a web map tile caching system
- (c) 2006-2007 MetaCarta, Inc. Distributed under the BSD license.
- Manual section:
- Manual group:
TileCache is a BSD licensed tile caching mechanism. The goal is to make it easy to set up a WMS or TMS frontend to any backend data services you might be interested in, using a pluggable caching and rendering mechanism.
TileCache was developed by MetaCarta Labs and released to the public under a BSD license.
The TileCache was designed as a companion to OpenLayers, the BSD licensed web mapping interface. If you are using TileCache with OpenLayers, please read the section of this readme which describes how to do so. For additional help with setting up TileCache for use with OpenLayers, please feel free to stop by #openlayers, on irc.freenode.net, or to send email to firstname.lastname@example.org.
Generally, installing TileCache is as simple as downloading a source distribution and unpacking it. For installation systemwide, you can also use the Python Package Index (aka pypi or Cheeseshop) to install TileCache. Simply type easy_install TileCache. Once this is done, you will need to install the TileCache configuration file. A tool to do this is installed, called tilecache_install_config.py. A full installation likely looks like:
$ sudo easy_install TileCache ... Installed /usr/lib/python2.5/site-packages/TileCache-2.0-py2.5.egg $ sudo tilecache_install_config.py Successfully copied file /usr/lib/python2.5/site-packages/TileCache-2.0-py2.5.egg/TileCache/tilecache.cfg to /etc/tilecache.cfg.
TileCache is also available as a Debian package from the TileCache homepage. This Debian package is designed to install on Debian etch releases or later. This Debian package should install on Ubuntu Feisty or Gutsy.
RUNNING UNDER CGI
Extract the code to some web directory (e.g. in /var/www).
Edit tilecache.cfg to point the DiskCache to the location you wish to cache tiles, and the layers to point to the map file or WMS server you wish to cache. On Debian, this file is in /etc/tilecache.cfg by default.
Permit CGI execution in the TileCache directory. For example, if TileCache is to be run with Apache, the following must be added in your Apache configuration, where /var/www/tilecache is the directory resulting from the code extraction. On Debian, this is typically /usr/lib/cgi-bin.
<Directory /var/www/tilecache> AddHandler cgi-script .cgi Options +ExecCGI </Directory>
http://example.com/yourdir/tilecache.cgi?LAYERS=basic&SERVICE=WMS &VERSION=1.1.1&REQUEST=GetMap&SRS=EPSG:4326&BBOX=-180,-90,0,90 &WIDTH=256&HEIGHT=256
If you see a tile you have set up your configuration correctly. Congrats!
Non-standard Python Location
If your Python is not at /usr/bin/python on your system, you will need to change the first line of tilecache.cgi to reference the location of your Python binary. A common example is:
Under Apache, you might see an error message like:
[Wed Mar 14 19:55:30 2007] [error] [client 127.0.0.1] (2)No such file or directory: exec of '/www/tilecache.cgi' failed
to indicate this problem.
You can typically locate where Python is installed on your system via the command which python.
Windows users: If you are using Windows, you should change the first line of tilecache.cgi to read:
C:/Python should match the location Python is installed under on your system. In Python 2.5, this location is C:/Python25 by default.
RUNNING UNDER MOD_PYTHON
Extract the code to some web directory (e.g. /var/www).
Edit tilecache.cfg to point the DiskCache to the location you wish to cache tiles, and the layers to point to the map file or WMS server you wish to cache
Add the following to your Apache configuration, under a <Directory> heading:
AddHandler python-program .py PythonHandler TileCache.Service PythonOption TileCacheConfig /path/to/tilecache.cfg
An example might look like:
<Directory /var/www/tilecache/> AddHandler python-program .py PythonHandler TileCache.Service PythonOption TileCacheConfig /var/www/tilecache/tilecache.cfg </Directory>
In this example, /var/www/tilecache is the directory resulting from the code extraction. If you’ve installed this from a Debian package, the location of your .cfg file is probably /etc/tilecache.cfg.
Edit tilecache.cfg to point to the location of your ‘Layers’ directory, as demonstrated inside the default tilecache.cfg.
Visit one of the URLs described above, replacing tilecache.cgi with tilecache.py
If you see a tile you have set up your configuration correctly. Congrats!
