[GRASSGUI] graphic overlays

Michael Barton michael.barton at asu.edu
Sun Mar 25 14:13:02 EDT 2007 

I can agree with all of this.

Please keep in mind, all, that I would love to be able to write scripts that
could display maps in a wxPython canvas with all kinds of nice decorations
like legends, arrows, and scales. In my lab, we are currently developing
dynamic GIS and cellular automata models, as well as coupled agent-based and
surface process models, that call for dynamic displays to track changing
landscapes. At the present, we can only do that via an x monitor, but I
would prefer to move this into a better display environment.

So I definitely favor a scriptable display architecture. However, this may
require code beyond basic GRASS C modules--like the new scripts that Jachym
has written. 

The scripting that we are doing in my lab adds iterative dynamics to GRASS
commands via bash shell scripting. This is not practical to do with
individual GRASS commands via the command line. That's why it's great that
GRASS is so scriptable. But we are definitely adding functionality to GRASS
that is not there in the C modules alone.

The same thing goes for GUI code. I agree with the direction that Daniel is
advocating--and in fact maybe we all agree in general. I'm just trying to be
practical and realistic. We might well be able to have command-line control
of complex displays, including NVIZ-like displays. However, we may only be
able to do this if we have access to wxPython and OGL. That is, it may not
be possible only with C code command modules and nothing else. I don't think
that is inherently bad. We use GDAL and PROJ code to handle import/export
and projection, and the current GUI won't run unless TclTk is available. So
there are precedents. I agree that we always need to minimize dependencies
as much as possible. But some are unavoidable, or at least highly desirable.

Since I don't write C code, maybe I'm being overly pessimistic. However,
Glynn is pretty much the only one on the team who is writing new and
sophisticated graphics rendering routines into GRASS C code. This has made
an enormous difference, but he hasn't had much help and I am unable to offer
any. 

Michael


On 3/25/07 9:43 AM, "Daniel Calvelo" <dca.gis at gmail.com> wrote:

