elvish shell

I decided to give the elvish shell a try today. Elvish is a shell written in golang, designed to be more expressive than bash.

Bash drops just short of being a programming language. For example, it doesn’t do floating point arithmetic, which is a nuisance. Elvish does, though.

I set myself a little task: write a program which reads from standard in, prints the number of lines read, then prints the input. It’s an interesting task because it must cache the input.

Here’s what I cam up with in elvish:

#!/usr/bin/env elvish
contents = [(cat )]
echo (count $contents)
for line $contents { echo $line }

Here’s the equivalent I write in awk:

{
 i += 1;
 line[i] = $0;
}

END{
 print i;
 for(idx=0; idx<i; idx++) print line[idx];
}

As you can see, elvish manages the same thing in less than half the number of lines.

Here’s my C++ version:

#include <iostream>
#include <string>
#include <vector>

int main()
{
 std::vector<std::string> lines;
 std::string line;
 int nlines = 0;
 while(std::getline(std::cin, line)) {
 nlines++;
 lines.push_back(line);
 }

std::cout << nlines << "\n";
 for(std::vector<std::string>::iterator it = lines.begin(); it != lines.end(); ++it)
 std::cout << *it << "\n";

return 0;
}

That’s 20 lines of code.

Here are the timings of running the scripts:

time awk -f count.awk <derive-2.txt >/dev/null
real 0m0.067s
user 0m0.004s
sys 0m0.002s

Although there are some odd timings:


time awk -f count.awk <derive-2.txt >/dev/null

real 0m0.006s
user 0m0.003s
sys 0m0.002s

$time elvish count <derive-2.txt >/dev/null

real 0m0.052s
user 0m0.025s
sys 0m0.011s

time ./a.out < derive-2.txt >/dev/null # c++ version

real 0m0.006s
user 0m0.004s
sys 0m0.001s

Awk seemed a little inconsistent in its performance. I’m not sure what’s going on there. It looks like Awk can be a real speed demon, and I’ve even seen it run a little faster than the C++ code. Those Awk guys certainly seem to know how to optimise stuff.

Elvish is certainly worth checking out. I’m not sure about its runtime speed, though.