[postgis-users] Questions related to some tests I did in PostGIS

[Bborie Park]

>
>
> So, my questions are:
> Why counting pixels was not faster in the second subset with pixels of 5x5
> meters?
>
>
I cannot tell based upon your queries. I can't distinguish which query
times are for the 5x5 vs 1x1. Have you tried doing outdb? It may be more
performant for a source raster of that size.


> How ST_Count works? Does it uses the ST_Value function in each pixel of the
> pixel matrix row by row in each tile to keep track of the nodata value
> pixels?
>
>
No. ST_Count goes a different pathway. It does need to check each pixel for
the NODATA value though. I suspect ST_Count is just returning the "count"
value of ST_SummaryStats as the general cost of counting pixels in a raster
is the majority required for summary stats.


> Is the raster support in postgis, in general, only suitable to get pixel
> statistics of some region in the raster?
>
>
Nope. That is the general use case but I know of places where modeling
surfaces is done in-database.


> Whats the advantages of using postgis raster to analyze rasters compared to
> some of the arcgis tools (for example)? I ask that because is much more
> easier to visualize the data in arcgis compared to postgis tables in QGIS
> plugin (wkt raster)
>
>
Depends on how big your dataset is. If you're working on longitudinal
datasets (over space and time) the database is better prepared for work on
these kinds of datasets.

One thing that I realize is that ST_Area could be faster at least in the
case of no skew by not going to geometry.

-bborie
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/postgis-users/attachments/20140917/64cb3351/attachment.html>