r.in.wms2 fails to install on 6.x

[hamish]

Hi,

the r.in.wms2 addon fails to build on 6.4 and devbr6 from g.extension(s).

The trouble seems to be in the Makefile, the helper scripts don't get installed. It relies on Python.make which is badly out of date in 6.x (still refers to SWIG), which may not be the main trouble, but it certainly isn't helping. Once that is fixed there may be more grass7->6 porting needed.

It is still somewhat unknown to me which python scripts need .bat wrappers & presence in the PATH to work in grass6.

currently (amazingly?) r.in.wms[.sh] run from the command line is the only working option on wingrass. see #820, #1163.

thanks, Hamish

[hamish]

see also #1555, #1196 and #1768.

[hamish]

Hi,

I've now done a bit more on this in grass6 addons svn. It now installs with 'make make MODULE_TOPDIR=$GISBASE' or similar from the source dir; I haven't tried either of the g.extensions yet though.

Also I did a bit more on the port to G6, the module runs now using the default settings, but I didn't test it heavily, just Richard's neowms example using the default download method

r.in.wms2.py url="http://neowms.sci.gsfc.nasa.gov/wms/wms?" \
  out=solar.8day layer=CERES_INSOL_E format=png

Probably a number of fixes there should be considered for porting back to trunk.

Hamish

[turek]

Hi Hamish,

thanks for the fixes. I have ported back to G7 most of them (r56991, r56992 and r56993).

So far the name of the o flag remains. In description of s flag there is written: Skip requests for fully transparent data

I am not sure if it is correct description. If this flag is not present, the data are fetched also with transparent layer (if available). Otherwise you can use bgcolor param with this flag and specify color for these transparent areas. The o was chosen as abbreviation for opaque.

Spaces between options are still kept, it seems more clear to me than one block of ~20 options. But it also can be changed.

Stepan

[hamish]

Replying to turek:

Hi Hamish,

thanks for the fixes. I have ported back to G7 most of them

I hope you found them useful, if they are bad you don't have to listen to me, it's your module after all. :)

So far the name of the o flag remains. In description of s flag there is written: Skip requests for fully transparent data

I am not sure if it is correct description. If this flag is not present, the data are fetched also with transparent layer (if available). Otherwise you can use bgcolor param with this flag and specify color for these transparent areas.

I misunderstood it then, in one of the TWMS/r.out.kml script I have somewhere it checks if the returned tile is all-white and skips warping/mosaicking that so that part of the background becomes NULL not 255. Is the idea to make a complete rectangle instead of missing tiles in the corners? I was wondering how the bgcolor option fit in. If that supports standard GRASS colors you might consider adding:

  #% gisprompt: old,color_none,color

so from the module GUI the user gets the color-picking tool.

The o was chosen as abbreviation for opaque.

I prefer to avoid -o,-q,-v as option letters if there are others available, since it is easy for users to confuse them with --o,--q,--v. fwiw I also like to avoid -l, -1, and -I, but for certain-font reasons.

Spaces between options are still kept, it seems more clear to me than one block of ~20 options. But it also can be changed.

shrug, author's choice; the convention elsewhere is compressed but it doesn't change the running of the script at all so just cosmetic, and I agree it gets a bit hard to read when there are many options. Don't fear the whitespace. :)

a couple observations/wishes:

• wx: switching from 1.1.1 to 1.3.0 clears selection and resizes window height?
• wx: how about being able to filter the list like the location wizard has for projection type Search? (often the servers have 50+ layers to search)
• support for png8 image format? (like qgis has)
• I get an error when BGCOLOR is unset, does the spec require it or just a picky WMS server?

{{{ <ServiceException> BGCOLOR incorrectly specified (0xRRGGBB format expected) </ServiceException></ServiceExceptionReport> }}}

• module gui: Interpolation type isn't part of the server request. -> Optional gui tab?

Also, I'm seeing a unicode error vs. layer Title string in the 6.4.3svn's r.in.wms1's wxGUI wrapper. It's not to do with r.in.wms2.py, so I'll post that to the -dev ML.

regards, Hamish

ps- the WMS/WFS I'm using to test with:

​http://wiki.openstreetmap.org/wiki/New_Zealand/Imagery

[turek]

Replying to hamish:

Hi Hamish, I am sorry for late response.

Replying to turek:

thanks for the fixes. I have ported back to G7 most of them

I hope you found them useful, if they are bad you don't have to listen to me, it's your module after all. :)

Yes, they are useful!

