C++ Library containing a lot of needful things: Stack Trace, Command Line Parser, Resource Handling, Configuration Files, Unix Command Execution, Directories, Regular Expressions, Tokenizer, Function Trace, Standard Extensions.
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
 
 
 
 
 

523 lines
20 KiB

/** @file
$Id$
$Date$
$Author$
@copy © Marc Wäckerlin
@license LGPL, see file <a href="license.html">COPYING</a>
$Log$
Revision 1.8 2005/11/29 12:39:42 marc
make it compilable with gcc 4.0.2 and newer doxygen
Revision 1.7 2005/04/20 18:12:55 marc
added kill() for PartialExec
Revision 1.6 2005/04/19 18:48:00 marc
new feature PartialExec
Revision 1.5 2004/12/14 20:30:09 marc
added possibility to pass string to stdin of child process
Revision 1.4 2004/10/07 09:27:01 marc
errors in documentation
Revision 1.3 2004/08/28 16:21:25 marc
mrw-c++-0.92 (mrw)
- new file: version.cpp
- new file header for all sources
- work around warning in mrw::auto<T>
- possibility to compile without log4cxx
- work around bugs in demangle.h and libiberty.h
- corrections in documentation
- added simple tracing mechanism
- more warnings
- small corrections in Auto<>::Free and a new test for it
- possibility to compile without stack trace
*/
#ifndef __MRW_EXEC_HPP__
#define __MRW_EXEC_HPP__
#include <string>
#include <list>
#include <memory>
#include <mrw/exception.hpp>
#include <mrw/unistd.hpp>
namespace mrw {
/** @defgroup CmdExec Execute UNIX Commands
There is no easy way to safely execute UNIX commands and to
return the output of the callee to the caller. @c system ist
first of all known to be unsafe, because it opens a shell, and
second there is no way to transfer the output back to the
caller. On the other hand, starting a new process with @c fork
and @c exec and passing the output of the callee to the caller
using pipes is quite complex and needs much more than one simple
line of code. This is the gap that is filled with this command
execution classes. There's a class for the command to be
executed and a class for the execution of the command.
Forking a subprocess and evaluating the result becomes so easy:
@code
try {
// execute the command: /bin/ls -l /tmp
mrw::Exec ls =
(mrw::Cmd("/bin/ls"), "-l", "/tmp").execute(false);
// evaluate the result
if (ls.success())
std::cout<<"Execution successful, result was:"<<std::endl;
else
std::cerr<<"Error in execution, error was:"<<std::endl;
std::cout<<ls.result()<<std::endl;
std::cerr<<ls.error()<<std::endl;
} catch (ExecutionFailedExc& x) {
// a fatal execution error occurred
// you can trace x.what() and x.stacktrace()
}
@endcode
It is also possible to pass an @c stdin input argument to the
subprocess called:
@code
try {
mrw::Exec cat = mrw::Cmd("/bin/cat")
.execute("this is passed to stdin");
// "cat" passes all from stdin to stdout, therefore:
assert(cat.result()=="this is passed to stdin");
} catch (...) {} // ignore
@endcode */
//@{
class Cmd;
//============================================================================
/** @brief Exception: Execution of command failed.
@pre \#include <mrw/exec.hpp>
This exception is thrown, if the exection of a command in
mrw::Exec is failed. That means, it was not possible to fork or
to create the necessary pipes, or the command executing process
terminated with an error. In the last case, you can access the
error stream from @c stderr respectively @c cerr with method
mrw::Exec::error(). */
class ExecutionFailedExc: public mrw::exception {
public:
ExecutionFailedExc(const std::string&, const std::string&)
throw(std::bad_exception);
virtual ~ExecutionFailedExc() throw() {}
virtual const char* what() const throw() {return _what.c_str();}
private:
std::string _what;
};
//============================================================================
/** @brief Execute a command in a new process.
