Fix string indexing

[certik]

The following code:

def f():
 s: str
 s = "abcdefg"
 print(s[1]) # prints "bcdefg", but should print "b"
 print(s[1:2]) # prints "cdefg", but should print "b"
 print(s[1:3]) # prints "defg", but should print "bc"

f()

Prints (correctly) using CPython:

$ python a.py 
b
b
bc

But with LPython it prints:

$ lpython a.py && ./a.out 
bcdefg
cdefg
defg

We have to fix it.

[namannimmo10]

C++ code for reference:

#include <iostream>
#include <string>
#include <vector>
#include <cassert>
#include <cmath>
#include <Kokkos_Core.hpp>
#include <lfortran_intrinsics.h>

template <typename T>
Kokkos::View<T*> from_std_vector(const std::vector<T> &v)
{
 Kokkos::View<T*> r("r", v.size());
 for (size_t i=0; i < v.size(); i++) {
 r(i) = v[i];
 }
 return r;
}

void _lpython_main_program()
{
 f();
}

void f()
{
 std::string s;
 s = "abcdefg";
 std::cout << s[1 + 1-1] << std::endl;
 std::cout << s[2 + 1-1] << std::endl;
 std::cout << s[3 + 1-1] << std::endl;
}

namespace {

void main2() {
 _lpython_main_program();
}

}
int main(int argc, char* argv[])
{
 Kokkos::initialize(argc, argv);
 main2();
 Kokkos::finalize();
 return 0;
}

[namannimmo10]

We should improve it as well.

[namannimmo10]

I guess the problem is here:

lpython/src/libasr/codegen/asr_to_llvm.cpp

Lines 1029 to 1043 in 3104565

	if (l == r) {
	this->visit_expr_wrapper(x.m_args[0].m_left, true);
	llvm::Value *idx = tmp;
	idx = builder->CreateSub(idx, llvm::ConstantInt::get(context, llvm::APInt(32, 1)));
	//std::vector<llvm::Value*> idx_vec = {llvm::ConstantInt::get(context, llvm::APInt(32, 0)), idx};
	std::vector<llvm::Value*> idx_vec = {idx};
	llvm::Value *str = builder->CreateLoad(array);
	llvm::Value *p = builder->CreateGEP(str, idx_vec);
	// TODO: Currently the string starts at the right location, but goes to the end of the original string.
	// We have to allocate a new string, copy it and add null termination.
	tmp = builder->CreateAlloca(character_type, nullptr);
	builder->CreateStore(p, tmp);
	//tmp = p;

[Smit-create]

This issue might be easier to handle once we move forward with: https://gitlab.com/lfortran/lfortran/-/merge_requests/1700