Basic
Unix File System Concepts
| Unix stores information in files;
a file is essentially just a bunch of computer bytes with a name assigned
to them. Programs, data, text, images, and a million other things
are all just files as far as Unix is concerned. A file may be
a short laundry list you typed in, a huge web page you downloaded, a complex
Microsoft Word document, an e-mail message you saved, an executable computer
program, etc. Unix works on files. One major way to get Unix
to do things for you with your files is to use a command
line interface to type in commands. Most commands operate on (do
something to or with) files. gcc,
for example, is the name of the Unix command that runs the C compiler;
so gcc my_program.c
commands Unix to compile the C
program stored in a file named my_program.c
(You'll use an editor to create your program files.)
File names: Although modern Unix
systems can let almost any collection of characters be a file name, it's a good idea to
follow some basic guidelines (not rules, when you know how and why, you can violate
any of these with near impunity):
- Start with a letter; after that first letter, you can use numbers, letters, and
(carefully) some special characters — but see the third guideline, on special
characters, below. Note: Unix, like most
modern systems, actually lets you start file names with numbers or even other characters,
but you may sometime need to move a file to an older or less forgiving system (like DOS,
for example) and there can be other difficulties with odd file names as well.
- Unlike Microsoft Windows,
Unix is case sensitive, so AAA
and AAa
would be two different Unix filenames.
- If the letters you type are being echoed (displayed) in upper case even though you're
not depressing the "Shift" key when you type, you have probably previously
inadvertently hit the "Caps Lock" key at the left of your keyboard. Hit it
again to turn off "Caps Lock". Now your keystrokes should be
displayed in lower case unless you explicitly hit the "Shift" key.
- Don't embed blank spaces. The string of characters "my
really cute filename" might or might not
work on any given Unix system; but even if it did, it would sometimes
require you to use the "quote marks" and sometimes not.
Besides, as above, there are other, older systems that you might someday
want to move your files to without renaming them. Instead of a
blank, use the _underscore_ (a capital -minus- sign, just to the
right of the zero on most keyboards). my_really_cute_filename
is a good, safe Unix file name.
- Don't use special characters (besides the underscore) until you know enough about Unix
to do so safely. The asterisk, question mark, slash, period, square brackets, single
or double quote marks, and other such characters can have special meanings to Unix or some
of its commands.
Extensions:
Many (not all, not even most) of the programs that make up Unix expect
file names to end with what is called an extension. An extension
is a period followed by up to three characters. Extensions are often
optional; it's usually not a big deal to work around missing or misleading
extensions, but it can be more work and in any case, extensions help you
keep track of what type of data you've got in your files. So a
file name of good_name.txt
would be presumed to be a text file (just characters about which
Unix knows nothing else — your laundry list, for example); while
good_name.c
might be the source code
of a C program you've written. After compilation,
the resulting executable program (also known as a binary file, or binary
image — what you'll actually tell Unix to execute to run your program)
might be stored in a file named good_name.exe
Extensions are more for your benefit than the computer's; they're customs
I recommend you follow. When you have hundreds of files, how will
you remember just by looking at its name that a file named my_first_program
contains a C program or an Ada program or a Basic program or whatever?
And every so often you'll run into a command that is touchy about
extensions — like our current Ada compiler. Naming an Ada program
source file good_name won't
work here; to be compiled by our Ada compiler, an Ada program must be
stored in a file whose name has an .adb
extension.
|
Directories and Paths:
To make it easier for you to keep track of files, Unix allows you to organize
them into a hierarchical structure of directories (some other systems call them
folders). You might, for example, keep all your C programs in a directory
named "CS125" and
all your personal correspondence in a directory named "letters".
If you had a lot of letters, you might have a sub directory for letters/home,
another for letters/girl_friends,
yet another for letters/IRS,
and so on. These sub-directories
can continue for as many levels downwards as you wish, as in letters/girl_friends/Mary,
letters/girl_friends/Jill,
etc. Technically, to Unix, a directory is just a file that
contains the names of other files as well as some information about them —
like, most importantly, where to find them on the hard disk, for example.
