Hi Peppe, Your explanation is clear.
I tend to be on the same opinion as Jukka on this topic because I generally use OpenJUMP as a toolbox, and I generally know exactly what I want to do with my data. But I admit that to visualize heterogeneous data, OpenJUMP has not much to offer to the user to solve projection problems, and the beginner can be bothered by the lack of assistance. Here are a few recommandation : 1 - Projection issues may be tricky. It is magic as long as the only need is visualization, but if the user need to reproject his dataset, he must be aware of the consequences (reversibility, topology consistency...). Last time I have been screwed by a projection problem is with FME. I imported shapefiles with a prj in a project using the "same" projection. It was supposed to be a no-op (doing nothing), except that FME did a transformation from projection A (defined by prj parameters) to projection A (defined by internal FME parameters), which resulted in an invisible switch of a few micrometers difficult to see, but which broke the consistency with another layer (which did not follow the same process). Of course this can be avoided in FME, but this is just an example to illustrate that without a great care, something supposed to be magic may become dramatic. 2 - From my point of view, one of the most difficult problem is to be able to recognize that two coordinate reference system with different origins (different registries, different formats, different libraries, different definitions) represent the same thing (see the above problem with FME). I think you already worked on that problem. 3 - Your mail explains quite clearly what already exists and where you want to go. I think that to anticipate difficulties, we can suppose that a SRID is associated to the task and try to define OpenJUMP behaviour in different situations : - default behaviour when creating a new task : asking for a srid or not ? it is a good thing if OJ can infer information from prj files or other sources, but I don't like having to answer esoteric questions before I can start working. - task without srid : does it take the srid of the first layer imported ? What if layers without srid are already imported ? - can we change the srid of a task if layers with srid are already imported ? - importing a layer with a different srid : 1) the layer is just tagged (layer srid mismatch task srid), 2) the layer is automatically reprojected by the renderer ? 3) the user is invited to reproject the layer ? 4) There are some options to define OpenJUMP behaviour - how to deal with layers without projection : can we import them in a task with a srid ? can we edit them ? Do we set the task projection to the layer projection automatically ? - if a reprojected layer is not editable, an interesting option would be to set the task srid to the selected layer srid (-> makes the selected layer editable, and reproject other layers) - etc. 4- Implementation : no real opinion. Ede's advice will certainly make the code more flexible, but also a bit more complex. And how to represent the coordinate system property ? Another difficult question. We already have SRID represented by an int at the geometry level (JTS) and a CoordinateSystem at the FeatureSchema level. IMHO, the first is a bit too lightweight (cannot handle non EPSG crs). The second is too lightweight if we want to use it to effectively transform coordinates (cannot handle much transformations) and too heavyweight if we just use it as a reference to be used by CTS library (or any other). My 3 cents Michael Le 03/08/2016 à 15:13, Gmail a écrit : > LoopThis thread needs a larger explanation. > I try to simplify it. > other GIS like Kosmo or GVSig implemented Coordinate system framework > following these steps: > a) first step they add a projection object to the task (usually as EPSG or > ESRI code). In Kosmo user has to set that. QGIS also allows to set Task > projection loading that from the first loadedf file (with SRID). > b) QGIS define the Unit of the task from SRID ( ex. 4326>degree, > 32632>metre) while GvSig And Kosmo require to set it manually. > c) a projection object is set to each loaded layer. This is done reading > layer metadata or manually > d) if the task and the layer projection object are different a > transformation should be set. Those software use ( for vector) proj4 > libraries. In this step Qgis and newer gvsig allows on fly reprojection. > e) this transformation is taken into account only by layer renderes on the > workbench. Which changes geometry before drawing it. > This transformation is saved into project file and taken into account > whenever the project file is loaded. > f) Note that Kosmo (and probably Gvsig) doesn't allow any spatial operation > on reprojected layers. The only way to modify them is to save them > reprojected. > > Recently I did few modifications on shape file reading in order to expand > capability to set layer SRId when reading file. Layer properties plugins > already have this capability for both raster and vector ( included > geotif*). Together with database and wfs capability to record layer srid we > probably get almost point C of my list. > > My idea is to work on point A and B, integrating parts if my measure plugin > in to OJ core in order to have measurements\zoom when task projection is > geographic or possibility that oj display meter or feet unit on > measurements \ scale bar. > The other points can be faced in the future, including in fly reprojection. > > My project: > 1) Oj already as a srid registry embedded that I added when I defined srid > detection capability from auxiliary files. It is a simple list of > projection, a series of lines with only srid number and a proj. description > ( ex <32632>;<WGS 84 UTM zone 32>), build using proj4 registries and excel. > I could expand each line with unit ( ex <32632>;<WGS 84 UTM zone 32>;<metre>) > 2) expand Task class with srid code and unit. User can define manually . > 3) modify measure /zoom plugins according units, meter, foot, degree ( in > this last case I would limit only to wgs84 ) using classes fro my measure > plugin. > > Peppe > > > > > > * > > > > Inviato con AquaMail per Android > http://www.aqua-mail.com > > > Il 03 agosto 2016 12:32:48 edgar.sol...@web.de ha scritto: > >> hey Peppe, >> >> On 03.08.2016 11:11, Giuseppe Aruta wrote: >>> Hi all, >>> The title explains what is my idea. In a possible future we can extend OJ >>> projection capabilities. And the 1st step I would explore is to add SRID >>> code to a task (to centralize possible transformations) >> can you elaborate? >> >>> and unit of measurements (retriving from SRID, which will affect other >>> plugins/tools like measure tools, measure area/length, display scales etc, >>> especially for Geographic coordinate systems). >> same here. >> >>> I gave a look at Task class , should I implement (srid and unit) as >>> properties into the associate xml file? Does it breaks compatibility? >> ..ede >> >> ------------------------------------------------------------------------------ >> _______________________________________________ >> Jump-pilot-devel mailing list >> Jump-pilot-devel@lists.sourceforge.net >> https://lists.sourceforge.net/lists/listinfo/jump-pilot-devel > > > ------------------------------------------------------------------------------ > _______________________________________________ > Jump-pilot-devel mailing list > Jump-pilot-devel@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/jump-pilot-devel > ------------------------------------------------------------------------------ _______________________________________________ Jump-pilot-devel mailing list Jump-pilot-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/jump-pilot-devel
