Ticket #1103 (reopened defect)

Opened 4 years ago

Last modified 3 years ago

Layer tolerance default value

Reported by: jlacroix Assigned to: jlacroix
Priority: high Milestone: FUTURE
Component: MapServer C Library Version: 4.4
Severity: normal Keywords:
Cc: jmckenna@dmsolutions.ca

Description 

The tolerance default value in the layer is 0. This works well with polygon
layers since this value will return query result for a click inside the polygon.
However for line and point layers a tolerance of 0 seems always return nothing
(It's the like the layer is not queryable). In the documentation the default
tolerance is marked as 3, but in the code it's 0.

Should we update the documentation or update the tolerance value?
Steve?

Change History

12/03/04 12:27:42 changed by sdlime 

The problem is that there is no perfect default value. Either 3 or 0 is 
problematic in certain cases.

I think we need a way to set a default based on layer type. That would be easy 
via straight CGI or simple mapfile loading via MapScript. It's harder 
(impossible?) when creating layers dynamically, unless we require layer type in 
the constructor.

In lieu of a fix as described above I would vote to change the documentation 
since people have probably already reacted to the behavior of the software. 
Changing documentation doesn't break applications...

Steve

12/03/04 13:35:58 changed by jlacroix

  • cc set to jmckenna@dmsolutions.ca.
I agree with you on this one. Since we don't know the layer type at the
beginning we can't set the default tolerance. I will add Jeff in CC to update
the Doc.

Jeff: The Layer TOLERANCE default value is 0, not 3 like stated in the doc.

Steve: Maybe we can reuse the same trick for layer tolerance default value that
we use for the symbol default size. Set it to -1 by default. When doing a query,
use the tolerance value or a default value based on the layer type if it's -1.
Just want to have a thought on this.

12/03/04 18:05:02 changed by sdlime 

Julien: Good idea on setting tolerance in the query functions. Let's pursue 
that for 4.6, but update the documentation now.

Steve

04/13/05 11:53:54 changed by jlacroix

  • milestone set to 4.6 release.
Aiming 4.6 for the dynamic tolerance default value.
I may take care of this one if I have time.

04/15/05 11:13:11 changed by jlacroix

  • owner changed from sdlime to jlacroix@dmsolutions.ca.
  • cc set to mapserver-bugs@dmsolutions.ca.
Reassign to me, I will start working on this today.

04/15/05 15:40:09 changed by jlacroix

  • status changed from new to closed.
  • resolution set to fixed.
I implemented this. The default is now -1. When doing a query, it check for the
layer type. If it's a POINT or a LINE, the default is 3. For all other layer
types, the default is 0.

Marking as FIXED.

Reopen if you see any other issue.

04/15/05 15:54:39 changed by sdlime 

Where specifically is the default being set now? Queries are notoriously 
complex and you would've had to modify a number of functions. Just want to make 
sure nothing was missed...

Steve

04/15/05 16:19:01 changed by jlacroix 

It seems that the query stuff was only using the tolerance value from the layer
object only at 4 places. you can grep on "layer_tolerance" to see where this is.

If I forgot something, please reopen and I will try to fix it.

04/15/05 17:04:07 changed by jmckenna@dmsolutions.ca 

julien did u update the docs for this, or does that still remain?

04/19/05 16:37:44 changed by jlacroix

  • dependson set to 1323.

04/19/05 16:39:14 changed by jlacroix 

Sorry for the late answer jeff. I opened bug 1323 for updating the documentation
about this.

Just to know, where are the files to update for the doc?

04/20/05 11:07:57 changed by jmckenna@dmsolutions.ca 

no problem

:pserver:msdoc@cvs.gis.umn.edu:/data2/cvsroot 
module: mapserver_docs
file: mapfile-reference.xml

i'm on it

04/20/05 13:45:04 changed by jmckenna@dmsolutions.ca

  • keywords set to VERIFIED.
tested.  nice improvement.

10/26/05 00:01:57 changed by sdlime

  • keywords deleted.
  • status changed from closed to reopened.
  • resolution deleted.
I think we may need to re-examine this. I just ran into a situation where these
mods didn't make sense. Specifically with the query by shape functions. I'm not
sure anything but a default tolerance of 0 for all layer types makes sense for
these functions. If you're not careful it's easy to over select points with
polygons by default. To me if I'm using a polygon to select points the default
should be only the points in the polygon. That's not the case now.

I'm of the opinion now that that default of 3 for points and lines makes sense
with point based queries only but in all other cases the default should be 0.

Thoughts?

Steve

11/05/05 01:13:04 changed by sdlime 

Any thoughts on this guys? It's really a problem because in some cases because
you can end up with more results than you should have. Depending on the
situation that's a very subtle error. I prefer less subtle errors like no points
found. I think we can do better. A default tolerance of 0 should be used for
points and lines when using doing box or polygon queries. A tolerance of 3 makes
sense with point or line-based queries.

Steve

01/10/06 15:16:49 changed by jmckenna@dmsolutions.ca 

i was just trying to test comment#13.  I tested with cgi (itasca demo,
rubberband box query) and phpmapscript (gmap demo, box query) and could not
over-select points (and i removed tolerance values set in mapfile)  I tested
with 4.8.0-rc2.

sorry.

01/10/06 15:40:50 changed by sdlime 

The place I ran into this with was when querying by shape within MapScript and 
it wreaked havoc trying to debug because points outside the query shape were 
being selected and the number selected varied by image resolution (where the 
tolerance is computed). The persons doing the development were expecting strict 
intersection when using non-point based selection and that's not that case. 

The software is working correctly, just not in the way the user expects.

On one hand this is simply a RTFM problem. On the otherhand a tolerance other 
than 0 only makes sense for point-based queries of points and lines.

Steve

01/10/06 15:59:12 changed by jmckenna@dmsolutions.ca 

k.  in that case I agree with your recommendations in comment#13.

01/10/06 16:38:05 changed by jlacroix 

Steve do you think it's possible to use different default values only in case of
point and line queries?

I think only in case of point query is feasible since it got its own function.

01/10/06 17:59:55 changed by sdlime

  • milestone changed from 4.6 release to FUTURE.
Yup, I think it's possible. The only place right now we should set a default 
tolerance other than zero would be in msQueryByPoint(). We could also do 
something in msQueryByFeature and msQueryByShape based on feature/shape type 
but those functions only support polygon selection shapes. 

Mind you, this is nothing that needs to be fixed ASAP so I'm marking as 
future...

Steve