@pre \#include <mrw/exec.hpp>
This class handles the execution of a command in a new process
and returns the two streams @c cout and @c cerr, also known as @c
stderr and @c stdout.
Method @c execute can optionally also take a string parameter
that is passed to @c stdin of the child process.
There are different ways of usage for this class. A simple way,
one line of code, to get only the resulting stream (no error)
is:
@code
string stdout =
(mrw::Cmd("/bin/ls"), "-l", "/tmp").execute(false).result();
@endcode
If you need not only the resulting @c stdout stream, but also
the error stream @c stderr, then you need to store the result:
@code
mrw::Exec ls =
(mrw::Cmd("/bin/ls"), "-l", "/tmp").execute(false);
if (!ls) ...; // command termianted with error
// ls.result() contains stdout
// ls.error() contains stderr
@endcode
@note Please note that the command execution may throw an exception. */
class Exec {
//................................................................ methods
public:
/** @brief Create an executor given a command.
Construction without passing a command is not possible. */
Exec(const mrw::Cmd&) throw(std::bad_exception);
Exec(const mrw::Exec&) throw(std::bad_exception);
~Exec() throw();
Exec& operator=(const mrw::Exec&) throw(std::bad_exception);
/** @brief Execute the command.
@param exc
- @c true throw an exception if return status is not zero
- @c false throw only an exception in case of a fatal error
@throw ExecutionFailedExc is thrown if
- fork fails
- creation or setup of pipes failed
- if given parameter is @c true (the default) also if the
executed program terminates with an error */
Exec& execute(bool exc=true) throw(std::exception);
/** @brief Execute the command, pass @c stdin.
@param input Input that is passed to @c stdin of the child process.
@param exc
- @c true throw an exception if return status is not zero
- @c false throw only an exception in case of a fatal error
@throw ExecutionFailedExc is thrown if
- fork fails
- creation or setup of pipes failed
- if given parameter is @c true (the default) also if the
executed program terminates with an error */
Exec& execute(const std::string& input, bool exc=true)
throw(std::exception);
/** @brief Execute the command, pass @c stdin.
@param input Input that is passed to @c stdin of the child process.
@param exc
- @c true throw an exception if return status is not zero
- @c false throw only an exception in case of a fatal error
@throw ExecutionFailedExc is thrown if
- fork fails
- creation or setup of pipes failed
- if given parameter is @c true (the default) also if the
executed program terminates with an error */
Exec& execute(char const*const input, bool exc=true)
throw(std::exception) {
return execute(std::string(input), exc);
}
/** @brief Executes the command if not done, streams @c stdout
into a string
If the command has not yet been executed successfully, it is
first executed, then the @c stdout output of the called
program is appended to the string.
@throw ExecutionFailedExc in case of any failure or if the
executed program does not return a zero exit status. */
Exec& operator>>(std::string&) throw(std::exception);
/** @brief Executes the command if not done, returns @c stdout as string
If the command has not yet been executed successfully, it is
first executed, then the @c stdout output of the called
program is returned.
@return @c stdout of the called program
@throw ExecutionFailedExc in case of any failure or if the
executed program does not return a zero exit status. */
operator std::string&() throw(std::exception);
/** @return
- @c true if the last execution was successful
- @c false if the last execution failed or the command was
never executed */
operator bool() throw(std::bad_exception);
/** @brief Executes the command if not done, returns @c stdout as string
If the command has not yet been executed successfully, it is
first executed, then the @c stdout output of the called
program is returned.
@return @c stdout of the called program
@throw ExecutionFailedExc in case of any failure or if the
executed program does not return a zero exit status. */
std::string& result() throw(std::exception);
/** @brief Executes the command if not done, returns @c stderr as string
If the command has not yet been executed successfully, it is
first executed, then the @c stderr error output of the called
program is returned.
