[mapserver-users] How Interval is defined for WMS-T

[Even Rouault]

Le mardi 16 février 2016 15:54:20, thomas bonfort a écrit :
> Even,
> I'm not sure I agree with your conclusion.

Which one ;-) ? 
I agree with your below analysis that there's a bug for resolution=second but 
my points were a bit different :
- the resolution selected is derived from the value of user specified time 
parameter
- there is apparently a bug in the way wms_timeformat enforces the specified 
formats (I've tried to set it to "YYYY-MM-DD HH:MM", but values matching 
"YYYY-MM-DD HH:MM:SS" are also accepted)

> 
> I think the intention would be to write
> 
> timecol between time0 and time0+resolution
> 
> which translates to:
> 
> time0<=timecol<=time0+resolution
> 
> whereas we want:
> 
> time0<=timecol<time0+resolution
> 
> thus the hack to use time0+resolution-1second which is clearly incorrect
> when the desired resolution is the second (and somewhat incorrect for the
> other resolutions)
> 
> 
> I'd say this is a bug.


> 
> --
> thomas
> 
> On 16 February 2016 at 15:43, Even Rouault <even.rouault at spatialys.com>
> 
> wrote:
> > Le mardi 16 février 2016 15:21:17, Yves Jacolin a écrit :
> > > Hello,
> > > 
> > > I have a WMS-T layer. The datasource is in a postgis layer. The SQL
> > 
> > request
> > 
> > > send by MapServer is something like this:
> > > 
> > > ("timestamp" between date_trunc('second',timestamp
> > > '2016-02-17T15:00:00.000Z') and date_trunc('second',timestamp
> > > '2016-02-17T15:00:00.000Z') + interval '1 second' - interval '1
> > > second')
> > > 
> > > I understand the + interval '1 second' - interval '1second', but I
> > > can't understand how the first interval resolution come from, see [1].
> > > 
> > > I guess this is calculated from wms_timedefault or wms_timeextent. Here
> > 
> > my
> > 
> > > METADATA LAYER of my mapfile:
> > >             "wms_timeextent" "2016-02-16T22:20/2016-02-17T21:20/PT1H"
> > >             "wms_timeitem" "timestamp"
> > >             "wms_timedefault" "2016-02-16T23:20"
> > > 
> > > Any idea how I can "force" the interval resolution to minute?
> > 
> > Yves,
> > 
> > The resolution comes from the value of time parameter passed in the WMS
> > request (or the timedefault value if time is not specified). To get a one
> > minute resolution, the time parameter must be expressed in minutes.
> > 
> >  This is
> > 
> > actually explained in http://mapserver.org/ogc/wms_time.html (
> > "Interpreting
> > Time Values", "Exceptions to this rule")
> > 
> > AFAICS there's no way to override that currently (although it might
> > potentially be desirable, for example by taking into account the
> > resolution in
> > the timeextent syntax)
> > 
> > One workaround to indicate the intent to the client is to specify
> > "wms_timeformat". Although it doesn't seem to work properly currently, ie
> > if
> > you specify "YYYY-MM-DDTHH:MM", a value with seconds will be accepted as
> > validating the beginning of the pattern. Hum looks like a bug.
> > 
> > Even
> > 
> > --
> > Spatialys - Geospatial professional services
> > http://www.spatialys.com
> > _______________________________________________
> > mapserver-users mailing list
> > mapserver-users at lists.osgeo.org
> > http://lists.osgeo.org/mailman/listinfo/mapserver-users

-- 
Spatialys - Geospatial professional services
http://www.spatialys.com