Submitting General Notes

Module behaviour: computational region settings

The current region or ​computational region is the actual setting of the region boundaries and the actual raster resolution. It can be considered as region of interest (ROI).

As a general rule in GRASS GIS (new module implementations should follow this!):

• Raster maps are always imported completely at their own resolution (exception: WMS imported layers).
• Vector maps are always imported completely.
• In computations,
  • raster input maps are automatically cropped/padded and rescaled (using nearest neighbour resampling) to match the current region in order to produce the output raster map or to query values.
  • Raster output maps have their bounds and resolution equal to those of the current computational region.
  • Vector maps are always considered completely.

Submitting code

Be sure to develop on top of the latest GRASS GIS code which is in a Git repository on GitHub. See ​CONTRIBUTING file there.

Commit messages

Generally, the commit message (log message) should give an information about what changed in the code and how the change affects the functionality. Additionally, the change of dependencies and changes of functionality of depended code should be discussed if applicable.

The general format of a message is:

module or library: description (possible Trac ticket, merge commit or related commit)

Examples (of single line commit messages):

r.to.vect: corrected x in the crowded message
g.mremove: Changed the interface to that of g.mlist and added exclude= (ticket #2228)
libraster: Added raster name and row info to get/put_row error messages
vlib/pg: column check must be case-insensitive 
wxGUI/lmgr: add measuring of distances also to Layer Manager
wxGUI: workaround for not visible toolbars on Mac with wxPython 3

Include ticket using hash mark and ticket number, e.g. #2228, and another commit (revision) using letter r and revision number, e.g. r60975. This will allow Trac (and perhaps other systems) to create a link to the ticket or commit (or revision). However, do not rely on this and always include the information about what and why the commit is changing and how (consider browsing commit messages in command line).

It is possible and allowed to do multiline commit messages but you should consider the following. First, multiline commit messages should be used to provide further details about the commit. They should not be used to describe large changes of code. Instead, these large changes should be split into the separate commits with shorted commit messages. Note that this not only simplifies writing of good, simple and readable commit messages but it also makes code review and regression testing easier. Second, if you have a lot to say about the commit you should perhaps include this in the comments or documentation (you can refer there to tickets or other commits too in the same way as in commit messages, although they will not be automatically linked).

Write the commit messages in the way that they can be used to create/update change logs, wiki:Release pages and news in general.

Don't include your name (or id) to commit message, this is stored separately and automatically. However, if you are committing someone's code (e.g. path) or you are writing the code together with someone, include his or her name.

Include the module, library or component name as a prefix separated by colon, e.g. libraster:. You don't have to include file names in the commit message, they are managed by SVN itself.

If you are not sure if your style is correct, ask on mailing list.

Some bad examples (of single line commit messages):

r.slope.aspect: fix compilation
(missing information what exactly was broken and reasoning behind the fix; it's clear that we are not trying to break something by the commit)

wxGUI/render: attempt to fix #560
(missing information what is #560 and how we are trying to fix it)

Add tests for Table and Columns classes
(we don't know where the classes belongs to, prefix pygrass: or pythonlib/pygrass would tell us in this case)

fix bug introduced in r60216
(missing information how the bug was fixed and which bug it was)

fix r60216 (i18n)
(it should probably say something like: wxGUI: fix insufficient handling of i18n (introduced in r60216))

libraster:Added raster name and row info to get/put_row error messages
(missing space after colon)

d.histogram launched from map display toolbar doesn't work
(this is description of what is wrong, not how it is fixed)

fixed loading from a file, should I backport it?
(commit messages are not for opening discussions or general communication)

Creating (legacy) diffs

Be sure to create unified (diff -u) format. "Plain" diffs (the default format) are risky, because they will apply without warning to code which has been substantially changed; they are also harder to read than unified.

Such diffs should be made from the top-level directory, e.g. git diff display/d.vect/main.c; that way, the diff will include the pathname rather than just an ambiguous main.c.

SVN Properties

When submitting new files to the repository set SVN properties,

usually for directory

      svn:ignore : *.tmp.html
      		   *OBJ*

or e.g. for C-file

      svn:mime-type : text/x-csrc
      svn:keywords : Author Date Id
      svn:eol-style : native

See ​http://svnbook.red-bean.com/en/1.4/svn.advanced.props.html

To set a property:

      svn propset svn:keywords 'Author Date Id' <file>
      svn propset svn:mime-type text/x-sh grass_shellscript.sh

To edit the svn:ignore property using your default text editor:

      svn propedit svn:ignore <directory>

To set the svn:ignore property non-interactively, first create a file containing the value:

      echo "*.tmp.html" > ignore.txt
      echo "*OBJ*" >> ignore.txt

then use:

      svn propset -F ignore.txt svn:ignore <directory>

List of mime-type:

      C++ files (.cpp): text/x-c++src
      C files (.c): text/x-csrc
      DTD files (.dtd): text/xml-dtd
      GIF files (.gif): image/gif
      Header files (.h): text/x-chdr
      HTML files (.html): text/html
      JPEG files (.jpg): image/jpeg
      Makefiles: text/x-makefile
      PNG files (.png): image/png
      Python files (.py): text/x-python
      Shell scripts (.sh): text/x-sh
      Text files (.txt): text/plain
      XML files (.xml): text/xml

(please update the list...)