RUNNING STANDALONE (UNDER WSGI)
TileCache as of version 1.4 comes with a standalone HTTP server which uses the WSGI handler. This implementation depends on Python Paste, which can be downloaded from:
For versions of Python earlier than 2.5, you will also need to install wsgiref:
Once you have all the prerequisites installed, simply run:
This will start a webserver listening on port 8080, after which you should be able to open:
to see your first tile.
RUNNING UNDER FASTCGI
TileCache as of version 1.4 comes with a fastcgi implementation. In order to use this implementation, you will need to install flup, available from:
This implementation also depends on Python Paste, which can be downloaded from:
Once you have done this, you can configure your fastcgi server to use tilecache.fcgi.
Configuring FastCGI is beyond the scope of this documentation.
RUNNING UNDER IIS
Installing TileCache for use with IIS requires some additional configuration.
Install Python for Windows
Follow “Using Python Scripts with IIS” to setup Python CGI for IIS. http://support.microsoft.com/kb/276494
Edit metabase.xml to get correct Security Permissions for IIS 6.0. http://blogs.msdn.com/david.wang/archive/2005/04/20/IIS6-CGI-Web-Service-Extension.aspx
Edit tilecache.cgi to specify your configuration file explicitly:
svc = Service.load("C:\\TileCache\\tilecache.cfg")
When using Tilecache with TMS style requests, you will need to change your IIS configuration to allow PATH_INFO to be delivered to the CGI script. http://support.microsoft.com/kb/q184320/ provides information on this topic. If you do not do this, you will get an error message like:
“The requested layer (tilecache.cgi) does not exist.”
TileCache is configured by a config file, defaulting to tilecache.cfg. There are several parameters to control TileCache layers that are applicable to all layers:
The bounding box of the Layer. The resolutions array defaults to having resolutions which are equal to the bbox divided by 512 (two standard tiles).
Whether to send debug output to the error.log. Defaults to “yes”, can be set to “no”
Layer description, used in some metadata responses. Default is blank.
File extension of the layer. Used to request images from WMS servers, as well as when writing cache files.
A string used to describe the layers. Typically passed directly to the renderer. The WMSLayer sends this in the HTTP request, and the MapServerLayer chooses which layer to render based on this string. If no layer is provided, the layer name is used to fill this property.
An integer, describing the number of ‘zoom levels’ or scales to support. Overridden by resolutions, if passed.
The absolute file location of a mapfile. Required for MapServer and Mapnik layers.
The maximum resolution. If this is set, a resolutions array is automatically calculated up to a number of levels controlled by the ‘levels’ option.
set to “yes” to turn on metaTiling. This will request larger tiles, and split them up using the Python Imaging library. Defaults to “no”.
an integer number of pixels to request around the outside of the rendered tile. This is good to combat edge effects in various map renderers. Defaults to 10.
A comma seperated pair of integers, which is used to determine how many tiles should be rendered when using metaTiling. Default is 5,5.
Comma seperate list of resolutions you want the TileCache instance to support.
Comma seperated set of integers, describing the width/height of the tiles. Defaults to 256,256
String describing the SRS value. Default is “EPSG:4326”
The type of layer. Options are: WMSLayer, MapnikLayer, MapServerLayer, ImageLayer
URL to use when requesting images from a remote WMS server. Required for WMSLayer.
The watermarkImage parameter is assigned on a per-layer basis. This is a fully qualified path to an image you would like to apply to each tile. We recommend you use a watermark image the same size as your tiles. If using the default tile size, you should use a 256x256 image. NOTE: Python Imaging Library DOES NOT support interlaced images.
The watermarkOpacity parameter is assigned on a per-layer basis. This configures the opacity of the watermark over the tile, it is a floating point number between 0 and 1. Usage is optional and will otherwise default.
Setting this to ‘loose’ will allow TileCache to generate tiles outside the maximum bounding box. Useful for clients that don’t know when to stop asking for tiles.
Setting this to “google” will cause tiles to switch vertical order (that is, following the Google style x/y pattern).
USING TILECACHE WITH OPENLAYERS
To run OpenLayers with TileCache the URL passed to the OpenLayers.Layer.WMS constructor must point to the TileCache script, i.e. tilecache.cgi or tilecache.py. As an example see the index.html file included in the TileCache distribution.
