Presizely is an easy-to-use online service for transforming images. The main focus is on resizing/scaling images, but several other types of transformations are also possible.
The service is currently under beta testing and should not be used in production environments.
$ git clone [email protected]:toreau/Presizely.git
$ cd Presizely
$ docker build -t presizely .
$ docker run -p 3000:3000 presizely
The simplest way of using Presizely is to just hand it the URL to the original image without specifying any transformations. The only advantage you get from this, is that every subsequent request will return a cached version of the image.
http://127.0.0.1:3000/http://example.com/image.jpg
One number to the left of an x sets a specific width and calculates the height based on aspect ratio.
http://127.0.0.1:3000/1024x/http://example.com/image.jpg
One number to the right of an x sets a specific height and calculates the width based on aspect ratio.
http://127.0.0.1:3000/x1024/http://example.com/image.jpg
Two numbers separated by an x resizes the image to the exact dimensions specified.
http://127.0.0.1:3000/1024x1024/http://example.com/image.jpg
If any of the numbers explained above is a decimal number, it will be treated as a ratio. Here the image will be resized to 50 % of its width, with aspect ratio preserved;
http://127.0.0.1:3000/0.5x/http://example.com/image.jpg
One number to the right of an r rotates the image a specific number of degrees clockwise. Use a negative number to rotate anticlockwise.
http://127.0.0.1:3000/r90/http://example.com/image.jpg
NOTE! The background color for corners is at the moment set to white. It should be possible to choose the color, or transparency, and this will be rectified in a later version.
A set of numbers to the right of a c crops the image;
http://127.0.0.1:3000/c1024x1024-0x0/http://example.com/image.jpg
The two first numbers are the width and the height of the "crop window", while the two last numbers are the x and y value for where you want the "crop window" to be positioned.
Use the bw parameter to make the image black and white.
http://127.0.0.1:3000/bw/http://example.com/image.jpg
Use the fh and/or fv parameter to flip the image horizontally and/or vertically, respectively.
http://127.0.0.1:3000/fh,fv/http://example.com/image.jpg
One number to the right of a q changes the quality of the image.
http://127.0.0.1:3000/q75/http://example.com/image.jpg
Use the o parameter to optimize the Huffman coding tables. Compared to the other actions, this will have a bigger performance hit, depending on the image size.
http://127.0.0.1:3000/o/http://example.com/image.jpg
Use the prog parameter to convert JPEG images to progressive images. You can read more about progressive JPEGs here and here.
It's possible to chain image actions by separating them with a comma. Please note that cropping is always performed first, resizing is always performed second, while the other actions are performed in an arbitrary order.
http://127.0.0.1:3000/1024x,q75,r90/http://example.com/image.jpg
If Presizely is given a URL to an HTML page, like a news article, it will look for the image representing the page, specifically the og:image and twitter:image meta tags and use the image URL found (if any).
http://127.0.0.1:3000/1024x,q75,r90/http://example.com/sports/article.html
Coming soon...
Presizely have two cache levels. First, the original image is cached. Second, the processed image (ie. after all the actions have been performed on the image) is cached based on the actions done to it.
The two cache levels behaves the same, except that the most frequently accessed images (after transformation) are also stored in an in-memory L1 cache.
- Original image can't be larger than 64MB, but this can be changed by setting the
MOJO_MAX_MESSAGE_SIZEconfiguration variable to the max. number of bytes you want. - Only the most popular (...) image formats are supported; JPEG, PNG and GIF are safe bets.
See also Presizely's issues page on github for more information.
Presizely's performance depends primarily on the hardware (specifically the CPU) it runs on when it downloads and transforms an image. Once an image is downloaded and transformed, it's cached, so performance shouldn't be a problem. However, if you cache your images on disk, IO can be an issue. Presizely therefore does all the caching in the background, so that the client won't have to wait for the transformed image to be delivered.
On my server, an Intel Xeon E3-1246 3.50GHz, with Presizely set up to run with 16 workers, it can easily handle 2,500 requests per second after the image has been cached. Your mileage may wary, of course.
- Perl
- Mojolicious
- Imager
- ...and many other excellent CPAN modules!
Tore Aursand © 2016, ALL RIGHTS RESERVED
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent
