NAME
Resources - handling application defaults in Perl.
SYNOPSIS
use Resources;
$res = new Resources;
$res = new Resources "resfile";
DESCRIPTION
Resources are a way to specify information of interest to program or packages.
Applications use resource files to specify and document the values of quantities or attributes of interest.
Resources can be loaded from or saved to resource files. Methods are provided to search, modify and create resources.
Packages use resources to hardwire in their code the default values for their attributes, along with documentation for the attibutes themselves.
Packages inherit resources when subclassed, and the resource names are updated dynamically to reflect a class hierarchy.
Methods are provided for interactive resource inspection and editing.
1. Resource Files.
A resource specification in a (text) resource file is a line of the form:
name: value
There may be any number of whitespaces between the name and the colon character, and between the colon and the value.
- name
-
A word not containing whitespaces, colons (':'), dots ('.') or asterisks ('*'), nor starting with '_'.
A sequence in one the following forms:
word.name word*name *name
No distinction is made between uppercase and lowercase letters.
The asterisks in a resource name act as wildcards, matching any sequence of characters.
- value
-
An unadorned word or a quoted sequence of whitespace-separated words. Both single (' ') and double quotes quotes (" ") are allowed, and they must be paired.
Any constant scalar constructor in Perl, including anon references to constant arrays or hashes.
The special words yes, true, no, false (case insensitive) are treated as boolean resources and converted 1 and 0, unless they are quoted.
Examples of valid resource specifications:
hotdog*pickles : plenty # A word.
hotdog.price : 22 # Another word
hotdog.recipe : 'Help me!' # A quoted sentence
hotdog*taste*sucks : yes # A boolean, converted to 1.
hotdog*smell*sucks : 'yes' # yes, aken verbatim
hotdog.size : [1, [2, 3]] # An anon array.
hotdog.lett : {"P"=>1, "Q"=>[2, 3]} # An anon hash.
Examples of illegal resource names:
hot dog # Whitespace in the name.
.hotdog # Leading dot in name.
hot._dog # Leading underscore in _dog.
hotdog* # Trailing asterisk.
hotdog. # Trailing dot.
A resource file may contain comments: anything from a hash ('#') character to the end of a line is ignored, unless the hash character appears inside a quoted value string.
2. Resource hashes
Resource hashes provide a way to specify the default values for the attributes of a package in its source code, along with documentation for the attributes themselves. The documentation itself is "dynamical" (as opposed to the static, pod-like variety) in that it follows a class hyerarchy and is suitable for interactive display and editing.
A resource hash is just a hash of ($Name => $Value) or ($Name => [$Value, $Doc]) things. Each hash key $Name is a resource name as above. Each hash value is either just the resource $Value, or a reference to an anon array [$Value, $Doc], with $Doc being the resource documentation.
The resource $Name cannot contain wildcard ('*') or colon (':') characters, nor start or end with a dot ('.'). Also, it must not be prefixed with the package name (since this is automatically prepended by the merge method, see below). Names starting with an underscore ('_') character are special in that they define "hidden" resources. These may not be specified in resource files, nor dynamically viewed/edited: they come handy to specify global parameters when you dont want to use global application-wide variables, and/or want to take advantage of the inheritance mechanism.
The resource $Value can be any constant scalar Perl constructor, including references to arrays and/or hashes of constants (or references thereof). Boolean values must be specified as 1 or 0.
The resource documentation is a just string of any length: it will be appropriately broken into lines for visualization purposes. It can also be missing for resources defined in base classes, in which case the base class documentation is used (see the merge method below).
The content of a resource hash is registered in a global Resource object using the merge method.
Here is an example of deafults specification for a package.
package HotDog;
@ISA = qw( Food );
%HotDogDefaults = (
taste => ["Yuck!", "Tongue feeling"],
meatcolor => ["Corpse", "Freshness indicator"],
stinks => [1, "Does it smell?"],
onions => [ { on => 2, ions => [3, 5] }, "Rings" ],
'food.good' => 0, # just a scalar, doc is in base class
'_ghost' => [0, "Hidden. Mr. Invisible"]
);
3. Objects and resources
The recommended way to use resources with Perl objects is to pass a Resource object to the "new" method of a package. The method itself will merge the passed resources with the package defaults, and the passed resource will override the defaults where needed.