@return @c stderr of the called program
@throw ExecutionFailedExc in case of any failure or if the
executed program does not return a zero exit status. */
std::string& error() throw(std::exception);
/** @return
- @c true if the last execution was successful
- @c false if the last execution failed or the command was
never executed */
bool success() throw(std::bad_exception);
//................................................................ methods
private:
Exec(); // no default constructor
//.............................................................. variables
private:
friend class PartialExec; // don't want the variables protected
mrw::Cmd* _cmd;
std::string _res, _err;
bool _success;
};
//============================================================================
/** @brief Execute a UNIX program in non blocking parts.
@pre \#include <mrw/exec.hpp>
A given UNIX command is executed, but the class does not wait
until it is finished, instead it gives back the control to the
caller. This behaviour is achieved using non blocking
communication. But the caller is responsible to retrieve all
information from the client, and if necessary to close the input
pipe of the client executable program. Therefore you have to
give back control from time to time, normally this is doen in a
@c while loop, where you can execute also different thing,
e.g. update a display of the result or similar.
With this class, you can communicate with a child process, and
do other things at the same time, without the need for multi
threading.
Execution of a program works the following way:
- do not use execute() (otherwise the behaviours is identical
to class mrw::Exec, you gain nothing, but also loose nothing)
- use start() to start the external program
- use start() or @c start(false) if you don't want to pass
input to the child process
- use @c start(true) if you want to pass input to the
child process
- if you called @c start(true), call finish() if you have no more
input to send to the child process (it's like an end-of-file)
- the execution is not terminated, before finished() returns
@c true
- while finished() is false, subsequently call read() to read
the output of the child process
@warning After calling finish(), or if you did not call
start() with parameter @c true, it is forbidden to
pass anything but an empty string as first parameter to
read()! Anything else is a programming error and
results in an assertion failure and a core dump!
@note If your program seems to hang, check if you call finish()
correctly!
Here an example:
@code
mrw::PartialExec exec = mrw::Cmd("/bin/cat").start(true);
std::string res = exec.read("This is a test\n").first;
res += exec.read("This is another test\n").first;
exec.finish(); // close the input pipe of @c cat
while (!exec.finished()) res+=exec.read().first;
@endcode */
class PartialExec: public Exec {
//................................................................ methods
public:
/** @brief Create an executor given a command.
Construction without passing a command is not possible. */
PartialExec(const mrw::Cmd&) throw(std::bad_exception);
/** @brief Copy construction invalidates the original object.
All opened pipes (opened with start()) are lost in the
original object and are then owned by the new object. */
PartialExec(mrw::PartialExec&) throw(std::bad_exception);
/** @brief Copy construction invalidates the original object.
@copydoc PartialExec(mrw::PartialExec&)
@warning @c const for the argument is a fake! It is casted away!
@param e @b Warning: const is casted away! */
PartialExec(const mrw::PartialExec& e) throw(std::bad_exception);
/** @brief Assignment invalidates the original object.
@copydoc PartialExec(mrw::PartialExec&) */
PartialExec& operator=(mrw::PartialExec&) throw(std::bad_exception);
/** @brief Close the input pipe of the child process.
If start() is called with argument @c false, then you
can pass input to @c stdin of the child process, but you @b
must call this method, after passing the last input
string. Otherwise, the child's input pipe won't be closed,
the child process does not stop waiting for more input! If
your program seems to hang, check if you call finish()
correctly! */
PartialExec& finish() throw();
/** @brief Check if there's more data left to read().
@return @c true if the child process has finished and all
data is read. */
bool finished() throw();
/** @brief Start a new child process.
At most one child process can run at the same time.
@throw mrw::runtime_error if a previous child has not finished() yet
@throw mrw::ExecutionFailedExc if the child process cannot be started
@param useInput
- @c true if input will be sent to the child's @c stdin
- pass all input in the first parameter of read()
- finish() must be called when all input is sent
- @c false if no input is sent to the child's @c stdin
- the first parameter of read must allways be passed an
empty string */
PartialExec& start(bool useInput=false) throw(std::exception);
/** @brief Read from the subprocess, optionally pass an @c input to
@c stdin of the subprocess.
