xara - Online in the Cloud

PROGRAM:

NAME

xara - GTK2 interface for the above

SYNOPSIS

Graphical interface (GTK2):
xara

The graphical interface allows the user to input queries and browse the results. Menu
options are provided for installing and removing the selected packages using apt-get. The
packages the user is interested in may be bookmarked.

Command-line interface (GTK2): A command-line interface, ara(1), is provided by the Debian
package ara.

Query syntax
See the EXAMPLES section for a quick introduction ; xara has some built-in help. The
syntax is described in detail below.

DESCRIPTION

ara and xara allow the user to search the Debian software package database (which includes
installed and uninstalled packages) using powerful queries made of boolean combinations of
regular expressions acting on fields given by patterns.

For example, the query section=utils & depends:(gtk or tk8 or xlibs or kde or gnome or qt)
& debian & package will display packages in the section utils that have graphical
interfaces (because they depend on graphical toolkits or X11 libraries), and whose
description contains the words debian and package.

RATIONALE

Debian users can easily install software with the commands dselect or apt-get install.
They can choose (on Debian 3.1 unstable) from over 30,000 packages. Finding the right
package can be quite difficult. Although packages are categorized in crude sections,
there are still too many packages and reading all descriptions is out of the question.

The database files are huge and their mail-like syntax makes them hard to search with
line-oriented tools like grep. There exist commands such as dpkg-iasearch(1) or
dpkg-dctrl(1) but their capabilities are limited. Graphical package management tools such
as aptitude or synaptic have search capabilities. Although ara can call apt to install or
remove packages, its orientation is that of a powerful search tool. Indeed, the name ara
comes from the imperative form of the Turkish verb aramak which means "to search".

THE DEBIAN PACKAGE DATABASE

The database of Debian packages is a huge text file at /var/lib/dpkg/available (or a
collection of text files under /var/lib/apt/lists/). These files are in a mailbox-like
format, and a typical entry looks like this:

Priority: required
Section: base
Installed-Size: 460
Origin: debian
Maintainer: Dpkg Development <[email protected]>
Bugs: debbugs://bugs.debian.org
Architecture: i386
Source: dpkg
Version: 1.10.24
Replaces: dpkg (<< 1.10.3)
Depends: libc6 (>= 2.3.2.ds1-4), ....
Filename: pool/main/d/dpkg/dselect_1.10.24_i386.deb
Size: 119586
MD5sum: c740f7f68dab08badf4f60b51a33500a
Description: a user tool to manage Debian packages
dselect is the primary user interface for installing, removing and
managing Debian packages. It is a front-end to dpkg.

Each package is thus described by a set of fields (like Package, Description, Version...).

QUERY SYNTAX AND SEMANTICS

Here we describe the query syntax in some detail. As of version 1.0, ara introduces new,
simplified syntax which is quite traditional and should be familiar to anyone having used
search engines. Search terms are simply combined with AND, OR and NOT boolean operators.
Having a look at the EXAMPLES section at the end of this manual should provide you a
starting point.

Consider the set D of Debian package descriptions contained in the file
/var/lib/dpkg/available (or in files under /var/lib/apt/lists/). Each description is a
set of couples of the form (f,v) where f and v are strings: f is the name of the field
(namely, Package, Description, Filename, Depends, etc.); v is its value. Thus D is a set
of set of couples, forming the universe. Queries select subsets of the universe D.
Output options select which fields of the selected part of the universe to display, and
how to display them.

Queries
A query is a boolean combination of atomic expressions. An atomic expression selects a
subset of the set D of descriptions. I call this set the meaning of the expression; if e
denotes an atomic expression, its meaning is denoted by [e]. The meaning of a boolean
combination of atomic expressions is just the boolean combination of the meaning of its
constituents. In other words, if e1 and e2 are atomic expressions, then e1 & e2 is a
query, whose meaning is the intersection of the meanings of e1 and e2; and the meaning of
e1 | e2 is the union of the meanings of e1 and e2.

Atomic expressions
Atomic expressions can be of the forms pattern, /regexp/, quoted_string, fieldspec
operator1 string, or fieldspec operator2 regexp.

Boolean operators and constants
e1 & e2 (also e1 AND e2, e1 and e2)
This is logical conjunction (set intersection). Returns the intersection of [e1]
and [e2], i.e. packages satisfying both e1 and e2.

e1 | e2 (also e1 OR e2, e1 or e2)
This is logical disjunction (set union). Union of [e1] and [e2], i.e. packages
satisfying e1, e2 or both.

!e1 (also NOT e1, not e1)
This is logical negation (set complementation). Complement of [e1], i.e. packages
not satisfying e1.

Please note that ~ stands for the current default field specifier and is not an
alias for the complementation operator.

true (also all)
The set of all descriptions, i.e. all packages.

false (also none)
The empty set, i.e. no packages.

Field specifiers
A field specifier fieldspec is a comma-separated list of field patterns.

Field patterns are like simple shell patterns and they may contain star characters (which
stand for anything) or question marks (which stand for any single character). They are
case-insensitive. They specify a set of fields.

For example description and Description specify the set of fields { Description }, whereas
de* specifies { Description, Depends }.

The special specifier ~ denotes the current default specifier (see below).

Current fields specifiers and simplified atomic expressions
The need to repeat the field specifier can make the above syntax cumbersome. That is why
there is a current field specifier. The current field specified is, by default,
Description,Package. Simplified atomic expressions are simply words or simplified shell
expressions (which do not need to be enclosed in double quotes) and they are searched in
fields in the current field specifier. They can be made of letters, digits, underscores,
dashes and periods. They may contain stars of question marks which are interpreted as for
field patterns (i.e., as simplified shell expressions). If double quotes are used, other
characters and spaces can be used.