> On 3/25/07, Michael Barton <michael.barton at asu.edu> wrote:
> [...]
>> This is the same for a GUI--which also augments GRASS commands with a
>> graphical and interactive environment. To say that we should not develop a
>> modern GUI for displaying GRASS GIS data in a rich, interactive, graphical
>> environment simply because it cannot be duplicated from typing textual
>> commands seems very restrictive.
> 
> Not restrictive, it is demanding.
> 
>> Should we drop transparency--which is a
>> very useful tool as well as a way to make visually appealing and informative
>> maps--simply because it is not easy to do by typing? Should we drop NVIZ
>> from GRASS simply because you can't get the same effects by typing on the
>> command line?
> 
> We shouldn't drop them, but we must strive at making the added
> functionality a part of the GRASS system as a whole. I mean, e.g., if
> we add mouse-based placement of north arrows, we must strive at
> providing similar functionality from the command-line, i.e. a
> d.northarrow command whose parameters may be produced by the
> interactive version.
> 
> This is not very hard to do, as making it in python is a fine first
> step. It is just a constraint (and I think a very good one) and
> discipline in the way we must structure and think about the "gui"
> code.
> 
> For instance, in the north arrow matter, it is better to use a python
> reader for GRASS' icon/symbol description language and build nice
> north arrow in that language than to use, say, SVG astounding
> gradient-filled and alpha-blended symbols, which cannot easily be used
> by the rest of GRASS code.
> 
>> I don't think that would serve the GRASS user community well.
> 
> Of course not. IMO the matter under discussion is reusability of
> current developments and newer functionality under the original GRASS
> philosophy.
> 
> BTW, anybody has pointers to a discussion that seems to have arisen
> with respect to GRASStep?
> 
> Daniel.
> 
>> Michael
>> 
>> 
>> On 3/25/07 12:52 AM, "Martin Landa" <landa.martin at gmail.com> wrote:
>> 
>>> Hi,
>>> 
>>> just *note* -- from my point of view it would be nice to design new
>>> GUI only as an interface (without any extra functionality). User can
>>> choose between CLI or GUI. But the functionality (including
>>> decorations) would be almost identical (i.e. accessible via GUI or
>>> CLI). E.g. I don't like that the user can set transparency level of
>>> raster layers via GUI in the current GIS Manager but not via CLI (I
>>> mean using directly d.rast).
>>> 
>>> Martin
>>> 
>>> 2007/3/25, Michael Barton <michael.barton at asu.edu>:
>>>> 
>>>> On 3/24/07 8:25 PM, "Glynn Clements" <glynn at gclements.plus.com> wrote:
>>>> 
>>>>> 
>>>>> Michael Barton wrote:
>>>>> 
>>>>>>>> I want to suggest that most or all graphic overlays be done via
>>>>>>>> wxPython
>>>>>>>> rather than using GRASS d.* commands.
>>>>>>>> 
>>>>>>>> These are: north arrow, scale, grid (and geodetic lines maybe), and
>>>>>>>> text.
>>>>>>>> We¹ll need someone to design some nice (and GRASS distinctive) north
>>>>>>>> arrows
>>>>>>>> and scales.
>>>>>>>> 
>>>>>>>> I suggest that these be accessed through an item or items (popup
>>>>>>>> checkbox
>>>>>>>> menu?) on the map display tool bar, rather than having them added as
>>>>>>>> layers.
>>>>>>>> They should be able to be turned on and off of course, as well as being
>>>>>>>> positioned with a mouse (the latter will be much easier if they are
>>>>>>>> wxPython
>>>>>>>> graphic objects). Each will need it¹s own dialog with options, but that
>>>>>>>> shouldn¹t be too difficult to do.
>>>>>>>> 
>>>>>>>> Vector labels could go either way<as something done from the map
>>>>>>>> display
>>>>>>>> tool bar or as an independent layer.
>>>>>>>> 
>>>>>>>> Other opinions?
>>>>>>> 
>>>>>>> I suggest that this is implemented in such a way that the user can
>>>>>>> easily choose to use the equivalent d.* commands instead, so that the
>>>>>>> display contents can be exported as a script which can be used outside
>>>>>>> of the GUI (i.e. with an X/PNG monitor).
>>>>>> 
>>>>>> This kind of negates the idea (I thought you were promoting too) of using
>>>>>> native GUI functions for all 'decorations'.
>>>>> 
>>>>> FWIW, I don't support requiring the GUI in order to render anything
>>>>> which might reasonably appear on a printed map (labels, scale bar,
>>>>> legend etc).
>>>> 
>>>> I don't think we should necessarily limit the GUI to only being able to do
>>>> what plain GRASS command modules can do. The programming that goes into the
>>>> GUI is value added to anyone who uses this. Underneath, it is still just
>>>> running GRASS modules. So anyone who wants to script as creatively can get
>>>> the same results. But that doesn't mean that simply running a series of
