path: root/doc/filename.n

'\"
'\" Copyright (c) 1995-1996 Sun Microsystems, Inc.
'\"
'\" See the file "license.terms" for information on usage and redistribution
'\" of this file, and for a DISCLAIMER OF ALL WARRANTIES.
'\" 
.so man.macros
.TH filename n 7.5 Tcl "Tcl Built-In Commands"
.BS
'\" Note:  do not modify the .SH NAME line immediately below!
.SH NAME
filename \- File name conventions supported by Tcl commands
.BE
.SH INTRODUCTION
.PP
All Tcl commands and C procedures that take file names as arguments
expect the file names to be in one of three forms, depending on the
current platform.  On each platform, Tcl supports file names in the
standard forms(s) for that platform.  In addition, on all platforms,
Tcl supports a Unix-like syntax intended to provide a convenient way
of constructing simple file names.  However, scripts that are intended
to be portable should not assume a particular form for file names.
Instead, portable scripts must use the \fBfile split\fR and \fBfile
join\fR commands to manipulate file names (see the \fBfile\fR manual
entry for more details).
.SH "PATH TYPES"
.PP
File names are grouped into three general types based on the starting point
for the path used to specify the file: absolute, relative, and
volume-relative.  Absolute names are completely qualified, giving a path to
the file relative to a particular volume and the root directory on that
volume.  Relative names are unqualified, giving a path to the file relative 
to the current working directory.  Volume-relative names are partially
qualified, either giving the path relative to the root directory on the
current volume, or relative to the current directory of the specified
volume.  The \fBfile pathtype\fR command can be used to determine the
type of a given path.
.SH "PATH SYNTAX"
.PP
The rules for native names depend on the value reported in the Tcl
array element \fBtcl_platform(platform)\fR:
.TP 10
\fBUnix\fR
On Unix and Apple MacOS X platforms, Tcl uses path names where the
components are separated by slashes.  Path names may be relative or
absolute, and file names may contain any character other than slash.
The file names \fB\&.\fR and \fB\&..\fR are special and refer to the
current directory and the parent of the current directory respectively.
Multiple adjacent slash characters are interpreted as a single
separator.  Any number of trailing slash characters at the end of a
path are simply ignored, so the paths \fBfoo\fR, \fBfoo/\fR and
\fBfoo//\fR are all identical, and in particular \fBfoo/\fR does not
necessarily mean a directory is being referred.
.RS
.PP
The following examples illustrate various forms of path
names:
.TP 15
\fB/\fR
Absolute path to the root directory.
.TP 15
\fB/etc/passwd\fR
Absolute path to the file named \fBpasswd\fR in the directory
\fBetc\fR in the root directory.
.TP 15
\fB\&.\fR
Relative path to the current directory.
.TP 15
\fBfoo\fR
Relative path to the file \fBfoo\fR in the current directory.
.TP 15
\fBfoo/bar\fR
Relative path to the file \fBbar\fR in the directory \fBfoo\fR in the
current directory.
.TP 15
\fB\&../foo\fR
Relative path to the file \fBfoo\fR in the directory above the current
directory. 
.RE
.TP
\fBWindows\fR
On Microsoft Windows platforms, Tcl supports both drive-relative and UNC
style names.  Both \fB/\fR and \fB\e\fR may be used as directory separators
in either type of name.  Drive-relative names consist of an optional drive
specifier followed by an absolute or relative path.  UNC paths follow the
general form \fB\e\eservername\esharename\epath\efile\fR, but must at
the very least contain the server and share components, i.e. 
\fB\e\eservername\esharename\fR.  In both forms,
the file names \fB.\fR and \fB..\fR are special and refer to the current
directory and the parent of the current directory respectively.  The
following examples illustrate various forms of path names:
.RS
.TP 15
\fB\&\e\eHost\eshare/file\fR
Absolute UNC path to a file called \fBfile\fR in the root directory of
the export point \fBshare\fR on the host \fBHost\fR.  Note that
repeated use of \fBfile dirname\fR on this path will give
\fB//Host/share\fR, and will never give just \fB//Host\fR.
.TP 15
\fBc:foo\fR
Volume-relative path to a file \fBfoo\fR in the current directory on drive
\fBc\fR.
.TP 15
\fBc:/foo\fR
Absolute path to a file \fBfoo\fR in the root directory of drive
\fBc\fR.
.TP 15
\fBfoo\ebar\fR
Relative path to a file \fBbar\fR in the \fBfoo\fR directory in the current
directory on the current volume.
.TP 15
\fB\&\efoo\fR
Volume-relative path to a file \fBfoo\fR in the root directory of the current
volume.
.TP 15
\fB\&\e\efoo\fR
Volume-relative path to a file \fBfoo\fR in the root directory of the current
volume.  This is not a valid UNC path, so the assumption is that the
extra backslashes are superfluous.
.RE
.SH "TILDE SUBSTITUTION"
.PP
In addition to the file name rules described above, Tcl also supports
\fIcsh\fR-style tilde substitution.  If a file name starts with a tilde,
then the file name will be interpreted as if the first element is
replaced with the location of the home directory for the given user.  If
the tilde is followed immediately by a separator, then the \fB$HOME\fR
environment variable is substituted.  Otherwise the characters between
the tilde and the next separator are taken as a user name, which is used
to retrieve the user's home directory for substitution.  This works on
Unix, MacOS X and Windows (except very old releases).
.PP
Old Windows platforms do not support tilde substitution when a user name
follows the tilde.  On these platforms, attempts to use a tilde followed
by a user name will generate an error that the user does not exist when
Tcl attempts to interpret that part of the path or otherwise access the
file.  The behaviour of these paths when not trying to interpret them is
the same as on Unix.  File names that have a tilde without a user name
will be correctly substituted using the \fB$HOME\fR environment
variable, just like for Unix.
.SH "PORTABILITY ISSUES"
.PP
Not all file systems are case sensitive, so scripts should avoid code
that depends on the case of characters in a file name.  In addition,
the character sets allowed on different devices may differ, so scripts
should choose file names that do not contain special characters like:
\fB<>:?"/\e|\fR.
'\""\" reset emacs highlighting
The safest approach is to use names consisting of
alphanumeric characters only.  Care should be taken with filenames
which contain spaces (common on Windows systems) and
filenames where the backslash is the directory separator (Windows
native path names).  Also Windows 3.1 only supports file
names with a root of no more than 8 characters and an extension of no
more than 3 characters.
.PP
On Windows platforms there are file and path length restrictions. 
Complete paths or filenames longer than about 260 characters will lead
to errors in most file operations.
.PP
Another Windows peculiarity is that any number of trailing dots
.QW .
in filenames are totally ignored, so, for example, attempts to create a
file or directory with a name
.QW foo.
will result in the creation of a file/directory with name
.QW foo .
This fact is reflected in the results of \fBfile normalize\fR.
Furthermore, a file name consisting only of dots
.QW .........
or dots with trailing characters
.QW .....abc
is illegal.
.SH "SEE ALSO"
file(n), glob(n)
.SH KEYWORDS
current directory, absolute file name, relative file name,
volume-relative file name, portability