rotate: rotate the image with desired degree
  • path: The path of image which is on source storage. Will be concatenated with location field of storage
  • format: The desired format of image. Current supported list are:
    • png
    • jpg
    • jpeg
  • width: The desired width of image. If this value is 0, the aspect ratio of output image is preserved
  • height The desired height of image. If this value is 0, the aspect ratio of output image is preserved
  • upscale: Whether if upscale the image has size smaller than desired size
  • quality: The desired quality of image
  • direction: The desired direction to flip the image
    • 'h' or 'horizontal'
    • 'v' or 'vertical'
  • degree: The desired degree to rotate the image

Todo

• DO NOT SUPPORT PYTHON 2.X
• Supports API authorization
• Supports asynchronous job for show mode
• Supports image uploading and processing with external image url
• Add some additional options such as fixed size, prefix and shard
• Provides a CLI for managing the fitter server: fitter run
• Supports more actions for dynamic image transforming
• Considering how use the fitter on serverless architecture (may use the AWS lambda mainly)
• Supports multi-clustering environment
• Test! Test! Test!
• More documentation
• Providing deploy process to use it as pluggable service easily