On Wed, 08 Apr 2009 10:10:29 -0400, Daniel Keep
<[email protected]> wrote:
Steven Schveighoffer wrote:
[snip]
-Steve
This is getting rather inflammatory, so let me clarify a few things.
What Yigal should have said was that when everything is a file, there is
no API for the CONTENTS of files. Yes, you can manipulate files, but
it's a bit like Variants in D: you can't do anything useful with the
contents unless you specifically know what they are.
Powershell is designed to work with managed (.NET) objects. It can
dynamically introspect these objects and pull them apart, mutate them,
convert them, etc. Imagine if every file on a UNIX system also carried
around a reference to its own parser, and a description of how to
interpret the data. THAT is what Powershell has over UNIX shells.
For example, the output of the "ls" equivalent is a table of objects.
By default, it gets toString'd and displayed as a table. However, you
can pipe this into a command that reorders columns. Or sorts. And you
don't have to dick about with ensuring the filenames don't contain
whitespace or newlines because the file names are carried around as
actual strings that know their own length.
Unlike UNIX, where everything is a file, to powershell everything is an
object. So the list command can be used on a directory or on an object
that is an aggregate of stuff. For example, you can use powershell to
browse the registry as if it was just another part of the filesystem.
This does sound useful. More so than was explained, sorry to Yigal for
being so inflammatory, but what can you expect when you say Unix is a
"stupid" design ;) It's just a different design, one that works very well
for script writers.
I won't say that "everything is a file" is a failure, but powershell
*definitely* exposes a superset of UNIX's functionality. Anything you
can do with a UNIX-style environment, powershell can potentially do
better.
This kind of thing is useful only for cases where OS objects were
converted into strings (such as via ls), actual files still have to be
read and parsed I'm guessing. But having .NET at your disposal gives you
a lot of potential parsers for things like XML files or whatever.
BTW, /proc is a lot like this. You don't need to have a manual to
understand what it means, since you can browse through it just like any
other directory. However, there are some files that have undescribed
values in them, where you do need a manual. But these files are built to
be parsed for the most part, so it doesn't take a lot of work to get
information out of it. I think the advantage to having everything be
represented as files is that you already have a bazillion tools to
manipulate and parse files in unix. The same is not for Windows, but it
does have .NET which has a great API to OS objects.
Microsoft really hit one out of the park with powershell, although it
remains to be seen if anyone will actually use it.
It sounds a lot like it is geared toward .NET programmers rather than
script writers. Do you use it?
-Steve
