Re^2: How to build system calls using '=>'

..or any passers by who are hunting for a way to make a path of directories they would be better following blue_cowdawg's advice in Re: "mkdir -p" equivalent? and using mkpath in File::Path

I always wonder how useful statements like "it's better to use X instead of Y" are if they don't say why X is better.

I'll bite. I say using the systems 'mkdir' is better than 'mkpath' because the arguments of 'mkpath' are hard to remember (without looking in the docs, what's the second argument of mkpath? How do I tell mkpath which mode I need? How do I create multiple directories in one go?). On top of that, 'mkpath' has a different name than 'mkdir'.

Sure, if you work on a system that doesn't have 'mkdir', you're better off to use File::Path. But then, programming in such an environment is like writing a novel without using the letter 'e'.

If they have a reason they can't use File::Path (even though it's a core module these days) they would be better using the built-in mkdir than shelling out to a system one.

And then you present a 17 line piece of code, which has a OS dependency as well, doesn't have an option to set a mode, and is verbose without an option to turn it off.

I strongly prefer to shell out using one line, have the ability to set the mode, and which, by default, is silent.

Perl --((8:>*

Comment on Re^2: How to build system calls using '=>'

Replies are listed 'Best First'.
Re^3: How to build system calls using '=>' by serf (Chaplain) on Dec 14, 2005 at 15:45 UTC
Why: mkpath doesn't require an external binary, doesn't create another process to do its job, and has full control over what it's doing and what comes back - there are less places that the attempt to create a directory can go wrong. Re remembering: The link to blue_cowdawg's advice in Re: "mkdir -p" equivalent? is there specifically because it shows an example of how to use mkpath, including modes etc - so they don't have to remember. The 17 line piece of code is an _example_. It is verbose so that the user knows where to report if they need to do so, they can always comment out the print, but it's where it is so they know if they want to report, that's where they should do it, i.e. it should only say it's done it once it's worked and not before. Shelling out to 'mkdir -p $path' is OS dependent as well - if you do that on Windows (tested on xp) you will end up with a directory called "-p" as well. In the example it points out where to change the directory seperator. mkdir allows you to set the mode as well, if you wish to - fair call, perhaps I could have added that in the example. :o)	[reply]
Re^4: How to build system calls using '=>' by Perl Mouse (Chaplain) on Dec 14, 2005 at 16:40 UTC
mkpath doesn't require an external binary Eeeck! Scary! An external binary! Code reuse! Bad! Bad! Bad! Perl is a glue language. There's nothing wrong with using a code someone else wrote. Whether that's "an external binary" (what's an "internal binary"?) or a module. doesn't create another process to do its job Yes, and? Are you worried about performance? Sure, creating another process takes a few microseconds. But we're about the muck around in the filesystem anyway. And compiling a piece of Perl isn't exactly resource free either. so they don't have to remember. Yeah, but then they have to look it up. And that takes valuable programmer time. Shelling out to 'mkdir -p $path' is OS dependent as well I never claimed it wasn't. The 17 line piece of code is an _example_. Of course. But: system 'mkdir', '-p', @dirs and die; is just one line. Even if you mimic it using 5 lines, it's still 400% more lines. Remember: programmer time is expensive! `Perl --((8:>*`	[reply]
Re^5: How to build system calls using '=>' by Anonymous Monk on Apr 17, 2009 at 16:34 UTC
Gotta agree with serf here. It's not about whether or not it's expensive or creating another process for one routine that you would want to do. It's more about practicing good programmer judgment when it comes to writing your code. Would you make system calls for everything you could (like, opening and reading in files, truncating files, moving files, parsing files) in a 3000-5000 line perl script or module you'd created? I doubt it. So why, for the love of god would you promote doing this simple routine with a system call as well. It's just poor programming practice. Serf has decidedly better points than you do about how to write this perl routine and why you would do it his way. And yes, there are several problems with using external binaries that are outside of perl to perform routines for your program. 1. Sometimes they will simply not port from one OS to another, talk about waste of valuable programmer time. You would need to re-write your code for every possible OS diference, if needed. 2. They can exercise a larger tax on system resources due to process spawning. Especially if this is the route you're going to take with large pieces of code. But, with complex pieces of code you want to be making very few system calls, if any at all. You may quickly find all of your spawned processes out of control. Lastly, while I don't promote putting in code simply for the sake of bloating your code with nonsensical information and handlers, if you're measuring your code by the number of lines of code you're producing, you're doing it wrong. You can write the same function many different ways, and while one way may take up 'less' space, it's almost certain to not have enough error handling and comments. I've never posted on Perl Monks, and I deeply appreciate all of the great folks who come out here and give freely their knowledge. But, I simply refuse to stand idly by and allow this to be the last comment on this thread as if it's the right thing to do. Peace.	[reply]