With the release of QGIS 1.7.0 it was announced that a new official plugin repository has been created with many features (a rating system, lists of recommended and recently added plugins, etc.). In addition, a dedicated section for 3rd-party plugins has been created in the QGIS bug tracker, where plugin authors can create a homepage, wiki, bug tracker, and code repository for their plugins. The main goal of all this is to provide a single repository for plugins and a single place to report bugs for both plugins and QGIS itself.
And although the new repository has been put into production and added as the default repository in the QGIS 1.9.90, plugin developers are still holding back on using it (at the time of writing, there are only 35 plugins in the new repository, for comparison, there are 111 plugins in the old repository). This can probably be explained partly by ignorance and partly by the somewhat confusing procedure for adding a plugin to the new repository.
Let’s try to figure this out.
I assume that the plugin code has been published to one of the public code repositories, such as GitHub or BitBucket, and now we want to set up a bug tracker and wiki on hub.qgis.org. This will make life a bit easier for users: they will be able to use their existing account to file bug reports. To do this:
go to hub.qgis.org and log in
open the “Projects” section and click on the “New Project” button, or go to the “User Plugins” subsection and click on the “New Subproject” button
fill in the form (note that in the “Subproject of” list you should select “User plugins”): enter the name of the plugin, a short description, and a link to its homepage (if available). Here, you can also select the necessary components of the bug tracker (wiki, tracker itself, calendar…). Most settings can be changed later
once you have completed the form, click on the ‘Save’ button. That’s it, a new project has been created
The bug tracker and code repository are ready. All we need to do now is upload the plugin to the plugins repository.
Before creating an archive and uploading it to the server, we need to fill in the metadata in the plugin’s metadata.txt file, specifying the URLs of the bugtracker, repository and homepage (more details). Here’s an example:
[general]name=Cool Plugindescription=Does some useful actions with your datacategory=Vectorversion=1.0.0qgisMinimumVersion=1.7.2icon=icons/pluginicon.pngauthorName=usernametags=vector, bounding box, bufferexperimental=Truedeprecated=Falsehomepage=http://someserver.com/coolplugin.htmltracker=http://hub.qgis.org/projects/coolpluginrepository=http://github.com/username/coolplugin
Save the changes in the metadata.txt file and create an archive with the plugin. Now we are ready to upload the plugin to the new repository:
use the “Share a plugin” link in the sidebar to open the upload form
select the plugin archive, activate the “Experimental” checkbox if necessary, and click “Upload”
If the plugin is packaged correctly and the metadata contains no errors, the archive will be uploaded to the server. Otherwise, an error message is displayed, and the upload is cancelled. Once the errors have been corrected, the upload should be repeated.
The uploaded plugin gets an unapproved status and is not immediately available to all users. It will only be added to the list of publicly available plugins after approval from the administrator. This will happen every time you upload a new version, even if your plugin has already been approved. This policy may be changed in the future.
There is also a small script to upload the plugin archive to the server. For example, you can add it to the post-commit hook or Makefile and upload new versions without a browser.
As you can see, there is nothing complicated about using the new infrastructure.
In general, the proposed changes were well received and sparked another wave of discussion about the plugin situation. At the same time, the main drawback of the proposed approach was pointed out — now, after activating a plugin, it is not known in which menu it will appear. To solve this problem, the Plugin Manager had to be modified, and another tag had to be added to the plugin metadata.
The new category metadata is for informational purposes only and should contain the name of the menu where the plugin will be placed after activation. This information is displayed in the Plugin Manager.
Category metadata in the Plugin Manager
If there is no category metadata, the default value “Plugins” is used.
Oh, and one more thing. Plugins written in C++ need to be recompiled. It is a small API break.
It is still the responsibility of the plugin developer to place the plugin in the correct menu, and there are many API calls available for this purpose. These methods can be used to add the plugin to any menu or toolbar.
What’s this all about? Well, about 20 minutes ago, it stopped being an experimental branch and was merged into master. So if you don’t find “Add Delimited Text Layer” in the “Plugins” toolbar tomorrow, don’t say you weren’t warned :-).
I am not going to discover America by saying that the number of different plugins for QGIS is constantly growing (190 at the time of writing) and it is difficult to navigate the “Plugins” menu with a dozen or two active plugins. This issue has been raised many times at hackfests, starting with the Vienna meeting, and on the mailing lists.
Periodically, new tickets were filled: #1602, #1734, #4069. One of the results was moving fTools to the “Vector” menu (the menu is created by the module itself), soon the “native” “Raster” menu was added, and a bit later — the “Database” menu (it appears when the first plugin belonging to this menu is activated). But the vast majority of plugins, including core plugins, continued to huddle together in the “Plugins” menu and place their buttons on the corresponding toolbar.
Meanwhile, in ticket #4395, Paolo again raised the question of moving at least the core plugins into the appropriate menus. This is where it all started.
Firstly, the Offline Editing and SPIT plugins have been moved to the “Database” menu. The “Oracle GeoRaster” and “SQLAnywhere” plugins lost their menu items altogether, and their buttons were moved to the “Layers” toolbar and duplicated in the “Layers” menu. After a brief discussion with Paolo, the “repression” continued:
raster plugins have been moved to the “Raster” menu and toolbar
a “native” “Vector” menu and toolbar have been created for plugins that work with vector data. fTools and a few other core plugins have been moved there
the “Add Delimited Text Layer” plugin has been moved to the “Layers” toolbar
the “GPS Tools” have been split: one button has been moved to the “Layers” toolbar and the other to the “Vector” menu
the “Database” toolbar has been created, and the “Offline Editing” and “SPIT” plugin buttons have been moved there
new methods have been added to allow plugin developers to place their plugins in the desired location
Some plugins have not been moved anywhere yet:
eVis (“Database”?)
