[mapserver-dev] RE: [UMN_MAPSERVER-DEV] Performance transparency

[Steve Lime]

I believe Thomas already has it working but I don't know that it is in SVN yet. It will only be supported for the AGG renderer  I think. There are certain caveats about it's use (for example it doesn't work well with certain types of linear data presentations. This should be in place for 5.2.

One question that remains is how to trigger it. I favor use of the opacity keyword as opposed to the specifying an alpha channel in a color.

Steve

>>> "Hallgren, Johan" <johan.hallgren at logica.com> 03/20/08 5:26 AM >>>
Hello

I have had a question last year related to performance when using transparency. Steve mentioned that there are plans to support feature-level transparency. Do we have any progress at that issue? When can I count on that feature?

/Johan



___________________________________



Johan Hallgren



Logica

Pelle Bergs backe 3

Box 1938, 791 19 Falun

Tel (och mobil): 023-547 46 (int: +46-2354746)

johan.hallgren at logica.com

http://www.logica.se



-----Original Message-----
From: UMN MapServer Developers List [mailto:MAPSERVER-DEV at LISTS.UMN.EDU] On Behalf Of Steve Lime
Sent: den 19 december 2007 20:31
To: MAPSERVER-DEV at LISTS.UMN.EDU
Subject: Re: [UMN_MAPSERVER-DEV] Performance transparency



Hi Johan: I've never quantified the performance hit but it is certainly expected. Layer-level transparency works like this:



  - a temporary image is created to render the transparent layer into, it is initialized to be completely transparent

  - rendering proceeds as normal in the temp image

  - the temp image is then "merged" into the base image, so for every non-transparent pixel in the temp image alpha blending is done and the base image pixel is modified

  - the temporary image is discarded



So, the processing time will increase with 1) image size and 2) vector layer density.



There are plans to support feature-level transparency but it won't work in all cases so layer level transparency will remain an option.



So, the results you are seeing are expected although I can't say if double is too much or not.



Steve





>>> On 12/19/2007 at 8:49 AM, in message

<D93CFEAA72FA764885CC1FDA94A1D89C035462F2 at WMRI000166.corp.wmdata.net>, Hallgren

Johan E <jhhal at WMDATA.COM> wrote:

> Hello Friends

>

> I a question regarding performance when using a transparent vectorlayer

> on top of a rasterlayer. The output is quite large (1400 * 1000 px) and

> vectordata is a shapefile with polygons.

>

> When turning the transparency on the vector layer the performance goes

> down quite much. It takes maybe double as long time to produce the map.

>

> Is there any knowledge about that?

>

> /Johan

>

> ___________________________________

>

>

> Johan Hallgren

>

> WM-data, Logica CMG

> Pelle Bergs backe 3

> Box 1938, 791 19 Falun

>

> Tel (och mobil): 023-547 46 (int: +46-2354746)

>

> johan.e.hallgren at wmdata.com

> <blocked::mailto:johan.e.hallgren at wmdata.com>

>

> http://www.wmdata.se

> <blocked::https://webaccess.wmdata.com/exchweb/bin/redir.asp?URL=http://

> www.wmdata.se/>

>

>