@param input a string to pass to the child processes @c stdin
@param exc
- @c true throw an exception if return status is not zero
- @c false throw only an exception in case of a fatal error
@return a pair containing the last read @c stdout and @c stderr
of the child
@throw ExecutionFailedExc is thrown if
- fork fails
- creation or setup of pipes failed
- if given parameter is @c true (the default) also if the
executed program terminates with an error
@note If start() was not called with parameter @c true, then
@c input must always be an empty string!
@pre start() was called */
std::pair<std::string, std::string> read(const std::string& input="",
bool exc=true)
throw(std::exception);
/// Terminates a running job by sending @c SIGTERM to the child process.
PartialExec& terminate() throw();
/// Kills a running job by sending @c SIGKILL to the child process.
PartialExec& kill() throw();
//................................................................ methods
private:
PartialExec(); // no default constructor
//.............................................................. variables
private:
bool _finished;
bool _finish;
std::auto_ptr<mrw::Pipe> _stdIn, _stdOut, _stdErr;
std::string _input;
int _num0, _num1, _num2, _lastPid, _pid;
};
//============================================================================
/** @brief A system command to be executed
@pre \#include <mrw/exec.hpp>
This class is used in conjunction with mrw::Exec. It must be
initialized with the command name, then the command parameters
are appended either with commas, or by streaming them into the
command, whatever you like.
You can stream the data into the class:
@code
mrw::Cmd ls("/bin/ls"); // the command to execute is: /bin/ls
ls<<"-l"<<"/tmp"; // the command is now: /bin/ls -l /tmp
@endcode
Or you can setup your command with commas:
@code
mrw::Cmd ls = (mrw::Cmd(/bin/ls), "-l", "/tmp");
@endcode
*/
class Cmd {
public:
/** @brief Create a command given the name of the executable
@param command the name of the program to execute (no parameter)
@note There is no default constructor. */
Cmd(const std::string& command) throw(std::bad_exception);
/** @brief Append a parameter to a command
@param param a parameter&nbsp;/ commandline argument
to append to the command */
Cmd& operator,(const std::string& param) throw(std::bad_exception);
/** @brief Append a parameter to a command
@param param a parameter&nbsp;/ commandline argument
to append to the command */
Cmd& operator<<(const std::string& param) throw(std::bad_exception);
/** @return the command including parameter */
operator std::string() const throw(std::bad_exception);
/** @return a mrw::Exec that's constructed with this class */
operator mrw::Exec() const throw(std::bad_exception);
/** @return a mrw::PartialExec that's constructed with this class */
operator mrw::PartialExec() const throw(std::bad_exception);
/** @brief Create a mrw::Exec and execute a child process.
@see Exec::execute(bool) */
Exec execute(bool exc=true) const throw(std::exception);
/** @brief Create a mrw::Exec and execute a child process.
@see Exec::execute(const std::string&, bool) */
Exec execute(const std::string& input, bool exc=true) const
throw(std::exception);
/** @brief Create a mrw::Exec and execute a child process.
@see Exec::execute(char const*const, bool) */
Exec execute(char const*const input, bool exc=true) const
throw(std::exception) {
return execute(std::string(input), exc);
}
/** @brief Create a new mrw::PartialExec and start a new child process.
@see PartialExec::start(bool) */
PartialExec start(bool useInput=false) const throw(std::exception);
private:
// Exec and PartialExec are allowed to call @c path() and @c args().
friend class Exec;
friend class PartialExec;
Cmd(); // No default constructor.
const char* path() const throw(std::bad_exception);
char** args() const throw(std::bad_exception);
typedef std::list<std::string> ArgList;
ArgList _cmd;
};
//@}
}
#endif