Resource inheritance via subclassing is then easily achieved, amounting to something like:
package HotDog;
@ISA = qw( Food );
# Called as: $hotdog = new HotDog ($resources);
sub new {
my $package = shift;
my $res = shift;
$res = new Resources if (ref($res) !~ /Resources/o);
$res->merge(\%HotDogDefaults);
my $hd = bless new Food ($res);
# Whatever initialization code
return $hd;
}
4. Methods in class Resources
4.1. Creation and initialization
- new Resources ($resfile);
-
Creates a new resource database, initialized with the defaults for class Resources (see below for a list of them).
If a nonempty file name is specified in $resfile, it initializes the object with the content of the so named resource file. For safe (non overwriting) loading, see the load method below.
If the special file name "_RES_NODEFAULTS" is specified, the object is created completely empty, with not even the Resources class defaults in it.
Returns the new object, or undef in case of error.
- load($resfile, $mode, $nonew);
-
Loads resources from a file named $resfile into a resource database.
The loading is controlled by the $mode argument as follows.
If $mode is omitted, or equal to "replace", the content of the file is loaded as it is, save for syntax checks.
If $mode is equal to "merge", resource merging via merge is performed if the database is not empty. This effectively prevents loaded resources from overwriting those already defined in the database.
The $nonew argument controls whether loading of non already defined resurces is allowed. If it is true, safe loading is performed: attempting to load non-wildcarded resource names that do not match those already present in the database causes an error. This can be useful if you want to make sure that only pre-defined resources (for which you presumably have hardwired defaults) are loaded. It can be a safety net against typos in a resource file.
Use is made of Safe::reval to parse values specified through Perl constructors (only constants, anon hashes and anon arrays are allowed).
Returns 1 if ok, 0 if error.
- merge(\%defaults);
-
Merges the resources of $res with the content of the resource hash %defaults. Only non-wildcarded names are admissible in %defaults. Merging does not change the value of resources already defined in $res.
Note that merging reflects subclassing in that the names of default resources of a base classe are prepended with the name of the caller package (lowercased and stripped of all foo::bar:: prefixes), followed by a dot. In the above example, the defaults of HotDog will be renamed, after merging as:
hotdog.taste, hotdog.meatcolor, ...
and, if Bell::Taco is a new blessed HotDog, then they will be translated again upon merging, to read:
taco.hotdog.taste, taco.hotdog.meatcolor, ...
And so on all the way up to the main package, but with the application name ($0, a.k.a $PROGRAM_NAME in English) prepended in place of "main". See also the documentation of the "byclass" method below for some conditions under which this will not work properly.
The new names are the ones to use when accessing these resources by name.
Again, if a resource is already defined when merge is entered, than its values overrides the defaults. See also the EXAMPLES section.
Returns 1 if OK, 0 if error.
4.2. Looking up resources
The values and documentation strings stored in a Resource object can be accessed by specifying their names in three basic ways:
- directly ("byname" methods)
-
As in "my.nice.cosy.couch" .
- by a pattern ("bypattern" methods)
-
As in "m??nice.*" .
- hierarchically ("byclass" methods)
-
If class My is a Nice, and Nice is a Cosy, and if the current stack of method calls goes from package My to Cosy, then asking for "couch" in package Cosy gets you the value/doc of "my.nice.cosy.couch". This behaviour is essential to proper resource subclassing, as explained in detail below.
The "bypattern" methods must search through the Resource object, while the "byclass" methods must work through call frames and ISA lists. Both ways tend to be slow in case of large resource databases, so the "byvalue" methods should always be used where appropriate. See the EXAMPLE section for more.
It is also possible to retrieve the whole content of a resource database ("names" and "each" methods)
Note that all the resource lookup methods return named (non "wildcarded") resources only. Wildcarded resources (i.e. those specified in resource files, and whose names contain one or more '*') are best thought as placeholders, to be used when the value of an actual named resource is set.