The default field specifier in a query query can be changed to fieldspec by simply
prefixing the query with fieldspec:. This gives fieldspec:query. However if query is
complex (i.e., contains binary boolean operators) you need to enclose query in
parentheses, as in fieldspec:(query1 or query2).

String literals
String literals can be given with or without double quotes; without double quotes, the
syntax is as for C identifiers, except that you can use dashes, you must start with a
latin letter ([a-zA-Z]) and you can continue with Latin letters, decimal digits or
underscore ([a-zA-Z0-9_]). Inside double quotes, all characters are allowed, except
double quotes, which must be preceded by a backslash.

Variables
Results of queries can be stored in variables, which may be recalled later. This isn't
very useful in batch mode but is useful in interactive and graphical modes.

Variable names start with a dollar and follow usual conventions for variables, i.e., they
can be any mix of alphanumeric characters and symbols such as underscore, dash, etc.

Variable names are case-sensitive so that $Installed and $installed are different.

To assign the result of a query (which is a set of packages) a variable named $variable
just execute the query $variable := query. You may then recall this particular set by
simply writing $variable.

Example: $installed := status:(installed & !not-installed)

Operators
Hierarchical comparison operators can be negated by changing the direction of the angle
brackets and adding or removing an equality sign at end (<= becomes >). Other operators
are negated as follows: = becomes != and =~ becomes !~.

fieldspec=string
Atomic expression selecting packages having a field in fieldspec having a value a
value exactly equal to string.

fieldspec<string (fieldspec<=string, fieldspec>string, fieldspec>=string)
Atomic expression selecting packages having a field in fieldspec whose value is
strictly less than string. The order used is the Debian versioning order. This
order is compatible with the natural order on integers and with Debian version
numbers. When comparing strings not containing special characters, letters sort
before numbers, as opposed to lexicographic ASCII order we are used to. This means
that hexadecimal numbers (such as MD5 sums) will not have their usual order.

Note that string must be on the right side of the operator (i.e., you cannot write
1000 < Size).

fieldspec=~/expression/ (also fieldspec:/expression/)
Selects descriptions whose field named fieldspec exists and whose value matches,
case-sensitively, the regular expression expression.

fieldspec=~/expression/i (also fieldspec:/expression/i)
Same as above, but the regular expression is case-insensitive.

fieldspec=~/expression/w (also fieldspec:/expression/w)
Same as above, but the regular expression is case-sensitive and matches only at
word boundaries. Note that letters-to-digit or digit-to-letter transitions are
considered to be word boundaries.

fieldspec=~/expression/iw (also fieldspec:/expression/iw)
The regular expression here is case-insensitive and matched at word boundaries.

Regular expressions
Regular expressions are given between a pair of slashes; the last slash can be followed by
a commutative sequence of letters denoting flags. Regular expression syntax is sed-like:
grouping parentheses and alternation must be backslashed. For more details, see the
Objective Caml manual chapter on the Str module. In short (x,x1,x2 are meta-symbols
denoting regular expressions):

/./ Any character.

/toto/ Literal string toto.

/x1x2/ Concatenation.

/x1\|x2/
Alternation.

\(x1\)*
Star closure.

[c-d] Character range.

\b Word boundaries.

/x/i Case insensitive.

/x/w At word boundaries.

Remark
Most queries will contain an appreciable amount of shell metacharacters. For example,
logical disjunction is denoted by the pipe character, which is used by all known shells.
The problem is aggravated by the fact that names of real commands are likely to appear in
the used expressions; successfully setting up a UNIX pipeline by error is therefore
plausible.

When calling ara from the command line in batch mode, You are strongly urged to protect
your queries by surrounding them with simple quotes; never write something like ara
Pack*=~/halt|reboot|shutdown/ as this will very likely reboot your system (and is
incorrect regular expression syntax, if halt or reboot or shutdown is meant: pipes must be
backslashed). Instead, one should write ara 'Pack*=~/halt\|reboot\|shutdown /'

OPTIONS

None.

EXAMPLES

Section=utils
List the name of every package in section utils.

Section=utils and !Depends:(gnome|kde|gtk)
... except those whose dependency field matches the regexp gnome\|kde\|gtk

Section=utils and Status:(installed & !not-installed)
List all installed packages in section utils.

section=utils and !depends:(gtk|gnome|kde) and priority=optional
... show only optional packages.

section=utils & (!depends:(gtk|gnome|kde) | size<100000) & priority=optional
Well, exclude gtk,gnome or kde stuff only if 100000 bytes or greater.

Section=games and not (Depends:(gtk|sdl|kde|opengl|gnome|qt)
or /shoot\|kill\|destroy\|blast\|race\|bomb/iw
or /multi\(-\|\)player\|strategy\|conquest\|3\(-\|\)d/iw)
and Depends:(xlibs or vga)
and Size <= 1000000

Display all packages in the games section whose size does not exceeding one million bytes,
and which do not depend on fancy stuff like GTK, SDL, KDE, OpenGL, Qt or Gnome, do not
mention some form of violence (to shoot, to kill, etc.) in their description, are not
described as multi-player, strategy, conquest or three-dimensional, and yet depend on
either xlibs or svga to exclude console-based games.

SPEED

xara reads the whole database into memory and then processes queries. Since the database
is usually big, this takes some time. However, queries then run quite fast. So specify
multiple queries or use the -interactive option to amortize the cost of reading the
database.

Use xara online using onworks.net services