(Almost everything to Unix is just a file of some sort; data
is in files, directories are files, your programs will be files, even the commands
you'll use are [mostly] just programs stored in files.)
So a directory is a file that contains information about other
files, possibly including other directories which can then contain other files
and directories, and so on. Computer scientists call such hierarchical
structures, where one container can contain other containers, "trees"
but draw them upside down as shown in Figure 1, below, with the root (or base)
of the tree at the top and the branches spreading down instead of up.
(I don't want to hear any comments here about computer scientists getting everything
ass backwards; there are good historic reasons for drawing trees like this;
you're just undoubtedly too young to know the history ;-)
A directory can contain any mix of both "ordinary" files and other
directories (known as sub-directories). Unix doesn't care, so long as everything
(except one directory, called the root directory, which we'll define in a few
moments) is contained within some higher level directory.
Figure 1. Sample Directory Hierarchy. The directory hierarchy forms
an upside down tree.
In Unix, every file has a basic name, which just differentiates
it from other files in the same directory (every file is stored in some
directory, remember), and a "full" name which allows it to be
differentiated from every other file in the entire Unix system. (The
analogy to peoples' first names and full names is pretty close.) In
Figure 1, for example, there are two files who appear to have the same name,
namely April_9.txt, but
actually don't because their full names are different. The fullname
of one of the files there is
/letters/girlfriends/Jill/April_9.txt
while the full name of the other is
/letters/IRS/April_9.txt
I'm sure you've noticed that that full name is fairly long and would be irritating
to have to type out in full any time you wanted to do something with it.
Fortunately, you won't have to type a file's full name very often; usually you
won't even know it. You and Unix are going to agree on a default directory and
as long as you and Unix don't get confused about what directory you're both
talking about, you'll just use basic file names. In the jargon of a Unix
shell, we are said to always be "in" some directory (known as the
"current", or "working" directory).
Obviously there will have to be commands to allow you to ask what directory
Unix thinks it's in, to tell it to change the directory its in, to get
it to tell you the full name of something, etc. We'll come to all those
commands later. But first we've got to define more precisely some of these
types of file names we've been talking about — basic and full, for example,
and there is another type as well.
File names can be either qualified or basic (a basic name is sometimes known as
simple or unqualified). A qualified file name starts with either a fully or
partially qualified path, which is a list of nested
directories; basic filenames do not include any path information at all. A path is
fairly analogous to your street address — it tells Unix how to find the file. So a
qualified file name is something like a snail mail address: John Doe, 1 Main Street,
Anytown, AZ. Note, however, that in snail mail addresses, reading from left to right
moves in the direction of increasing generality, while for a Unix path, left to right is
increasingly specific, moving down the directory tree. Anyway, the fully qualified
file name
/letters/girlfriends/Jill/April_9.txt
consists of:
- a fully qualified path (the hierarchical set of directories, starting and
ending with the slash / character), and
- a basic file name, which in the example above is April_9.txt
A path whose very first character is a slash,
like /letters/girlfriends/Jill/,
is a fully qualified path (we'll cover partially
qualified in a moment, be patient). This one, /letters/girlfriends/Jill/,
tells Unix to start following the path to the file starting in the directory
letters which, as shown in Figure 1,
is located within the very top level directory of the entire Unix system.
Since the file system is a hierarchical tree structure, there must be
exactly one such topmost directory — no more, no less, just exactly
one. In conversation, you and I and all other Unix hacks call it the root
directory, but when talking to Unix, the name of the root is just a slash, all
by itself, like so: /
- The root directory is the exception I mentioned earlier — the root is
the only directory that's not nested within (we also sometimes say
"beneath") some other directory. Every other file (and hence
directory, since a directory is a file) is somewhere beneath the root.
