[cfe-dev] [PATCH] LibTooling docs

Manuel Klimek klimek at google.com
Fri Apr 20 09:40:43 PDT 2012 

On Fri, Apr 20, 2012 at 6:36 PM, Erik Verbruggen <erikjv at me.com> wrote:
> I re-read Chris' tutorials, and they are all active. So ignore that. :-) All your changes look fine!
>
> Off the record: I am also a non-native English speaker (as I assume you are, correct me if I am wrong), but I learned that apparently "basically" is abused so much that some people will question if you really know what you are talking about :-/ And unfortunately, in my experience, I am afraid they are right. I do not doubt you knowledge, but it is something I had to learn "the hard way".

Yea, the basically was basically crap :) I'm also non-native, so I'm
always happy to get different opinions :)

>
> -- Erik
>
>
> On 20 apr. 2012, at 10:30, Manuel Klimek <klimek at google.com> wrote:
>
>> On Thu, Apr 19, 2012 at 5:42 PM, Erik Verbruggen <erikjv at me.com> wrote:
>>>
>>> On Apr 19, 2012, at 16:36, Manuel Klimek wrote:
>>>
>>>> Please find a first version attached.
>>>>
>>>> The next step is a short intro on how to write a FrontendAction, which
>>>> is common to clang plugins and libtooling, and thus I thought I'd put
>>>> it into a separate doc.
>>>
>>>
>>> Comments inline
>>>
>>> Cheers,
>>> Erik.
>>>
>>>
>>>> LibTooling
>>>>
>>>> LibTooling is a library to support writing standalone tools based on Clang. This document will provide a basic walkthrough of how to write a tool using LibTooling.
>>>>
>>>> Introduction
>>>>
>>>> Tools built with LibTooling, like Clang Plugins, run FrontendActions over code. In this tutorial, we'll demonstrate the different ways of
>>>
>>> Don't use "we". "This tutorial demonstrates ..." is better. Same for constructs like "we'll": in colloquial communication that is okay, but otherwise you should use "we will". The only kind-of-exception is "don't" or any use of "n't" for "not".
>>
>> I disagree with the idea that we need to be overly formal in those
>> tutorials. Now, if you think that some of my bad grammar makes the
>> sentences hard to understand / read, I'm happy to change, but I don't
>> think that a style that's more dry makes it easier to read...
>>
>>>> running clang's SyntaxOnlyAction, which basically runs a quick syntax check, over a bunch of code.
>>>
>>> Try not to use "basically", as it might sound like you just don't know what really happens. Now in all honesty, I have no better way of phrasing this sentence...
>>
>> Done (just removed the "basically" :)
>>
>>>> Parsing a code snippet in memory...
>>>
>>> Ellipsis? Why not a dot/full-stop like in other headers?
>>
>> Done.
>>
>>>> If you ever wanted to run a FrontendAction over some sample code, for example to unit test parts of the Clang AST, runToolOnCode is what you looked for. Let me give you an example:
>>>>
>>>>   #include "clang/Tooling/Tooling.h"
>>>>
>>>>   TEST(runToolOnCode, CanSyntaxCheckCode) {
>>>>     // runToolOnCode returns whether the action was correctly run over the
>>>>     // given code.
>>>>     EXPECT_TRUE(runToolOnCode(new clang::SyntaxOnlyAction, "class X {};"));
>>>>   }
>>>>
>>>> Writing a standalone tool.
>>>>
>>>> Once you unit tested your FrontendAction to the point where it cannot possibly break, it's time to create a standalone tool. For a
>>>
>>> "it is"
>>>
>>>> standalone tool to run clang, it first needs to figure out what command line arguments it needs for a specified file. To that end we want to create a CompilationDatabase.
>>>
>>> ".. we want to ..." could be read as one-of-the-ways-to-do-it. Is it? Otherwise just state: "To that end a CompilationDatabase is needed."
>>
>> I try to avoid passive voice wherever possible. Changed to "To that
>> end we create a CompilationDatabase."
>>
>>>> Creating a compilation database.
>>>>
>>>> CompilationDatabase provides static factory functions to help with parsing compile commands from a build directory or the command line. To allow both explicit specification of a compile command line, as well as retrieving the compile command lines from a database, we can write:
>>>
>>> "To allow....can write:" -> "The following code allows for both explicit specification of a compile command line, as well as retrieving the compile commands lines from a database."
>>
>> Done.
>>
>>>> int main(int argc, const char **argv) {
>>>>   // First, try to create a fixed compile command database from the command line
>>>>   // arguments.
>>>>   llvm::OwningPtr<CompilationDatabase> Compilations(
>>>>     FixedCompilationDatabase::loadFromCommandLine(argc, argv));
>>>>
>>>>   // Next, use normal llvm command line parsing to get the tool specific
>>>>   // parameters.
>>>>   cl::ParseCommandLineOptions(argc, argv);
>>>>
>>>>   if (!Compilations) {
>>>>     // In case the user did not specify the compile command line via positional