For example, a line in a resource file like
*background : yellow
fixes to yellow the color of all resources whose name ends with "background". However, your actual packages will never worry about unless they really need a background. In this case they either have a "background" resource in their defaults hash, or subclass a package that has one.
- valbyname($name);
-
Retrieves the value of a named resource from a Resource database. The $name argument is a string containing a resource name with no wildcards.
The resource is first looked up with the program name ($0) automatically prepended to the passed name. If this fails, a second search is done using the passed name verbatim.
Returns the undefined value if no such resource is defined.
- docbyname($name);
-
Retrieves the documentation string of a named resource from a Resource database. The $name argument is a string containing a resource name with no wildcards.
The resource is first looked up with the program name ($0) automatically prepended to the passed name. If this fails, a second search is done using the passed name verbatim.
Returns the undefined value if no such resource is defined.
- bypattern($pattern);
-
Retrieves the full names, values and documentation strings of all the named (non wildcarded) resources whose name matches the given $pattern. The pattern itself is string containing a Perl regular expression, not enclosed in slashes.
Returns a new Resource object containing only the matching resources, or the undefined value if no matches are found.
- valbypattern($pattern);
-
Retrieves the full names and values of all named (non wildcarded) resources whose name matches the given pattern.
Returns a new Resource object containing only names and values of the matching resources (i.e. with undefined doc strings), or the undefined value if no matches are found.
- docbypattern($pattern);
-
Retrieves the full names and documentation strings of all named (non wildcarded) resources whose name matches the given pattern.
Returns a new Resource object containing only names and docs of the matching resources (i.e. with undefined resource values), or the undefined value if no matches are found.
- byclass($suffix);
-
Resources are naturally ordered with respect to a hierarchy of classes, in that if two resource names share a common suffix, then the resource defined in the subclass overrides the one of a base class (unless the former is wildcarded in a resource file).
For example, if class HotDog has a default for a resource named "hotdog.pickles", and class Taco (a subclass of HotDog) has a default for "taco.hotdog.pickles", then we expect that the latter should override the former in a Taco object.
Note however that we do not want "mc.donalds.hotdog.pickles" to override "taco.hotdog.pickles", for a Mc class unrelated to Taco. This means that finding the right default in the database is not just a matter of counting dots, but of following a class hierarchy,
This method returns a 3 element list containing the name/value/doc tuple of the the most subclassed resource from the current class, among those having a common suffix. In the above example, a call like
package HotDog ... ($name,$value,$doc) = $res->byclass("pickles");
will set $name, $value and $doc equal to those of the "taco.hotdog.pickles" resource, if this HotDog is subclassed by Taco, and to those of "mc.donalds.hotdog.pickles", if it is subclassed by Mc via Donalds instead.
To do this, the algorithm walks upward in the current subroutine call stack, and looks up the ISA list of each caller package to determine whether the callers are subclasses or not. Since it depends on the current call stack, it will work only when calls are really nested, i.e. with inherited methods or explicitly nested calls (e.g. in the "new" method of subclasses).
The passed name suffix must not contain wildcards, nor start with a dot.
Be careful not to confuse the "byclass" with the "byname" and "bypattern" retrieval methods: they are used for two radically different goals. See the EXAMPLES section for more.
Returns the empty list if no resources are found for the given suffix, or if the suffix is incorrect.
- namebyclass($suffix);
-
As the byclass method above, but returns just the resource name (i.e. the suffix with all the subclasses prepended).
- valbyclass($suffix);
-
As the byclass method above, but returns just the resource value.
- docbyclass($suffix);
-
As the byclass method above, but returns just the resource documentation.
- each;
-
Returns the next name/[value,doc] pair of the named (non wildcarded) resources in a resource database, exactly as the each Perl routine.
- names;
-
Returns a list of the names of all named (non-wildcarded) resources in a resource database, or undef if the databasee is empty.
4.3. Assigning and removing Resources
- put($name, $value, $doc);
-
Writes the value and doc of a resource in the database. It is possible to specify an empty documentation string, but name and value must be defined.
Wildcards ('*' characters) are allowed in the $name, but the $doc is ignored in this case (documentation is intended for single resources, not for sets of them).
