[ka-Map-users] Some thoughts/questions about tile.php

[Stephen Woodbridge]

Tim,

Thanks, for the link. I knew this was possible but since I don't have 
all the times generated I'm stuck using auto for the time being.

Did anyone get mod_rewrite or a 404 handler to redirect a cache hit to 
tile.php if the file did not exist. This does not solve the problem of a 
partially written file in the cache, should whould require changes like 
I outlined below.

-SteveW

Tim Schaub wrote:
> Just a reminder that there is a "tile_source" "cache" option for layer
> metadata.  If you set this in your mapfile, the call to tile.php is
> avoided completely (tile source is determined on the client side).  This
> does speed things up a bit.  I agree that it would be good to optimize
> tile.php - anything that you could do would be somewhere between the
> speed of the cache option and the current tile.php speed.
> 
> Terse docs:
> http://ka-map.ominiverdi.org/wiki/index.php/Mapfile_Metadata_Configurati
> on#Mapfile_LAYER_Section
> 
> Tim
> 
> -----Original Message-----
> From: ka-map-users-bounces at lists.maptools.org
> [mailto:ka-map-users-bounces at lists.maptools.org] On Behalf Of Stephen
> Woodbridge
> Sent: Thursday, July 13, 2006 1:33 PM
> To: ka-map-users at lists.maptools.org ML
> Subject: [ka-Map-users] Some thoughts/questions about tile.php
> 
> Hello Paul, et al,
> 
> I have been looking at tile.php and had some thoughts/questions about
> how it works and how it might be made faster.
> 
> 1) currently the first thing it does is open lock file on the metatile
> 2) if the file exist, it gets a shared lock
> 3) serves the image
> 4) removes the lock
> 
> It seems like it would be faster to avoid this locking, ie: if the file
> exists just serve it. If image creation were written to a tmp file name
> and renamed to the actual tile name when the write was done and the file
> closed, this would be atomic and should avoid the problem that the code
> above is dealing with.
> 
> Since the whole concept of tiled mapping is to generate once and serve
> many times it seems this would be a valuable performance improvement by
> avoiding 3 file operations per tile requested.
> 
> 
> Second issue is that directory structure is pretty flat as Paul has
> mentioned in the past. I was thinking that a very trivial change would
> be to split the directories tnnnnn-lnnnnnn into tnnnnn/lnnnnnn which
> would significantly reduce the number of entries in any given directory.
> 
> Any thoughts or issues on these would be appreciated.
> 
> Best regards,
>    -SteveW
> _______________________________________________
> ka-Map-users mailing list
> ka-Map-users at lists.maptools.org
> http://lists.maptools.org/mailman/listinfo/ka-map-users
>