Skip to main content

A tile-on-demand tile server built with PIL and Tornado

Project description


A tile-on-demand tile server built with PIL and Tornado.


We want to store a high-resolution image in memory on the server as a numpy
array. Then when a client requests a particular tile, we can make the PNG of the
requested tile by slicing the numpy array and using PIL to write the resulting
PNG back to the client through a StringIO stream.

This is primarily intended for building interactive visualizations in research
settings where we might want to skip the time- and/or disk-intensive tile
generation step required by typical tile servers.


Install projectile

$ pip install projectile

Serve a test image from the [USC-SIPI Image Database](

$ projectile sanfran

Manually request a particular tile by navigating to <http://localhost:8000/2/1/2.png>.

Try zooming and panning in the demo client by navigating to <http://localhost:8000/>.

Serve one of your own images (any format readable by PIL) with

$ projectile some_image.tiff

or, if you have data in a numpy `.npy` file,

$ projectile some_image.npy

Load an image in grayscale mode and apply a colormap:

$ projectile --mode L --cmap viridis pentagon

Stress testing

To test the performance limits of on-demand tiling, download this
[high resolution map of Great Britain](
(8,150 × 13,086 pixels, file size: 102.74 MB) from Wikimedia Commons.

Grayscale performance test:

$ projectile britain.jpg -m L

Reducing tile resolution when running in color:

$ projectile britain.jpg --tile_size 128


- `numpy>=1.13.3`
- `Pillow>=4.3.0`
- `tornado>=4.5.2`
- `matplotlib>=2.1.0`


### URL scheme

The server will serve grayscale and RGB images in their original colors at


where `<z>` is the zoom level, `<x>` and `<y>` specify the coordinates of the
tile at that zoom level (`0/0` is the top left tile), and `<s>` specifies the
image tile resolution in pixels (must be a power of 2).

The server will serve colormapped versions of a grayscale image at


where `<cmap>` is the name of a matplotlib colormap, and `<vmin>` and `<vmax>`
specify the range of image pixel values linearly interpolate against the
colormap (pixel values outside this range will be clipped).

### Using a custom client

If you like the projectile backend but just want to use a simple custom client
contained in a single HTML file `custom_client.html`, you can run

$ projectile array.npy --client custom_client.html

to make projectile serve your client instead of the included demo client.

### Using projectile in your existing Tornado web application

The core functionality is exposed in the `TileHandler` class defined in
[](projectile/, which you can use in your own Tornado web

from tornado import web
from projectile.server import TileHandler


app = web.Application([
(r'/([0-9]+)/([0-9]+)/([0-9]+)/([0-9]+).png', TileHandler,


### Launching projectile from your own Python code

You can also launch a server from your own Python code with the `run()` function
defined in [](projectile/

from projectile.server import run



The demo client is lifted from <>, with the
addition of a small filtering check to prevent the client from requesting tiles
which lie beyond the image boundaries.

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for projectile, version 0.0.4
Filename, size File type Python version Upload date Hashes
Filename, size projectile-0.0.4-py2-none-any.whl (12.0 kB) File type Wheel Python version py2 Upload date Hashes View
Filename, size projectile-0.0.4.tar.gz (23.8 kB) File type Source Python version None Upload date Hashes View

Supported by

AWS AWS Cloud computing Datadog Datadog Monitoring DigiCert DigiCert EV certificate Facebook / Instagram Facebook / Instagram PSF Sponsor Fastly Fastly CDN Google Google Object Storage and Download Analytics Pingdom Pingdom Monitoring Salesforce Salesforce PSF Sponsor Sentry Sentry Error logging StatusPage StatusPage Status page