Imported Discussion of MapsForge file format

78 views
Skip to first unread message

Paul Daisey

unread,
Mar 14, 2012, 10:30:16 AM3/14/12
to geospatial-mobile-da...@googlegroups.com

From: Paul Daisey [mailto:pdais...@gmail.com]
Sent: Wednesday, March 14, 2012 9:30 AM
To: Frantz, Nathan P ERDC-TEC-VA
Cc: 'je...@imagemattersllc.com'; Backe, Kevin AGC
Subject: Re: Fw: MapsForge Tiel Format

 

Nathan:

From reading the description it looks to me like the images are contained with everything else in one or more binary files.  (It is unclear whether "subfiles" are separate files or just segments of one file.)  The index design (header points to tile indexes which point to images, poi and ways data) does not constrain where the image blobs are located in the binary file.  I found no mention of the actual image formats.  This is as simple and crude an interface as is possible - a pointer to a file offset and a length.  Pro: that makes it possible to build very light weight clients; Con: requires custom software for all load, access and update functions.

There are a lot of reasons why software has evolved from fixed formats like this (and hand-optimized code to deal with them) to a layering and modularity componentization specialization of functions with standard interfaces.  These are often called "ilities" - portability, scalability, extensibility, adaptability, reliability, maintainability, evolve-ability, security, interoperability - and lower total lifecycle cost.    The inflexibility of the MapsForge vector storage format is a case in point.  On the other side is the rise of Java, which despite its overhead has become a standard programming interface within both Oracle and Android as well as a wide variety of other platforms.

I'd argue that given the requirement for a container that holds tiles, rasters, and vectors, that we're better off incurring the relatively minimal overhead of an SQL interface like the one provided by SQLite.  Pro: standard interfaces (ODBC / JDBC) are widely implemented; Con: larger ( < 1m http://www.sqlite.org/footprint.html) application footprint.  We'd have to benchmark the alternatives to know how much slower the SQL impl would be; they both might use the same low level file I/O seek / read routines under the hood. 

Paul

On 3/13/2012 10:00 AM, Frantz, Nathan P ERDC-TEC-VA wrote:

I definitely have no interest in the vector storage. Ben has just suggested we use this as a possible tiling spec but I can’t find very much information on the format.

 

Thanks,

 

Nathan Frantz
Research Geographer
Topographic Engineering Center
Engineer Research and Development Center
703-428-6919

BB 571-535-0567
www.agc.army.mil
Nathan....@erdc.usace.army.mil
 

 

From: Paul Daisey [mailto:pdais...@gmail.com]
Sent: Tuesday, March 13, 2012 9:55 AM
To: Frantz, Nathan P ERDC-TEC-VA
Cc: 'je...@imagemattersllc.com'
Subject: Re: Fw: MapsForge Tiel Format

 

Nathan:

    I looked at MapsForge data structure a bit more yesterday.  Point features are stored as POIs.  ALL linear and area features are stored as "ways".  Both record the first coordinate as an offset from the tile corner.  Successive points in ways are recorded as differences to save space.  This may make for fast displays, but I'd expect it to be very bad for supporting analysis.  There is just an offset to the tile data; I could not find anything about image formats, other than a statement referencing OSM tiles.  But OSM does not say anything about image formats either.  That is apparently left for the web browser to handle.

Paul

On 3/7/2012 12:58 PM, Frantz, Nathan P ERDC-TEC-VA wrote:

It supports tiles as well.. I was just curious as to the format and how it was encoding the data

 

Nathan Frantz
Research Geographer
Topographic Engineering Center
Engineer Research and Development Center
703-428-6919

BB 571-535-0567
www.agc.army.mil
Nathan....@erdc.usace.army.mil
 

 

From: Paul Daisey [mailto:pdais...@gmail.com]
Sent: Wednesday, March 07, 2012 11:55 AM
To: Frantz, Nathan P ERDC-TEC-VA
Cc: 'pdais...@gmail.com'; 'je...@imagemattersllc.com'
Subject: Re: Fw: MapsForge Tiel Format

 

It is aimed at fast display of OpenStreetMap data, which is a very small and specialized subset of vector feature data.

On 3/6/2012 10:48 AM, Frantz, Nathan P ERDC-TEC-VA wrote:

Initial thoughts on this format?
 

From: Tuttle Benjamin -Ben- T NGA-IIG USA CIV [mailto:Benjamin...@nga.mil]
Sent: Tuesday, March 06, 2012 09:33 AM
To: Frantz, Nathan P ERDC-TEC-VA
Subject: MapsForge Tiel Format
 

Reply all
Reply to author
Forward
0 new messages