Since there's only and always exactly one root to a Unix file system, it doesn't
really need a name; so the original Unix designers, with their perverted love
of cryptic names for the sake of crypticism, used or designated the /as
the symbol for the root. Using the slash elsewhere, like the other
slashes (not the first one) in the path /letters/girlfriends/Jill/
is not really a good human-machine interface design practice.
But it's way too late to change Unix now; we all just live with it.
(There are actually other reasons for doing it that way [the inventors of
Unix were pretty bright]; but that's way beyond our scope here.) Anyway,
the very first slash in a fully qualified path or file name is actually by
itself the name of a directory, namely the root; while all the other slashes
just separate the names of each successive directory from the next one beneath
it in the path.
- Another way to think of it, is that a slash after the end of a filename
indicates that it is the name of a directory file. In the fully qualified
file name shown earlier, letters/
is a directory, girlfriends/
is a directory, Jill/ is
a directory, April_9.txt
is not a directory. The root, being unique and hence not requiring its
own name is just designated by a bare /
with no other name in front of it.
- If you were silly enough, you could get away with creating and naming a
directory "root".
You might then write /root/ as a path
but all that would mean would be that there was a directory named
"root" within the real root, whose name was really
/
Clear as mud? Probably a good idea not to name any of your directories
"root" — even though
it won't confuse Unix, it would surely confuse anyone you talked to.
In any event, nobody wants to type those
long, fully qualified file names
out all the time so Unix has a convention regarding a "working
directory", also known as the current
directory, and sometimes even as various
combinations of those terms (like, current working directory). At given instant, you and Unix agree on a working directory
(I'll explain how in a moment) and when you enter a command and give it just
an unqualified (or basic or simple) file name like prog1.c,
Unix looks for a file by that name only in the current working directory.
(If there is no file by that name in the current directory, Unix gives you some
sort of error message.) So, referring back to Figure 1, if the current
working directory were CS125,
you could display that file named prog1.c
with the cat command, by entering cat
prog1.c But you can always give a Unix command
a fully qualified file name and it will try to find the file without caring
about the current working directory at all; so the command cat
/CS125/prog1.c
would display the correct file (the one within the CS125
directory) no matter what the current working directory was. (That would
be true even if the current working directory also contained a file whose basic
name was prog1.c
That file in the current directory would not be the one displayed,
/CS125/prog1.c
would be displayed.)
You control the working directory. When you login to
Unix, your initial working directory, before you start moving around, is some
directory created for you by a system administrator when your Unix account was
created. (Your system administrator is said to have created a home
directory for you). The full name of my home directory here on the campus
Unix file system is /facstaff/j/jaffem;
so if I had a file there whose basic name was my_resume.doc,
it's fully qualified name would be:
/facstaff/j/jaffem/my_resume.doc
Anyway, back to directories: You can make new sub directories with the mkdir command. You can change the
current working directory via the cd
command. You can get Unix to tell you the name of the current working directory by
means of the pwd command. You
can get a list of the files (including other directories) within the current directory by
means of the ls command. All these
commands and others are explained more fully in the Basic Unix Shell Commands section. But first,
there's one last type of path name to learn.
A partially qualified
(sometimes just called qualified) or relative
path is a set of nested directories not starting with a slash/
Unix must start following a path from somewhere it knows about. If it
can't start at the root, about the only other place it knows about for sure
is the current working directory.
So AAA/BBB/CCC.dd
would tell Unix to look for a directory named AAA
within the current working directory and then look for directory BBB
within (or beneath) AAA and then look
for the file CCC.dd within BBB.
Thus AAA/BBB/CCC.dd would
be a partially qualified file name with AAA/BBB/as
a relative (or partially qualified) path and CCC.dd
as the basic file name (including the dd
extension). In Figure 1, for example, if the current working directory were
letters, the command
cat girlfriends/jill/June_5.txt
would work But if the current working directory were CS125,
the same command wouldn't work at all — since Unix wouldn't be able
to find the relative path girlfriends/jill/
starting from within CS125.