The value is written unchanged unless the resource database already contains a wildcarded resource whose name includes $name (foo*bar includes foo.bar, foo.baz.bar, etc.). In this case the value of the wildcarded resource overrides the passed $value.
Returns 1 if ok, 0 if error.
- removebyname($name);
-
Removes the named (non wildcarded) resources from the database.
Returns 1 if OK, 0 if the resource is not found in the database.
- removebypattern($name);
-
Removes from the database all resources (both named and wildcarded) whose name mathes $pattern. An exactly matching name must be specified for wildcarded resources (foo*bar to remove foo*bar).
Returns the number of removed resources.
4.6. Viewing and editing resources.
- view;
-
Outputs the current content of a Resource object by piping to a pager program.
The environment variable $ENV{RESPAGER}, the resource "resources.pager" and the environment variable $ENV{PAGER} are looked up, in this very order, to find the pager program. Defaults to /bin/more if none of them is found.
The output format is the same of a resource file, with the resource names alphabetically ordered, and the resource documentation strings written as comments.
Returns 1 if ok, 0 if error.
- edit($nonew);
-
Provides dynamical resource editing of a Resource object via an external editor program. Only resource names and values can be edited (anyway, what is the point of editing a resource comment on the fly?).
The environment variables $ENV{RESEDITOR}, the resource "resouces.editor", the environment variables $ENV{VISUAL} and $ENV{EDITOR} are looked up, in this very order, to find the editor program. Defaults to /bin/vi if none is found.
The editor buffer is initialized in the same format of a resource file, with the resource names alphabetically ordered, and the resource documentation strings written as comments. The temporary file specified by the "resources.tmpfil" resource is used to initialize the editor, or '/tmp/resedit<pid>' if that resource is undefined.
When the editor is exited (after saving the buffer) the method attempts to reload the edited resources. If an error is found the initial object is left unchanged, a warning with the first offending line in the file is printed, and the method returns with undef. Controlled resource loading is obtained by specifying a true value for the $nonew argument (see load).
If the loading is successful, a new (edited) resource object is returned, which can be assigned to the old one for replacement.
After a successful edit, the value of the resource "resources.updates" (which is always defined to 0 whenever a new resource is created) is increased by one. This is meant to notify program the and/or packages of the resource change, so they can proceed accordingly if they wish.
4.5. Miscellaneous methods
- write($filename);
-
Outputs all resources of a resource database into a resource file (overwriting it).
The resource documentation strings are normally written as comments, so the file itself is immediately available for resource loading. However, if the boolean resource "resources.writepod" is true, then the (non wildcarded) resources are output in POD format for your documentational pleasure.
As usual in Perl, the filename can allo be of the form "|command", in which case the output is piped into "comma1nd".
For resources whose value is a reference to an anon array or hash, it produces the appropriate constant Perl contructor by reverse parsing. The parser itself is available as a separate method named _parse (see package source for documentation).
Returns 1 if ok, 0 if error.
5. Resources of Resources
As you may have guessed at this point, the default configuration of this package itself is defined by resources. The resource class is, of course, "resources" (meaning that all the names have a leading "resource.").
To prevent chaos, however, these resources cannot be subclassed. This should not be a problem in normal applications, since the Resource package itself is not meant to be subclassed, but to help building a hierarchy of classes instead.
The currently recognized resources, and their default values, are:
- resources.appname : ''
-
The application name of this Resource object.
- resources.editor : /bin/vi
-
Resource editor.
- resources.pager : /bin/more
-
Resource pager.
- resources.separator : ':'
-
Pattern separating names from values in resource files.
- resources.tmpfil : ''
-
Editor temporary file.
- resources.updates : 0
-
Number of resource updates.
- resources.verbosity : 1
-
True to print warnings.
- resources.viewcols : 78
-
Width of view/edit window.
- resources.writepod : false
-
Boolean. True if the write method should output in POD format.
EXAMPLES
Here is a more complex example of hyerarchical inheritance of resources. Package HotDog is subclassed from Junk and Food. The subclass has defaults for two resources defined by the base classes ("food.edible" and "junk.germs"), and their values will override the base class defaults.
