TileIndexes and paths - a proposal

[Brent Fraser]

----- Original Message -----
From: "Frank Warmerdam" <warmerdam at POBOX.COM>
To: <MAPSERVER-USERS at LISTS.UMN.EDU>
Sent: Friday, October 22, 2004 1:32 PM
Subject: Re: [UMN_MAPSERVER-USERS] TileIndexes and paths - a proposal


> Brent,
>
> If we can agree on how the tile index names, I am happy to implement it.
> I haven't thought through your proposal carefully yet, but hopefully I
> will get a chance to.
>
> Brent Fraser wrote:
> > A couple of more questions...
> > ======================
> > 1.  What if the tileindex is a database?
>
> I presume you mean, what if the tileindex dosn't have a meaningful path?
> Good question.
>
At the first MUM, Sean G. had asked about using a database as a tileindex,
but at that time it was no-go due to the single layer architecture for the
tileindex and tiles.  Since Steve L. added the two layer architechure for
the tileindex and tiles, this looks more feasible, but my guess is that it
is not implemented.

So let's say it could be done.  How would that impact the tile paths?  One
solution is to ignore the DATA object in the tileindex LAYER (have to, since
it's not a path), so basically the full path is kept in the database (except
for the optional DATA object, AKA the filename, in the layer definition).
If you move the tiles, you would need to issue an Update SQL statement.  Not
great, but OK.


> The other side of this is to keep in mind that the names in the tileindex
> aren't necessarily filenames.  In the case of GDAL formats, they could
> be a gltp: url for OGDI data, or a DODS url for DODS servers.
>

My thought is that the fully qualified tile path should be something that at
least one of the GDAL drivers could open.  For example suppose I wanted to
serve up VMAP0 data in it's original format (probably not a good idea
performance-wise).    Currently I think the mapserver code has some logic to
determine if the path is relative or absolute; it would have to be enhanced
to recognize the following as an absolute path:

gltp:/vrf/usr4/mpp1/v0eur/vmaplv0/eurnasia 'watrcrsl at hydro(*)_line'

and pass it the GDAL library.  Likely some other issues as well.  On the
other hand, I don't think anyone has asked about this, so maybe it is not an
issue....


> > 2.  What if the tileindex is in a different coordinate system than the
> > tiles?
>
> At this time that is not permitted.  All tiles in a tile index have to
> be in the same coordinate system as the tileindex.

This is an enhancement that I would like to see.  I implemented it in my
local copy of MS 4.0 and could pass the source to some brave soul willing to
integrate it into CVS.  Basically it required having a .prj file with the
tileindex and ensuring the tiles themselves had georeferencing (either
internal for GeoTiffs, or .prj files for shapefiles etc.).  The reason I
didn't submit the code is that I believe the .prj file code should actually
be in GDAL so formats without built-in coordinate system definition support
could be more easily manipulated.  See my email from May 12, 2003
"Projection AUTO, GDAL and ECW files".

So many enhance requests, so little time...
Brent