I'm using this wrapper for myself. I've created a composer package installable from packagist.org. I've created this branch to keep your namespace and contribute my changes. If you want them. Here is what I did:

• added .gitignore
• made it psr-0 compatible
• ready to add to packagist.org
• passed through php-cs-fixer (minor fixes, indentation, etc)
• moved ShellWrap.php to src/MrRio for autoloading
• moved examples to new directory
• added LC_ALL=en_US to testDate() to avoid error in non-english environment

NOTE: The diff on ShellWrap is the complete file due to directory changes..

If you use function chdir() to change working directory to a different directory first, following two statements should still return same directories. However, in Travis CI they return different directories back (the 2nd one returns the directory where the PHP script was invoked).

1. getcwd()
2. (string) ShellWrap::pwd()

Thus we need to have the 4th parameter (getcwd()) specified explicitly here, also ideally it's unnecessary.

In some PHP IDE (e.g., PhpStorm) you may see that method _\MrRio\ShellWrap::_toString() is highlighted with notification saying

__toString() method must return a string

This could be triggered with following code piece where no actual shell command invoked:

<?php
require_once __DIR__ . '/vendor/autoload.php';

echo (new MrRio\ShellWrap());
?>

Although in reality people may not make method calls like that, it's better to get it fix to make variables consistent. Property \MrRio\ShellWrap::$output should be initialized and used as a string but not an array, as in the fix.

I've opened an issue against Boris here, however, I am also going open an issue here for anyone else experiencing this. When using shell wrap with a composer version >1.3.0, PHP gets reloaded to disable XDebug and causes an error to be thrown in the terminal. E.g.,

PHP Warning:  Module 'pcntl' already loaded in Unknown on line 0

Warning: Module 'pcntl' already loaded in Unknown on line 0

As Boris is only used for the interactive debugger, is there any chance of a workaround? Boris hasn't had any commits or changes in years, maybe Psysh would be a good drop-in replacement?

My case is currently that I'm streaming use log files using sh::cat() and using a callback closure to pass the data (after a filter) to an PHPExcel object. The log files can easily get over 500MB. Looking through the code it seems that even when using a callback/closure, the output of the cat command still gets saved to a variable, making memory usage go up and up.

I cannot grep on the log files as it's filled with IDs, possible conflicting with other data in the same line.

Is this currently possible in a less memory consuming way? Else I'm also willing to submit a PR with some extra functionality on this. Reasoning could be that when streaming through a closure, no need is left to save things to the self::$output variable.

echo sh::grep('html', sh::curl('http://example.com', array(
    'location' => true
)));

In current API I don't like 2 things:

1. inverted order of commands (grep -> curl) when actually curl | grep is intended
2. "wrap" or functional style a(b(c()))

It would be nice to have fluid API like this

sh::curl('http://example.com')->pipe('grep', 'html')->pipe('whatever');

or

sh::curl('http://example.com')->sh('grep', 'html')->sh('whatever');

or

sh::curl('http://example.com')->grep('html')->whatever();

@MrRio what do you think?