Manual of Lua-URI: file
lua-uri-file - File URI support for Lua URI library
Description
The class uri.file
is used for URIs with the file
scheme. It inherits
from the uri class.
A file URI without an authority doesn't have a well defined meaning. This
library considers such URIs to be invalid when they have a path which does not
start with '/' (for example file:foo/bar
). It is likely that any such URI
should really be a relative URI reference. If the path does start with a slash
then this library will attempt to 'repair' the URI by adding an empty authority
part, so that file:/foo/bar
will be changed automatically to
file:///foo/bar
.
A host value of localhost
is normalized to an empty host, so that
file://localhost/foo
will become file:///foo
. An empty path is
normalized to '/'.
The path part is always considered to be case sensitive, so no case folding is done even when converting to a filesystem path for Windows.
Query parts and fragments are left alone by this library, but are not used in converting URIs to filesystem paths.
Converting between URIs and filesystem paths
A uri.file
object can be converted into an absolute path suitable for
use on a particular operating system by calling the filesystem_path
method:
local uri = assert(URI:new("file:///foo/bar"))
print(uri:filesystem_path("unix")) -- /foo/bar
print(uri:filesystem_path("win32")) -- \foo\bar
This method will throw an exception if the path cannot be converted. For example, a file URI containing a host name cannot be represented on a Unix filesystem, but on a Win32 system it will be converted to a UNC path:
local uri = assert(URI:new("file://server/path"))
print(uri:filesystem_path("unix")) -- error
print(uri:filesystem_path("win32")) -- \\server\path
To convert a filesystem path into a URI, call the class method
make_file_uri
:
local FileURI = require "uri.file"
local uri = FileURI.make_file_uri("/foo/bar", "unix")
print(uri) -- file:///foo/bar
uri = FileURI.make_file_uri("C:\foo\bar", "win32")
print(uri) -- file:///C:/foo/bar
To convert a relative URI reference (a uri._relative
object) into a filesystem path you should first resolve it against an
appropriate file
URI, and then call the filesystem_path
method on that.
Methods
All the methods defined in lua-uri(3) are supported. The userinfo
,
and port
methods will always return nil, and will throw an
exception when passed anything other than nil. The host
method will
normalize localhost
to an empty host name, and will throw an exception if
given a new value of nil. The path
method will normalize an empty path
or nil value to '/'.
In addition to the standard methods, file URIs support the filesystem_path
method, and the uri.file
class contains the make_file_uri
function,
both of which are described above.
Operating systems supported
The conversion between a file URI and a path suitable for use on a particular
operating system are defined in additional classes, which are loaded
automatically based on the operating system name supplied to the two conversion
functions. For example, passing the string win32
to the functions will
invoke the implementation in the class uri.file.win32
. An exception will be
thrown if no class exists to support a given operating system. The following
operating system classes are provided:
uri.file.unix
A URI containing a host name will cause an exception to be thrown, as there is no obvious way for these to be represented in Unix paths. If the path contains an encoded null byte (
%00
) or encoded slash (%2F
) then an exception will be thrown.Attempting to convert a relative path to a URI will cause an exception.
uri.file.win32
Forward slashes ('/') in URIs will be converted to backslashes ('\') in paths, and vice versa.
URIs containing host names will be converted to UNC paths, starting with a '\\' followed by the hostname and then the path part. If the path part of a URI appears to begin with a drive letter, then the first slash will be removed so that the resulting path starts with the letter. Encoded pipe characters ('%7C') will be recognized as equivalent to colons for the purpose of identifying drive letters, since they have been historically used in that way, but I believe they are not allowed to occur in the path unencoded in a URI nowadays.
The operating system names are case insensitive, and are folded to lowercase before being converted into a Lua module name.
Currently there is no way for this library to recognise the operating system it is running on, since Lua has no built-in way of providing that information.
References
The most up to date IETF standard for the file
URI scheme is still
RFC 1738 section 3.10, but this does not specify exactly how to convert
between URIs and filesystem paths on particular platforms. It does however
specify the equivalence between 'localhost' and an empty host.
The correct form of file URI to represent a Windows filesystem path is described in a blog article: http://blogs.msdn.com/ie/archive/2006/12/06/file-uris-in-windows.aspx
There is a standard of sorts describing the conversion between Unix paths and file URIs: http://equinox-project.org/spec/file-uri-spec.txt