question about shp2tile

[Stephen Woodbridge]

Zhonghai Wang wrote:
> Hi guys,
>  
> I've tried to make a efficient map layer with the commands shp2tile, 
> tile4ms, and shptree, but something seems not right, because MapServer 
> can not draw the maps. I've made a test in a seperate folder, and I've 
> done the following processes:
>  
> ***
> data -- Forests.shp (for a country)
> ***
> 1. >shp2tile -r 50 -c 50 Forests.shp Forests_Test.shp
>  >>>results for this step: shp, shx, and dbf files, there is no prj file 
> for the output file
>  >>>error info on the console: failed to create shp Forests_Test.shp -1833

This is likely because you are trying to create 2501 files (50 x 50 + 1) 
and they all half to be open at the same time for this mode so you 
likely ran into a process file handle limit. Also how many point do you 
have in this layer? What is the value of Num_points/2500, this number 
should not be smaller than 8000 - 10000.

> 2. >tile4ms ---- get the tileindex.shx, tileindex.shp and tileindex.dbf 
> files for the Forests_Test.shp, there is still no prj file for tileindex.shp
>  >>>no error messages appear at this step

There are no prj files created. Mapserver does not use them.

> 3. >shptree -- to genetare .qix file for all shapefiles in this subfolder
>  >>> results for this step: Forests.qix, Forests_Test.qix, and tileindex.qix
>  >>>no error info on the console
>  
> but, only if I set DATA "Forests" in the layer object the map will be 
> rendered, others like DATA "Forests_Test" or TTILEINDEX "tileindex" 
> TILEITEM "location" do not work, the server simply sends back a blank 
> image.
>  
> did I make any mistakes or it really do not work well. (I am using MS4W 
> 4.8.1)
>  
> I guess some errors occur when I perform the command shp2tile.

What version of shp2tile are you using?

woodbri at carto:/data/mdata$ ~/dev/shptools/shp2tile -v
$Id: shp2tile.c,v 1.13 2005/12/05 22:38:08 woodbri Exp $

If it does not respond with the Id string above you need to upgrade as 
there is a serious crashing bug for point data in the earlier version if 
you are using the -q option. It should work for the row col option, just 
try to decrease the number of tiles.

Also if you do not have something like R x C files in your directory 
then the process failed. For the row col option you can also specify 
--no-write to just get a stats report of how the data is put into the tiles.

-Steve

> thanks for any further info.
>  
> zhonghai
>  
> On 5/18/06, *Zhonghai Wang* <zhonghaiw at gmail.com 
> <mailto:zhonghaiw at gmail.com>> wrote:
> 
>     Hi Bob, Steve,
>      
>     thank you very much for all these helpful clues, now I think I've
>     got the points of the shp2tile command, it's really a good tool to
>     slice shapefile.
>      
>     zhonghai
>      
>     On 5/18/06, *Stephen Woodbridge* <woodbri at swoodbridge.com
>     <mailto:woodbri at swoodbridge.com>> wrote:
> 
>         Zhonghai Wang wrote:
>>  Hi folks,
>>
>>  I have a large shapefile, now I am trying to use shp2tile
>         command to
>>  slice it into pieces, with -r and -c is ok, but I do not fully
>>  understand the -q parameter, what does it actually mean? and
>         what number
>>  should a use for this parameter normally?
>>
>>  or something like this? -- >shp2tile -q 10000 input_shapefile
>>  output_shapefile
> 
>         Hi Zhonghai,
> 
>         The -r -c option breaks the extents of your shapefile into R x C
>         rows
>         and columns and then tries to fit the objects into the best
>         tile. I any
>         tile crosses a tile boundary by 5-10% then it is put into a
>         "supertile"
>         the could be the same extents as the original shape file. So
>         typically
>         you will end up with r X c + 1 tiles.
> 
>         The -q N option splits the extents in half either vertically or
>         horizontally and then sorts the objects into the 2 halves or put
>         them in
>         a supertile. Then if the either of the two halves has more than N
>         objects it is again split in half and this continues until all files
>         have less than N objects. This can cause some strange effects
>         like tiles
>         with 1 or a small number of objects and most tiles will have
>         less than N
>         objects in them. Since this algorithm tends to spatially cluster
>         objects
>         in a file, there is a good chance that if you need the file that
>         all or
>         most objects in the file will be used.
> 
>         I recommend trying numbers like 10,000 and 20,000 as you initial
>         tries.
>         I think you should probably not use numbers less then 8000, but
>         it is
>         really up to you to try and measure the results to find what
>         works best
>         for your data.
> 
>         -Steve W.
> 
> 
>