For your convenience use the source:grass-addons/tools/module_svn_propset.sh​ script.

SVN Property Id

We don't want the $ID$ in source code any more as it causes problems for the SVN branches.

Comments

PLEASE take the time to add comments throughout your code explaining what the code is doing. It will save a HUGE amount of time and frustration for other programmers that may have to change your code in the future.

End Of Line

Make sure a new line is at the end of each file and UNIX style newlines are used (\n).

Branches and backports

This section applies to the core developers with access to the main repository (other contributors can safely ignore it).

GRASS GIS Subversion repository has trunk and several branches. All the development should happen in trunk (trunk is a "development branch"). All the other branches are usually associated with releases and should contain stable code which is being prepared towards the given release.

When a bug is fixed the fix should be committed to trunk, tested there, and than backported to the release branch or branches. The testing before backport should include user testing, running automated test (if available), and compilation of the whole source tree (ideally after make distclean). Note that thinks like testing should be done also before the original commit to trunk. Also note that not all these steps has to be done manually, you can take an advantage of ​Travis CI or ​automated runs of the GRASS GIS testing suite.

Often there is more than one active stable branch, if you are backporting, make sure you always backport to all the branches between trunk and the furthest branch you are backporting to. For example, let's say we have trunk and branches 7.2 and 7.0, if you backport to 7.0, you have to also backport to 7.2. You can also choose to backport only to the closest branch, in our example 7.2.

Backport only complete fixes. When you are not sure if the fix is complete or if there is an possibility that some details such as wording will change, wait with the backport for a little while and backport all the changes together to reduce number of commits which needs to be reviewed (right now or in the future). You can also backport multiple commits from trunk as one commit if you think it is appropriate.

Include the number of the commit (or commits) you are backporting, into the commit message, for example: less verbose messages (backport r89436). This will help matching the file content in between branches and tracking if the commits were backported.

As a developer you should maintain a list of commits which you plan to backport. One way how to do it is to subscribe to the ​grass-commit mailing list and filter your commits in your email.

Do not add svn merge property to the commit. Review tools directory in Addons repository for scripts which will help you with backporting.

Makefiles

When writing Makefiles, use the current standard.

If you have to use commands, please check for:

            avoid     | use instead
    ------------------+---------------
    make target       | $(MAKE) target
    mkdir target      | $(MKDIR) target
    cp  (executable)  | $(INSTALL) -m 755 file target
    cp  (normal file) | $(INSTALL) -m 644 file target
    ar                | $(AR)

rm: be VERY careful with recursive remove. Also beware of removing $(FOO)* if $(FOO) has any chance of being empty.

Examples: see below examples or others

source:grass/trunk/raster/r.info/Makefile​
source:grass/trunk/vector/v.edit/Makefile​

If you are unsure, please ask on the GRASS Developers list.

AutoConf

If you need to add support for a different library in the 'configure' script, you should first seek consent in the grass-dev mailing list (see below), then you need to expand 'configure.in' and run subsequently autoconf-2.13 (later versions will not work) to re-generate 'configure'.

Naming Conventions

Have a look at source:grass/trunk/INSTALL​

For consistency, use README rather than README.txt for any README files.

Variables

GRASS/Environment variables:

If you add a new variable, please follow the naming convention. All variables are described in source:grass/trunk/lib/init/variables.html​

Modules

Try to use module names which describe shortly the intended purpose of the module.

The first letters for module name should be:

	d. 	- display commands
	db. 	- database commands
	g. 	- general GIS management commands
	i. 	- imagery commands
	m.	- miscellaneous tool commands
	ps. 	- postscript commands
	r. 	- raster commands
	r3. 	- raster3D commands
	v. 	- vector commands

Some additional naming conventions

• export modules: (type).out.(format) eg: r.out.arc, v.out.ascii
• import module: (type).in.(format) eg: r.in.arc, v.in.ascii
• conversion modules: (type).to.(type) eg: r.to.vect, v.to.rast, r3.to.rast

Avoid module names with more than two dots in the name. Example: instead of r.to.rast3.elev use r.to.rast3elev

Code Quality

Follow the best writing practices specified by GRASS submitting rules for a given language.

Write tests for your code. See the testing guide or existing examples (g.list).

Contact us

Tell the other developers about the new code using the following e-mail:

grass-dev@…

To subscribe to this mailing list, see ​http://lists.osgeo.org/mailman/listinfo/grass-dev

In case of questions feel free to contact the developers at the above mailing list.

​http://grass.osgeo.org/development/