全部博文（184）
分类：
2009-08-05 12:26:59
HTML::Mason::Params - Mason configuration parameters
This document lists all of the Mason configuration parameters that are intended to be used by end users.
Each parameter has two names: a Perl version and an Apache version. The Perl version uses lowercase_with_underscores, while the Apache version uses StudlyCaps with a Mason prefix. The conversion from one version to the other is otherwise very predictable. For example,
autohandler_name <--> MasonAutohandlerName
comp_root <--> MasonCompRoot
data_cache_defaults <--> MasonDataCacheDefaults
The Apache parameter names are used in the Apache configuration file in an .
The Perl parameter names are used from Perl code, i.e. anywhere other than the Apache configuration file. For example,
In an custom wrapper-based configuration, you can pass most of these parameters to the constructor.
In a , you can pass most of these parameters to the constructor.
When launching a , you can pass any of the HTML::Mason::Request parameters to .
List of variable names, complete with prefix ($@%), that you intend to use as globals in components. Normally global variables are forbidden by strict, but any variable mentioned in this list is granted a reprieve via a "use vars" statement. For example:
allow_globals => [qw($DBH %session)]
In a mod_perl environment, $r (the request object) is automatically added to this list.
Title that you want this ApacheHandler to appear as under Apache::Status. Default is "HTML::Mason status". This is useful if you create more than one ApacheHandler object and want them all visible via Apache::Status.
Method to use for unpacking GET and POST arguments. The valid options are 'CGI' and 'mod_perl'; these indicate that a CGI.pm or Apache::Request object (respectively) will be created for the purposes of argument handling.
'mod_perl' is the default under mod_perl-1 and requires that you have installed the Apache::Request package. Under mod_perl-2, the default is 'CGI' because Apache2::Request is still in development.
If args_method is 'mod_perl', the $r global is upgraded to an Apache::Request object. This object inherits all Apache methods and adds a few of its own, dealing with parameters and file uploads. See Apache::Request for more information.
If the args_method is 'CGI', the Mason request object ($m) will have a method called cgi_object available. This method returns the CGI object used for argument processing.
While Mason will load Apache::Request or CGI as needed at runtime, it is recommended that you preload the relevant module either in your httpd.conf or handler.pl file, as this will save some memory.
True or false, default is true. Indicates whether Mason should automatically send HTTP headers before sending content back to the client. If you set to false, you should call $r->send_http_header manually.
See the section of the developer's manual for more details about the automatic header feature.
NOTE: This parameter has no effect under mod_perl-2, since calling $r->send_http_header is no longer needed.
True or false, default is false. Indicates whether to flush the output buffer ($m->flush_buffer) after every string is output. Turn on autoflush if you need to send partial output to the client, for example in a progress meter.
As of Mason 1.3, autoflush will only work if has been set. Components can be compiled more efficiently if they don't have to check for autoflush. Before using autoflush you might consider whether a few manual $m->flush_buffer calls would work nearly as well.
File name used for . Default is "autohandler". If this is set to an empty string ("") then autohandlers are turned off entirely.
Number of bytes to preallocate in the output buffer for each request. Defaults to 0. Setting this to, say, your maximum page size (or close to it) can reduce the number of reallocations Perl performs as components add to the output buffer.
Specifies the maximum number of components that should be held in the in-memory code cache. The default is 'unlimited', meaning no components will ever be discarded; Mason can perform certain optimizations in this mode. Setting this to zero disables the code cache entirely. See the section of the administrator's manual for further details.
The class into which component objects are blessed. This defaults to .
The component root marks the top of your component hierarchy and defines how component paths are translated into real file paths. For example, if your component root is /usr/local/httpd/docs, a component path of /products/index.html translates to the file /usr/local/httpd/docs/products/index.html.
Under and , comp_root defaults to the server's document root. In standalone mode comp_root defaults to the current working directory.
This parameter may be either a scalar or an array reference. If it is a scalar, it should be a filesystem path indicating the component root. If it is an array reference, it should be of the following form:
[ [ foo => '/usr/local/foo' ], [ bar => '/usr/local/bar' ] ]
This is an array of two-element array references, not a hash. The "keys" for each path must be unique and their "values" must be filesystem paths. These paths will be searched in the provided order whenever a component path is resolved. For example, given the above component roots and a component path of /products/index.html, Mason would search first for /usr/local/foo/products/index.html, then for /usr/local/bar/products/index.html.
The keys are used in several ways. They help to distinguish component caches and object files between different component roots, and they appear in the title() of a component.
