RFC 69: C/C++ Code Formatting¶
This document proposes and describes desired code formatting style used across C and C++ source code in GDAL.
This RFC is based on GEOS RFC 4 by Mateusz Łoskot.
TODO: Give examples of formatting.
The document proposes and describes desired default code formatting style guidelines for GDAL programming in C and C++ languages.
The goal of this document is to initiate process to reach an agreement for the default code formatting style.
There is a need to decide on format of GDAL source code and apply such globally consistent format to GDAL C/C++ codebase.
A uniform, codebase-wide formatting style makes reading and comprehending existing code easier, writing code focused on important aspects of new developments and more pleasant, removes burden during a patch or pull request code reviews and prevents bikeshedding religious arguments. Even in small projects, contributing developers discover the problems of working without an agreed upon code format.
The utility of such guidelines has been proven by many open source software projects.
The scope of the proposal is specifically limited to formatting style guidelines. It is not an intention to develop a general coding guide covering other aspects of writing software like naming, etc.
It is important to make effortless for developers to produce properly formatted code.
The proposal suggests to use clang-format version 3.8 or higher to define C++ code formatting rules for GDAL code.
clang-format is a tool to automatically format C/C++ code, so
that developers don’t need to worry about style issues. Unlike other
tools which use own parsers,
clang-format uses the Clang tokenizer
and supports the same C++ source code as the Clang compiler. This
guarantees correct output is produced and offers unique features (eg.
wrapping long lines whether of code, strings, arrays - something which
AStyle has no way of doing).
The style settings can be defined in a
file, however to make it as easy as possible, we will use the default
style (LLVM style?).
clang-format is straightforward to run and can support
development workflow as standalone tool or as one of many editor
integrations or other bespoke utilities (eg.
git cl format
No automation of code reformatting is proposed. It would be treating the symptoms, no cause: developers not following the code formatting standard.
Although no means to enforce the default formatting style are proposed, currently used CI services (eg. Travis CI) may be employed as a post-commit safety valve - a clang-format lint failure as a compile break (e.g. clang_format.py build script used by MongoDB). Alternatively, a gatekeeper may be installed in SVN/Git, rejecting commits with code not conforming to the code formatting style.
Code Formatting Rules¶
What code formatting rules to use?
“A mature engineers know that a standard is more important than which standard.” ~[MongoDB]
clang-format offers several defaults (eg. LLVM, Mozilla, Linux,
Google C++ Style).
The proposal recommends to use one of the base styles without any modification. It is possible to fine-tune the configuration, but this RFC aims for simplicity.
The reasons are two-fold:
make GDAL code unified with the wide spectrum of well-established C/C++ projects
long arguments and religious wars prevention.
The hope is to avoid requiring a .clang-format file in the code base.
EditorConfig is currently in use and
.editorconfig file is provided to automatically tell popular code
editors about the basic style settings like indentation, whitespaces and
end-of-line markers for distinguished types of plain text files.
.editorconfig file will have to be updated to match the chosen
.clang-format settings if they are required.
clang-format does not enforce line endings.
The EOL marker is considered to be a part of a file encoding decision and not part of any coding style.
The EOL marker can be enforced as project-wide setting controlled with
However, it shall still be left as configurable setting in developer’s
environment of choice (eg.
git config) independently from the
What to do about the existing code?
The proposal recommends to just do one big reformat of the codebase.
While it may seem causing clutter in the repository log (eg.
svn blame), if it occurs infrequently (eg. yearly) and is applied to
the entire codebase at that time, it should not be very disruptive to
the source code history. One way to cope with skewed history is to use
git blame -w which ignores whitespace when comparing commits.
Partial application of the code formatting rules would create more work without delivering the full benefit [MongoDB] leading to codebase with different styles mixed.
Branches to run the big reformat in are:
After Big Reformat¶
How to work against the natural entropy in a codebase:
It is highly recommended to use
clang-formatintegration while writing a code.
Format changed code before committing or opening pull requests.
If you have to commit change in code formatting, do it in separate commit. Avoid commits with a mixture of code and formatting changes.
There is downside of history clutter in repository, but this proposal states that a codebase with different styles across is even worse.
“After all, every moment of time wasted on code formatting or discussion thereof is eliminated.” ~[MongoDB]
Set up Travis CI “style safety valve” build dedicated to run
clang-format lint based on the approach used in
script by MongoDB.
Those who build GDAL with GCC 6+ may appreciate consistent code format style as it will help to avoid some dozens of the new compiler warnings:
src/geom/Polygon.cpp: In member function ‘virtual int geos::geom::Polygon::getCoordinateDimension() const’: src/geom/Polygon.cpp:154:5: warning: this ‘if’ clause does not guard... [-Wmisleading-indentation] if( shell != NULL ) ^~ src/geom/Polygon.cpp:157:2: note: ...this statement, but the latter is misleadingly indented as if it is guarded by the ‘if’ size_t nholes=holes->size(); ^~~~~~