This plugin use clang for accurately completing C and C++ code.
How-to use it? Simply put it on ~/.vim/plugin
It should work immediately.
Configuration:
- g:clang_complete_auto:
if equal to 1, automatically complete after ->, ., ::
Default: 1
- g:clang_complete_copen:
if equal to 1, open quickfix window on error.
Default: 0
- g:clang_hl_errors:
if equal to 1, it will highlight the warnings and errors the
same way clang does it.
Default: 1
- g:clang_periodic_quickfix:
if equal to 1, it will periodically update the quickfix window
Note: You could use the g:ClangUpdateQuickFix() to do the same
with a mapping.
Default: 0
- g:clang_snippets:
if equal to 1, it will do some snippets magic after a ( or a ,
inside function call. Not currently fully working.
Default: 0
- g:clang_conceal_snippets:
if equal to 1, vim will use vim 7.3 conceal feature to hide <#
and #> which delimit a snippets.
Note: See concealcursor and conceallevel for conceal configuration.
Default: 1 (0 if conceal not available)
- g:clang_exec:
Name or path of clang executable.
Note: Use this if clang has a non-standard name, or isn't in the
path.
Default: 'clang'
- g:clang_user_options:
Option added at the end of clang command. Useful if you want to
filter the result, or if you want to ignore the error code
returned by clang: on error, the completion is not shown.
Default: ''
Example: '|| exit 0' (it will discard clang return value)
- g:clang_use_library:
Instead of calling the clang/clang++ tool use libclang directly. This
gives access to many more clang features. Furthermore it automatically
caches all includes in memory. Updates after changes in the same file
will therefore be a lot faster.
Default : 0
- g:clang_library_path:
If libclang.[dll/so/dylib] is not in your library search path, set
this to the absolute path where libclang is available.
Default : ''
Thanks to:
Xaizek
Note:
- This plugin is incompatible with omnicppcomplete due to the
unconditionnaly set mapping done by omnicppcomplete. So don't forget to
suppress it before using this plugin.
- If you find that completion is slow, please read the PCH section
below.
- The LICENSE file does not cover the files that come from the LLVM
project, namely, cindex.py and __init__.py, which are covered by the
LLVM license.
- Because libclang provides a lot of information about completion, there
are some additional kinds of completion along with standard ones (see
':help complete-items' for details):
'+' - constructor
'~' - destructor
'e' - enumerator constant
'a' - parameter ('a' from "argument") of a function, method or template
'u' - unknown or buildin type (int, float, ...)
'n' - namespace or its alias
'p' - template ('p' from "pattern")
PCH:
In case you can not or you do not want to install libclang, a precompiled header
file is another way to accelerate compilation, and so, to accelerate the
completion. It is however more complicated to install and is still slower
than the use of libclang.
Here is how to create the <vector> pch,
on linux (OSX users may use -fnext-runtime instead of -fgnu-runtime):
clang -x c++-header /path/to/c++/vector -fno-exceptions -fgnu-runtime -o vector.pch
You just have to insert it into your .clang_complete:
echo '-include-pch /path/to/vector.pch -fgnu-runtime' >> .clang_complete
One of the major problem is that you cannot include more that one pch,
the solution is to put the system headers or non changing headers
into another header and then compile it to pch:
echo '#include <iostream>\n#include <vector>' > pchheader.h
clang -x c++-header ./pchheader.h -fno-exceptions -fnu-runtime -o ./pchheader.pch
And then add it to the .clang_complete file.
Known Issues:
- If clang is not able to compile your file, it cannot complete
anything. Since clang is not supporting every C++0x features, this is
normal if it can do any completion on C++0x file.
- There is no difference in clang's output between private
methods/members and public ones. Which means that I cannot filter
private methods on the completion list.