>>>> GRASS commands will necessarily achieve the same result. Especially for
>>>> graphic rendering, a good GUI may well be able to some things that GRASS
>>>> outside the GUI cannot do. I don't see this as an issue. If people want to
>>>> use the features of the GUI they can. If they don't want to they don't have
>>>> to. If they want to recreate the GUI coding in another platform, that's
>>>> fine
>>>> too. But a bash shell script may not be able to achieve what can be done in
>>>> a wxPython script.
>>>> 
>>>>> 
>>>>> These should all be achievable using d.* commands, and the GUI should
>>>>> provide a convenient interface to those commands (e.g. using the mouse
>>>>> to identify coordinates for placement).
>>>> 
>>>> This is fine as far as it goes. d.grid is fairly complete as is. However,
>>>> d.barscale is very limited in its style of north arrows and scales. And
>>>> there is no way to adjust the size of either. If someone wants to program
>>>> in
>>>> a bunch of arrow and scale styles, along with ways of setting scale
>>>> parameters and arrow sizes, that's fine. But it may well be much easier to
>>>> to do this in the GUI than in a C module. I don't advocate getting rid of
>>>> d.barscale. But I don't think the GUI should be limited by its
>>>> restrictions.
>>>> d.text is similarly limited. You can do quite a bit more with text in the
>>>> GUI, and do so more easily (e.g., built in font-setting dialogs that are
>>>> appropriate to whatever platform it's run on). For all of these,
>>>> interactive
>>>> placing and moving with a mouse is difficult if we only use the d.* command
>>>> modules. While I can click and send coordinates to d.barscale (like we can
>>>> do in the TclTk GUI), I don't know how I could grab a north arrow with a
>>>> mouse and drag it to reposition it in another locale on the screen using
>>>> d.barscale. However, I can do that with wxPython generated graphics.
>>>> 
>>>>> 
>>>>>> It's not hard to add a layer
>>>>>> managers for d.grid, d.text, d.scalebar, etc. But it could clutter up the
>>>>>> main control more.
>>>>>> 
>>>>>> What if we could make it scriptable using wxPython graphic elements?
>>>>> 
>>>>> I'm not entirely sure what you mean by that.
>>>>> 
>>>>> But, essentially, I feel that it should be possible to save the
>>>>> contents of the display as a script which will work on a system with
>>>>> only core GRASS modules (no Python, no X server; can you render
>>>>> wxPython graphics into an image file on a system with no X server?).
>>>> 
>>>> NVIZ can be scripted, but you can't render an NVIZ image without NVIZ. I
>>>> don't see that as a problem. If you want to render a 2.5D or 4D image, you
>>>> need NVIZ. There is no way to render this using plain vanilla GRASS
>>>> commands. I wouldn't want to not have this tool simply because we couldn't
>>>> render its images with plain command line C modules.
>>>> 
>>>> I'm a very big advocate of GRASS's scripting potential. I've written a
>>>> number of scripts. And this scriptability is what makes it possible to wrap
>>>> the commands in a nice GUI. But as you've pointed out, there are
>>>> limitations
>>>> to graphic output rendering from command-line modules alone. Continued
>>>> improvements to ps.map may be a way to achieve this. Perhaps it could be
>>>> made to render to a PNG or PPM file as well as a postscript one. Just some
>>>> thoughts.
>>>> 
>>>> Michael
>>>> 
>>>> __________________________________________
>>>> Michael Barton, Professor of Anthropology
>>>> School of Human Evolution & Social Change
>>>> Center for Social Dynamics & Complexity
>>>> Arizona State University
>>>> 
>>>> phone: 480-965-6213
>>>> fax: 480-965-7671
>>>> www: http://www.public.asu.edu/~cmbarton
>>>> 
>>>> 
>>>> 
>>>> _______________________________________________
>>>> grassgui mailing list
>>>> grassgui at grass.itc.it
>>>> http://grass.itc.it/mailman/listinfo/grassgui
>>>> 
>>> 
>> 
>> __________________________________________
>> Michael Barton, Professor of Anthropology
>> School of Human Evolution & Social Change
>> Center for Social Dynamics & Complexity
>> Arizona State University
>> 
>> phone: 480-965-6213
>> fax: 480-965-7671
>> www: http://www.public.asu.edu/~cmbarton
>> 
>> 
>> 
>> _______________________________________________
>> grassgui mailing list
>> grassgui at grass.itc.it
>> http://grass.itc.it/mailman/listinfo/grassgui
>> 
> 

__________________________________________
Michael Barton, Professor of Anthropology
School of Human Evolution & Social Change
Center for Social Dynamics & Complexity
Arizona State University

phone: 480-965-6213
fax: 480-965-7671
www: http://www.public.asu.edu/~cmbarton

More information about the grass-gui mailing list