Advices and Gotchas when using cling and jupyter¶

The cling and xeus docs are good, but some informations are missing : this notebook adds more details.

Cling official docs: https://root.cern.ch/cling

.x filename.cxx - loads filename and calls void filename() if defined
.L library | filename.cxx - loads library or filename.cxx
.printAST - shows the abstract syntax tree after each processed entity
.I path - adds an include path

xeus-cling official docs: https://xeus-cling.readthedocs.io/en/latest/index.html

Some more advices / missing docs¶

Add include path / And library path / load libraries¶

Add library and include path¶

#pragma cling add_library_path("...") will add a path to the library path
#pragma cling add_include_path("...") will add a path to the include path

Load libraries¶

#pragma cling load("mylibrary") will load a library. You do not need to add the library extension (.so, '.lib, or .dylib). You can also load a library using .L.

Do not put a ";" at the end if you want to see the computation result¶

The cell below will output a result:

In [1]:

2*3

Out[1]:

The cell below will not, because of the ending ";"

In [2]:

2*3;

Create a .cpp or .h script in order to set your library/include path and load common libraries¶

See for example init_stdlibs.h or init.opencv.cpp

Below is an example usage:

In [3]:

#include "init_stdlibs.h"
#include "init.opencv.cpp"

This is C++ : everything can be defined only once¶

If you run twice a code that defines the same variable or a function, you will encounter a compiler error, with a bad looking Red Error Message (error: redefinition of '...'). There are several ways to work around this :

Method 1: restart the compiler from scratch

select "Kernel/Restart" and re-run the different cells

Method 2: Enclose the code contained in a cell in an anonymous code-block in order to create a closed scope for your variables.

For example, a cell like the one below can be run several times:

In [4]:

{
  std::string s("Hello how are you");
  int nbspaces = count(s.begin(),s.end(),' ');
  std::cout << "nbspaces=" << nbspaces << "\n";
}

nbspaces=3

Method 3: use Immediately Invoked/Executed Lambdas (IIL), for example:

const auto your_var = []{ YourCodeHere; return something; }();

In [5]:

[] { // This is the start of the Immediately Invoked Lambda 
     // (the "()" can be avoided, since it does not have params)
  cv::Mat m = cv::imread("data/lena.jpg"), blur;
  cv::resize(m, m, cv::Size(200, 140));
  cv::blur(m, blur, cv::Size(15, 15));
  return im::show(blur);
}() 
// * note the "()" on the last line: it immediately invokes 
//   the lambda (hence the IIL name)
// * Also note that the ending ";" is missing, so that the 
//   last computation result will be shown in the notebook

Out[5]:

Note: Immediately Executed Lambdas are similar to the javascript idiom immediately invoked function expression (IIFE).

Some gotchas with cling and jupyter¶

Do not define two functions in the same cell¶

Failure example¶

this cell will give an error : "function definition is not allowed here"

In [6]:

int foo() { return 42; }
int foo2() {return 43; }

input_line_17:3:13: error: function definition is not allowed here
 int foo2() {return 43; }
            ^

Interpreter Error:

Correction¶

Create two cells

In [7]:

int foo() { return 42; }

In [8]:

int foo2() {return 43; }

Add ";;" after lambdas¶

Failure example¶

If you run the two cells below, the notebook "Kernel" will fail, and you will need to restart it (menu "Kernel / Restart")

In [9]:

auto same = [](int i) { return i; }; // need to add a second ";" !

In [ ]:

same(1)

Correction: add ";;" after the lambda¶

In [1]:

// Strangely it will work with ";;"
auto same = [](int i) { return i; };;

In [2]:

same(1)

Out[2]:

Higher order functions : pass input functions by universal reference¶

Failure example : passing input function by copy¶

In [3]:

int add1(int i) { return i + 1; }

In [4]:

// This higher order function capture its input function (f) by copy
auto call_twice = [] (auto f) {
    return [&](auto x) {
      return f(f(x)); 
    };
};;

In [5]:

auto add2 = call_twice(add1);;

In [ ]:

// this call will cause the kernel to fail
add2(2);

Correction : pass the param via universal reference (&&)¶

In [1]:

int add1(int i) { return i + 1; }

In [2]:

// This higher order function capture its input by universal reference (&&)
auto call_twice = [] (auto && f) {
    return [&](auto x) {
      return f(f(x)); 
    };
};;

In [3]:

auto add2 = call_twice(add1);;

In [4]:

add2(2)

Out[4]:

When running cling from the console¶

Command line arguments (when using cling from the console)¶

cling is a version of clang that includes REPL. Thus, it accepts the standard clang command line arguments.

For example:

cling -I/opt/include -Lopencv_core --std=c++14

Add `using namespace std`¶

If you try to display a std::string or a std::vector<std::string> you might get some errors like this inside cling, when you are on the console (not inside notebook)

error: no template named 'basic_string'; did you mean 'std::basic_string'?

You can solve this by adding a line using namespace std

In [ ]:

Advices and Gotchas when using cling and jupyter¶

Some more advices / missing docs¶

Add include path / And library path / load libraries¶

Add library and include path¶

Load libraries¶

Do not put a ";" at the end if you want to see the computation result¶

Create a .cpp or .h script in order to set your library/include path and load common libraries¶

This is C++ : everything can be defined only once¶

Some gotchas with cling and jupyter¶

Do not define two functions in the same cell¶

Failure example¶

Correction¶

Add ";;" after lambdas¶

Failure example¶

Correction: add ";;" after the lambda¶

Higher order functions : pass input functions by universal reference¶

Failure example : passing input function by copy¶

Correction : pass the param via universal reference (&&)¶

When running cling from the console¶

Command line arguments (when using cling from the console)¶

Add using namespace std¶

Add `using namespace std`¶