>>>>     // command line arguments after "--", try to load the compile commands from
>>>>     // a database in the specified build directory.
>>>>     std::string ErrorMessage;
>>>>     Compilations.reset(CompilationDatabase::loadFromDirectory(BuildPath,
>>>>                                                               ErrorMessage));
>>>>
>>>>     // If there is still no valid compile command database, we don't know how
>>>>     // to run the tool.
>>>>     if (!Compilations)
>>>>       llvm::report_fatal_error(ErrorMessage);
>>>>   }
>>>> ...
>>>> }
>>>>
>>>> Creating and running a ClangTool.
>>>>
>>>> One we have a CompilationDatabase, we can create a ClangTool and run our FrontendAction over some code. For example, to run the
>>>
>>> OnCe (note the C :-) ). But better: "Once the CompilationDatabase is created, ...."
>>
>> Again, my strong preference for active voice trumps the idea of more formality.
>>
>>>> SyntaxOnlyAction over the files "a.cc" and "b.cc" one would write:
>>>
>>> "... one would write:" -> "the following code-snippet can be used:"
>>>
>>>>   // A clang tool can run over a number of sources in the same process...
>>>>   std::vector<std::string> Sources;
>>>>   Sources.push_back("a.cc");
>>>>   Sources.push_back("b.cc");
>>>>
>>>>   // We hand the CompilationDatabase we created and the sources to run over into
>>>>   // the tool constructor.
>>>>   ClangTool Tool(*Compilations, Sources);
>>>>
>>>>   // The ClangTool needs a new FrontendAction for each translation unit we run
>>>>   // on. Thus, it takes a FrontendActionFactory as parameter. To create a
>>>>   // FrontendActionFactory from a given FrontendAction type, we call
>>>>   // newFrontendActionFactory<clang::SyntaxOnlyAction>().
>>>>   int result = Tool.run(newFrontendActionFactory<clang::SyntaxOnlyAction>());
>>>>
>>>> Putting it together - the first tool.
>>>>
>>>> Now we combine the two previous steps into our first real tool. This example tool is also checked into the clang tree at tools/clang-check/ClangCheck.cpp.
>>>
>>> "To combine the two..."
>>
>> Not sure where this would go?
>>
>>>>   #include "llvm/Support/CommandLine.h"
>>>>   #include "clang/Frontend/FrontendActions.h"
>>>>   #include "clang/Tooling/CompilationDatabase.h"
>>>>   #include "clang/Tooling/Tooling.h"
>>>>
>>>>   using namespace clang::tooling;
>>>>   using namespace llvm;
>>>>
>>>>   cl::opt<std::string> BuildPath(
>>>>     cl::Positional,
>>>>     cl::desc("<build-path>"));
>>>>
>>>>   cl::list<std::string> SourcePaths(
>>>>     cl::Positional,
>>>>     cl::desc("<source0> [... <sourceN>]"),
>>>>     cl::OneOrMore);
>>>>
>>>>   int main(int argc, const char **argv) {
>>>>     llvm::OwningPtr<CompilationDatabase> Compilations(
>>>>       FixedCompilationDatabase::loadFromCommandLine(argc, argv));
>>>>     cl::ParseCommandLineOptions(argc, argv);
>>>>     if (!Compilations) {
>>>>       std::string ErrorMessage;
>>>>       Compilations.reset(CompilationDatabase::loadFromDirectory(BuildPath,
>>>>                                                                 ErrorMessage));
>>>>       if (!Compilations)
>>>>         llvm::report_fatal_error(ErrorMessage);
>>>>     }
>>>>     ClangTool Tool(*Compilations, SourcePaths);
>>>>     return Tool.run(newFrontendActionFactory<clang::SyntaxOnlyAction>());
>>>>   }
>>>>
>>>> Running the tool on some code.
>>>>
>>>> When you check out and build clang, clang-check is already built and available to you in bin/clang-check inside your build directory.
>>>>
>>>> You can run try clang-check on a file in the llvm repository by specifying all the needed parameters after a "--" separator:
>>>>
>>>>   $ cd /path/to/source/llvm
>>>>   $ export BD=/path/to/build/llvm
>>>>   $ $BD/bin/clang-check . tools/clang/tools/clang-check/ClangCheck.cpp -- \
>>>>     clang++ -D__STDC_CONSTANT_MACROS -D__STDC_LIMIT_MACROS \
>>>>     -Itools/clang/include -I$BD/include -Iinclude -Itools/clang/lib/Headers -c
>>>>
>>>> As an alternative, you can also configure cmake to output a compile command database into its build directory:
>>>>
>>>>   # Alternatively to calling cmake, use ccmake, toggle to advanced mode and
>>>>   # set the parameter CMAKE_EXPORT_COMPILE_COMMANDS from the UI.
>>>>   $ cmake -DCMAKE_EXPORT_COMPILE_COMMANDS=ON .
>>>>
>>>> This creates a file called compile_commands.json in the build directory. Now you can run clang-check over files in the project by specifying the build path as first argument and some source files as further positional arguments:
>>>>
>>>>   $ cd /path/to/source/llvm
>>>>   $ export BD=/path/to/build/llvm
>>>>   $ $BD/bin/clang-check $BD tools/clang/tools/clang-check/ClangCheck.cpp
>>
>> Thanks for the review!
>>
>> /Manuel
>> <LibTooling.html>

More information about the cfe-dev mailing list