Remember that after merging all resources names are prefixed with the current class name.
use Resources;
package Food;
%FoodDefaults = ( edible => "Sure" );
sub new {
my ($pack, $res) = @_;
$res = new Resources unless $resources;
$res->merge(\%FoodDefaults) || die ("can't merge defaults");
my $food= bless {};
$food->{Edible} = $res->valbyclass("edible");
# Use valbyclass so a subclass like HotDog can change this by its
# defaults.
}
# A Food method to say if it can be eaten.
sub eat {
my $food=shift;
return $food->{Edible};
}
package HotDog;
use Junk;
use Food;
use Carp;
@ISA = qw( Junk Food );
%HotDogDefaults =
(
taste => "Yuck!",
meatcolor => "Cadaveric",
stinks => 1,
'food.edible' => "Perhaps", # Quotes needed in these
'junk.germs' => "Amoeba" # names because of the dots
);
sub new {
my ($package, $res) = @_;
$res = new Resources unless $resources;
$res->merge(\%HotDogDefaults) || die ("can't merge defaults");
my $food = new Food ($res); # Override Food's edible.
my $junk = new Junk ($res); # Override Junks's germs.
my $hd = bless { %{food}, %{junk} }; # Merge and subclass.
$hd->{Meatcolor} = $res->valbyclass("meatcolor");
# Same reason as above
return $hd;
}
# A HotDog method to get a taste via resources
sub taste {
my ($hd, $res) = @_;
print $res->valbyname("hotdog.taste");
}
package main;
# Whatever
#
$res = new Resources("AppDefltFile") || die;
$snack = new HotDog($res);
$gnam = $snack->eat(); # hotdog.food.edible overridees food.edible,
# so here $gnam equals "Perhaps"
$slurp = $snack->taste($res) # $slurp equals "Yuck!", unless
# this resource too was overridden
# by a subclass of HotDog , or
# differently specified in
# "AppDefltFile"
Note the different use of valbyclass and valbyname to access resources:
- valbyclass
-
Is used for the resources like "edible" in Food, i.e. resources which are defined by the current class in its defaults. This because a subclass of it (e.g. HotDog) might have overridden them (by having an entry for "food.edible" in its defaults.
Generally speaking, you should not use valbyclass outside the new (creation and initialization) method of a class.
- valbyname
-
Is used after the initialization, when all subclassed resources have been resolved nicely. In any case, its arguments is the complete resource name, with all its dots.
SEE ALSO
Safe(3).
BUGS
The underlying idea is to use a centralized resource database for the whole application. This ensures uniformity of behaviour across kin objects, but allow special characterizations only at the cost of subclassing.
AUTHOR
Francesco Callari <franco@cim.mcgill.ca>
Artifical Perception Laboratory,
Center for Intelligent Machines,
McGill University.
WWW: http://www.cim.mcgill.ca/~franco/Home.html
COPYRIGHT
Copyright 1996 Francesco Callari, McGill University
Permission to use, copy, modify, and distribute this software and its documentation for any purpose without fee is hereby granted without fee, provided that the above copyright notice appear in all copies and that both that copyright notice and this permission notice appear in supporting documentation, and that the name of McGill not be used in advertising or publicity pertaining to distribution of the software without specific, written prior permission. McGill makes no representations about the suitability of this software for any purpose. It is provided "as is" without express or implied warranty.
MCGILL DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL MCGILL BE LIABLE FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
10 POD Errors
The following errors were encountered while parsing the POD:
- Around line 857:
You forgot a '=back' before '=head2'
- Around line 859:
'=item' outside of any '=over'
- Around line 992:
You forgot a '=back' before '=head2'
- Around line 994:
'=item' outside of any '=over'
- Around line 1106:
You forgot a '=back' before '=head2'
- Around line 1108:
'=item' outside of any '=over'
- Around line 1482:
You forgot a '=back' before '=head2'
- Around line 1495:
'=item' outside of any '=over'
- Around line 1620:
'=item' outside of any '=over'
- Around line 1636:
You forgot a '=back' before '=head1'