OpenStreetMap (a new category “Web” is proposed)
MapServer Export (a new category “Web” is proposed)
RoadGraph (probably “Vector”)
GRASS
Coordinate Capture
Diagram Overlay
So far, all this exists as a reorganise-plugins branch in my fork. The question of merging it has been raised on the mailing list. The main disadvantage (relative in my opinion) - when activating the plugin, the user will not know where exactly, in which of the 4 menus it (the plugin) will appear. One solution is to add a tag to the plugin’s metadata that specifies which section the plugin will be placed in after installation.
Quite often, I have come across the question: How can I use the georeferenced rasters (raster + map file) from OziExplorer in QGIS? In order to use such data in a GIS, it is usually necessary to convert them first, e.g., with GlobalMapper. With the release of GDAL 1.7.1, there is no need for proprietary software, as GDAL now supports OZI map files.
Just three hours ago, a patch by Nathan Woodrow was added to QGIS that allows expression-based feature labelling.
Expressions support basic arithmetic operators, parentheses, string functions, type conversion functions, etc. There is also an expression builder dialogue where you can:
search functions by their names
check expression validity as you type
preview expression results in real-time
read a short description of the selected function (currently only available for some functions)
load the first 10 unique values or all field values from the context menu
Julien Malik has announced the release of a new plugin for QGIS that provides an interface to the Orfeo Toolbox tools. The new plugin is based on the Processing Framework and requires the OTB libraries and Python bindings in the system. For Windows, you can get the necessary files here. For Linux users, it is a bit more complicated, but the author promises to prepare packages for Ubuntu soon.
Over time, the list of available tools will be expanded.
As you know, GDAL 1.9 now supports ESRI ArcObjects (read only) and File Geodatabase (read and write). This means that after rebuilding QGIS with the latest version of GDAL, you will be able to work with these data sources. For Linux users, there is no need to rebuild QGIS; it is enough to make the correct symlinks.
GDAL is gradually moving towards full Unicode support: RFC 23 has already been implemented, and there are ongoing efforts to implement RFC 5.
Another step was the implementation of recoding of Shapefile attributes to UTF-8 when reading and from UTF-8 when writing. Except for that… the encoding is determined by reading the LDID (Language Driver ID) from the DBF header. In general, this is the right approach, but I can’t remember the last time I saw shapfiles with the encoding specified correctly. Mostly, there are files with LDID set to 87, which corresponds to the default value.
This is where the most interesting part begins. It is clear that this default is different for everyone. And in the current implementation, the LDID/87 value is interpreted as ISO8859_1 (Latin-1). The problem with this approach, I think, is clear to everyone. The proposed solution is either to edit the existing files and set the required DBF encoding or to override the interpretation of the LDID value by setting an environment variable. The first method requires more work (it is necessary to find out the encoding of each Shape file and write it in the DBF header), but it is also the most correct. The second one is actually an ugly workaround because, with this approach, only files in one, overridden, encoding will be read correctly. All others will still be displayed as unreadable garbage, which is unacceptable when using data in different encodings.
A few months ago, the question of integrating OSSIM into QGIS was raised. The discussion was interesting and active, the OSSIM developers even published some experimental code, but it all fizzled out.
One of the reasons, in my opinion, is that the developers of the OSSIM data provider published the code not as a normal patch but as a set of modified files. Moreover, all this was done not on the latest master but based on outdated SVN sources. Another reason is that, at the same time, we had an active discussion about a new Processing Framework, and its development started. And the possibility of integrating OSSIM through this framework was also considered.
I didn’t feel like writing a report for my contest work today, and I didn’t feel like working on an Android app either. So I decided to see what the guys from OSSIM had written. In general, it doesn’t look too bad, although there are a few questionable solutions. For example, for some reason, they invented their own class for saving settings, even though there is QgsOptions and a settings dialogue. I have tried to integrate their code into master. The results are in my fork on GitHub, branch ossim_provider. I haven’t tested it myself yet, as I haven’t found a pre-built OSSIM package, and it will take quite some time to build it on my laptop.
Giuseppe Sucameli has successfully completed his GSoC 2011 project — the DB Manager plugin for QGIS.
DB Manager combines the functionality of the PGManager, SLManager and RT_Sql_Layer plugins and supports SQLite/SpatiaLite and PostgreSQL/PostGIS databases (including raster support). With this plugin, you can:
view a list of database tables
view detailed information about the selected table
view data in tabular form and on a map
rename and delete tables using a graphical interface
execute SQL queries
drag and drop database tables (both spatial and geometryless) into your QGIS project
In addition, while working on the module, Giuseppe implemented the ability to import layers from one data source to another (commit 1a70dddca1). This allows you to easily import data from a shapefile into a PostGIS or SpatiaLite database, and vice versa. This functionality is available in both C++ and Python.
Besides Pascal bindings for GDAL, I also have Pascal bindings for Proj.4 — a library for working with coordinate reference systems.
But while in the case of GDAL there were no Pascal bindings at all, the situation with Proj.4 is much more complicated. There are Pascal modules for old Proj versions 4.4.3, 4.4.6 and 4.4.7 (by Alexander Weidauer), which I used as a starting point for my work. The FPC also has a built-in package (based on the newer Proj 4.6.1). Unfortunately, I found out about this package after I had written my module.
So now I’m asking myself, should I make my code public or not?