I prefer to avoid -o,-q,-v as option letters if there are others available, since it is easy for users to confuse them with --o,--q,--v. fwiw I also like to avoid -l, -1, and -I, but for certain-font reasons.

That is good point. Do you have any other idea for name of this flag or just changed it to -s?

a couple observations/wishes:

• wx: switching from 1.1.1 to 1.3.0 clears selection and resizes window height?

The selection is not cleared, there exists independent panels for every web service because it is not guaranteed that capabilities file will be same for WMS 1.3.0 and 1.1.1 (if you select something it stays selected in the web service panel). Ideal solution would be existence of one panel, which would adapt according to set web service. But still it should probably behave same way. The resize problem should be fixed in r57025.

• wx: how about being able to filter the list like the location wizard has for projection type Search? (often the servers have 50+ layers to search)

It is good idea. Please add it to: ​https://trac.osgeo.org/grass/wiki/WebServices. There it should not be forgotten. Also you can add there the #% gisprompt: old,color_none,color idea.

• support for png8 image format? (like qgis has)

Done in r57024, r57023. It need to be tested.

• I get an error when BGCOLOR is unset, does the spec require it or just a picky WMS server?

Should be fixed in r57022.

• module gui: Interpolation type isn't part of the server request. -> Optional gui tab?

Good idea! Done in r57027 and r57026.

Best, Stepan

[hamish]

Hi,

I will try to test the new changes tomorrow, some quick comments though:

opaque ~ solid, so -s isn't too far away. I'm still trying to put what that flag does into better words, but not having much luck.

in devbr6 I just added r57028 I added ".decode('utf8')" to the WMS layer title-reading code to avoid a traceback*. I'm not sure if it will fill the gui window with ugly \u1234 text or if it will display them ok. do you know a utf8 enabled non-english wms server to test with? (works ok in trunk)

[*] traceback:

Re: [GRASS-dev] [GRASS GIS] #2010: r.in.wms2 fails to install on 6.x
grass-dev ML 2013/7/5
...
> return _gizmos.TreeListCtrl_SetItemText(*args, **kwargs)
> UnicodeDecodeError
> :
> 'ascii' codec can't decode byte 0xe2 in position 36: ordinal
> not in range(128)

thanks, Hamish

[hamish]

Hi,

some notes & observations:

• png8 support missing for WMS 1.3.0

• color picker greyed out in the "Add web service layer" WMS gui.

• loss of region precision in the saved map,

computational region:

g.region -p
north:      4941600
south:      4848400
west:       1343900
east:       1436600
rows:       1864
cols:       1854

right click on WMS layer, save from computational region. new raster:

r.info 
 |   Rows:         1861 
 |   Columns:      1852  
 |        N: 4941600.00001201    S: 4848400.00000474   Res:    50     |
 |        E: 1436600.00000344    W: 1343899.99999621   Res:    50     |

• WFS mixes in WMS layers, and has lot of traceback parse errors, but realistic data plots; can't query it in the map display window though. is WFS supposed to work yet or is it still a work in progress?

thanks, Hamish

[turek]

Hi,

Replying to hamish:

• png8 support missing for WMS 1.3.0

In capabilties file of WMS 1.1.1 there are two formats: image/png; mode=8bit and image/png8. In WMS 1.3.0 there is only image/png; mode=8bit format. I am not sure which one to use. Currently is used image/png8 therefore you can not see png8 in WMS 1.3.0. Maybe it should be used image/png; mode=8bit, because WMS 1.3.0 is more restrictive regarding formats. It requires only MIME types formats. WMS 1.1.1. is more benevolent. It could mean that image/png; mode=8bit is MIME type and the other is not. Or it just changes randomly from server to server?

• color picker greyed out in the "Add web service layer" WMS gui.

That is not very user friendly. If you check do not request transparent data, the button will be activated. Should be the tooltip added there or the button should be hidden if the choice is not checked? Or just let it activated also when it is unchecked?

• loss of region precision in the saved map,

Probably it is caused by reprojection (using gdalwarp).

• WFS mixes in WMS layers, and has lot of traceback parse errors, but realistic data plots; can't query it in the map display window though. is WFS supposed to work yet or is it still a work in progress?

WFS is not supported, parser just tries interpret capabilities file somehow. The parser is written in tolerant way, so it does not enforce standards strictly, instead of this it tries to interpret everything what it can understood.

Best Stepan

[martinl]

Remove Milestone from Addons bugreports.

[sbl]

Is this still relevant, or can we close it? Anyone planning to work on it?