Note: index.html assumes TileCache is set up under CGI (see above). If you set up TileCache under mod_python you’d need to slighly modify index.html: the URL passed to the OpenLayers.Layer.WMS constructor must point to the mod_python script as opposed to the CGI script, so replace tilecache.cgi with tilecache.py. Similarly, you would need to edit this URL if you were to use TileCache with the standalone HTTP Server or FastCGI.
The most important thing to do is to ensure that the OpenLayers Layer has the same resolutions and bounding box as your TileCache layer. You can define the resolutions in OpenLayers via the ‘resolutions’ option or the ‘maxResolution’ option on the layer. The maxExtent should be defined to match the bbox parameter of the TileCache layer.
If you are using TileCache for overlays, you should set the ‘reproject’ option on the layer to ‘false’.
USING TILECACHE WITH MAPSERVER
MapServer has a map level metadata option, labelcache_map_edge_buffer, which is set automatically by TileCache to the metaBuffer plus five when metaTiling is on, if it is not set in the mapfile.
If you are using MetaTiling, be aware that MapServer generates interlaced PNG files, which PIL will not read. See http://mapserver.gis.umn.edu/docs/faq/pil_mapscript on how to resolve this.
SEEDING YOUR TILECACHE
The tilecache_seed.py utility will seed tiles in a cache automatically. You will need to have TileCache set up in one of the previously described configurations.
tilecache_seed.py <url> <layer> [<zoom start> <zoom stop> [<bbox>]]
Seeding by center point and radius
If called without zoom level arguments, tilecache_seed.py will assume that it needs to read a list of points and radii from standard input, in the form:
<lat>,<lon>,<radius> <lat>,<lon>,<radius> <lat>,<lon>,<radius> <lat>,<lon>,<radius> <ctrl + d>
The format of this file is:
the position(s) to seed longitude
the position(s) to seed latitude
the radius around the lon/lat to seed in degrees
An example with zoom levels 5 through 12 would be like;
$ tilecache_seed.py "http://example.com/yourdir/tilecache.cgi?" Zip_Codes 5 12 "-118.12500,31.952162238,-116.015625,34.3071438563"
The bbox can be dropped and defaults to world lonlat(-180,-90,180,90):
$ tilecache_seed.py "http://example.com/yourdir/tilecache.cgi?" Zip_Codes 0 9
In center point/radius mode, the zoom level range is not specifiable from the command-line. An example usage might look like:
$ tilecache_seed.py "http://example.com/yourdir/tilecache.cgi?" Zip_Codes -118.12500,31.952162238,0.05 -121.46327,32.345345645,0.08 <Ctrl+D>
… the seeding will then commence …
CLEANING YOUR TILECACHE
The tilecache_clean.py utility will remove the least recently accessed tiles from a cache, down to a specified size.
tilecache_clean.py [options] <cache_location>
show program’s version number and exit
- -h, --help
show this help message and exit
- -s SIZE, --size=SIZE
Maximum cache size, in megabytes.
- -e ENTRIES, --entries=ENTRIES
Maximum cache entries. This limits the amount of memory that will be used to store information about tiles to remove.
The –entries option to tilecache_clean.py is optional, and is used to regulate how much memory it uses to do its bookkeeping. The default value of 1 million will hopefully keep RAM utilization under about 100M on a 32-bit x86 Linux machine. If tilecache_clean.py doesn’t appear to be keeping your disk cache down to an appropriate size, try upping this value.
tilecache_clean.py is designed to be run from a cronjob like so:
00 05 * * * /usr/local/bin/tilecache_clean.py -s500 /var/www/tilecache
Note that, on non-POSIX operating systems (particularly Windows), tilecache_clean.py measures file sizes, and not disk usage. Because most filesystems use entire file blocks for files smaller than a block, running du -s or similar on your disk cache after a cleaning may still return a total cache size larger than you expect.
Occasionally, for some reason, when using meta tiles, your server may leave behind lock files. If this happens, there will be files in your cache directory with the extension ‘.lck’. If you are seeing tiles not render and taking multiple minutes before returning a 500 error, you may be suffering under a stuck lock.
Removing all files with extension ‘.lck’ from the cache directory will resolve this problem.
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.