filetypes.txt

操作系统文件管理

z								1998-08-09
								Emil Brink

				gentoo FileTypes

INTRODUCTION
This document describes the gentoo filetyping architecture, i.e. the
means by which gentoo clasfsifies each file it shows as belonging to
exactly one previously defined file type. This classification is then
used to determine how gentoo is to deal with that file; it helps
specifying display colors and/or pixmaps, action commands, and so on.
	In this document, the word "file" is used in a very general
sense. A file is anything that can be found in a disk directory. Since
gentoo runs on Linux (and perhaps other Unices?), this means that a file
can be a directory, a plain file, a (soft) link, a (block or character)
device, a socket, or a FIFO. At least I _think_ that's all...

THE FILETYPE
A filetype is, technically, a named tuple of identification methods.
These methods can be applied to a dirpane line to determine if the disk
entry described by the line belongs to the type or not.
	A filetype can use anything from one to four available "levels"
of identification methods in order to determine if a given line has the
type. These levels are:

1) inode type
This is the simplest, fastest, most low-level check. The inode
type tells if the line is a directory, a soft link, plain file,
device, et cetera. A filetype MUST ALWAYS specify exactly ONE
inode type.

2) Protection
Using the protection matching rule, you can require files to have a
certain combination of protection bits. You can currently only specify
which bits must be set, not which must NOT be set (i.e. clear). This
means that you can find all files that are writable, but not all files
that are NOT writable... I don't know how important this really is,
but it feels kind'a incomplete right now.

3) Suffix Match
The suffix match is used to require that the file name ends in a certain
string. This is very useful to find normal extensions, since a typical
extension is just a suffix with a dot first. For example, you could use
the suffix match to require that candidate lines have names ending in
".gif". This check is highly specialized and therefore fairly quick.
	Since the suffix checker is meant to be lean, mean and quick,
it only allows one suffix rather than a list. If you need to check
against several suffixes (suffici? suffixen?), you'll need to move to
the next leve, the regular expression matcher.

4) Name Regular Expression Match
If the names you're trying to find have some similarity more complex
than just a plain suffix, maybe you can use this level. It lets you
specify a full regular expression (using the V8 syntax found elsewhere
in gentoo) against which file names are checked. For example, if you
where looking for MOD files on an Amiga partition, you might specify
a name RE of "^mod\..*". Or you could identify tar+gzipped files with
"(\.tar\.gz$)|(\.tgz$)".

5) 'file' Regular Expression Match
When none of the previous levels suffice, you can resort to the final
secret weapon; the 'file' regular expression matcher. This runs the
UNIX standard utility 'file' on the dirpane line in question, and then
tries to match the regular expression you supplied against file's
output.
	Using this level, you can perform virtually any kind of file
identification necessary, since you can add things to the "magic" file
/etc/magic used by file.
	As an example, you can identify executable files in the ELF
format (standard on modern Linuxes) by specifying a 'file' matching RE
of e.g. "ELF 32-bit (L|M)SB executable.*", or just plain "ELF.*".
Note that the RE you specify is only checked against the part of file's
output that is after the first colon.
	Please remember that using 'file' RE matching in your file
typing SEVERLY degrades gentoo's performance. Beware. I'm waiting for
a new release of the 'file' package that incorporates a tiny fix which
will allow gentoo to use it with much greater efficiency...


FILETYPING STRUCTURE
Gentoo maintains a set (e.g. a list) of filetype definitions; there is
no hierarchy or any other complex structure involved. The definitions
are applied one by one in search for a match. As soon as any a match is
found, the search is terminated and the line as considered to have the
type that matched.
	If no type matched, the built-in always present type "Unknown"
is assigned to the line, in order to guarantee that all dirpane lines
always have exactly one type assigned to them.
	The ordering of the filetypes internally influences the speed
of the matching process; this is something I will investigate once the
filetyping has been more fully implemented and is somewhat stable.


STYLES
A filetype is just a way of specifying how to identify files that have
things in common. Once the files have been partitioned into groups by
their file types, it would be nice to take advantage of this grouping
when working with the files. This is done through the use of styles.
	A style contains information that gentoo uses when rendering
the dirpane; colors and pixmaps for example. It also holds data used
when you manipulate the file (double-clicking it, or trying to view it
with the built-in View command).
	Styles, unlike filetypes, form a tree structure. There is a
single, always-present, "root" style. Other styles then appear as either
internal or leaf nodes in the tree based in the root style.


STYLE PROPERTIES
The following properties can be specified in a style:

DISPLAY
	Unselected
		Background Color
		Foreground Color
		Pixmap Icon
	Selected
		Background Color
		Foreground Color
		Pixmap Icon
ACTION
	Double-click
	View
	Edit
	Print
	Play


PROPERTY INHERITANCE
The reason why the styles are arranged into a tree structure is to
facilitate inheritance of properties. For each style (except the
root) you specify which properties are to be changed from the ones
of the parent style. Those that are not changed are then taken from
the parent, recursively. The root style always specifies all its
properties.
	If you're into C++ and object-oriented programming in general,
this should be right up your alley.