When you specify a single path for a component root, this is actually translated into
[ [ MAIN => path ] ]
If you have turned on , you may modify the component root(s) of an interpreter between requests by calling $interp->comp_root with a value. However, the path associated with any given key may not change between requests. For example, if the initial component root is
[ [ foo => '/usr/local/foo' ], [ bar => '/usr/local/bar' ], ]
then it may not be changed to
[ [ foo => '/usr/local/bar' ], [ bar => '/usr/local/baz' ],
but it may be changed to
[ [ foo => '/usr/local/foo' ], [ blarg => '/usr/local/blarg' ] ]
In other words, you may add or remove key/path pairs but not modify an already-used key/path pair. The reason for this restriction is that the interpreter maintains a component cache per key that would become invalid if the associated paths were to change.
The class to use when creating a compiler. Defaults to .
A code reference used to handle errors thrown during component compilation or runtime. By default, this is a subroutine that turns non-exception object errors in components into exceptions. If this parameter is set to a false value, these errors are simply rethrown as-is.
Turning exceptions into objects can be expensive, since this will cause the generation of a stack trace for each error. If you are using strings or unblessed references as exceptions in your code, you may want to turn this off as a performance boost.
The $m->cache API to use:
'1.1', the default, indicates a Cache::Cache based API.
'chi' indicates a CHI based API.
'1.0' indicates the custom cache API used in Mason 1.0x and earlier. This compatibility layer is provided as a convenience for users upgrading from older versions of Mason, but will not be supported indefinitely.
A hash reference of default options to use for the $m->cache command. For example, to use Cache::Cache's MemoryCache implementation by default:
data_cache_defaults => {cache_class => 'MemoryCache'}
To use the CHI FastMmap driver by default:
data_cache_api => 'CHI',
data_cache_defaults => {driver => 'FastMmap'},
These settings are overriden by options given to particular $m->cache calls.
The data directory is a writable directory that Mason uses for various features and optimizations: for example, component object files and data cache files. Mason will create the directory on startup, if necessary, and set its permissions according to the web server User/Group.
Under , data_dir defaults to a directory called "mason" under the Apache server root. You will need to change this on certain systems that assign a high-level server root such as /usr!
In non-Apache environments, data_dir has no default. If it is left unspecified, Mason will not use , and the default will be MemoryCache instead of FileCache.
True or false, default is true. Indicates whether Mason should decline directory requests, leaving Apache to serve up a directory index or a FORBIDDEN error as appropriate. See the section of the administrator's manual for more information about handling directories with Mason.
Escape flags to apply to all <% %> expressions by default. The current valid flags are
h - escape for HTML ('<' => '<', etc.)
u - escape for URL (':' => '%3A', etc.)
The developer can override default escape flags on a per-expression basis; see the section of the developer's manual.
If you want to set multiple flags as the default, this should be given as a reference to an array of flags.
One of "always", "auto", or "never". This determines whether or not an %ARGS hash is created in components. If it is set to "always", one is always defined. If set to "never", it is never defined.
The default, "auto", will cause the hash to be defined only if some part of the component contains the string "ARGS". This is somewhat crude, and may result in some false positives, but this is preferable to false negatives.
Not defining the args hash means that we can avoid copying component arguments, which can save memory and slightly improve execution speed.
File name used for . Default is "dhandler". If this is set to an empty string ("") then dhandlers are turned off entirely.
True or false, defaults to false. Indicates whether the can be modified on this interpreter between requests. Mason can perform a few optimizations with a fixed component root, so you should only set this to true if you actually need it.
True or false, default is true. Indicates whether components are compiled with support for . The component can be compiled to a more efficient form if it does not have to check for autoflush mode, so you should set this to 0 if you can.
Indicates how errors are formatted. The built-in choices are
brief - just the error message with no trace information
text - a multi-line text format
line - a single-line text format, with different pieces of information separated by tabs (useful for log files)
html - a fancy html format
The default format under and is either line or html depending on whether the error mode is fatal or output, respectively. The default for standalone mode is text.
The formats correspond to HTML::Mason::Exception methods named as_format. You can define your own format by creating an appropriately named method; for example, to define an "xml" format, create a method HTML::Mason::Exception::as_xml patterned after one of the built-in methods.
Indicates how errors are returned to the caller. The choices are fatal, meaning die with the error, and output, meaning output the error just like regular output.
The default under and is output, causing the error to be displayed in the browser. The default for standalone mode is fatal.
A hash reference of escape flags to set for this object. See the section on the for more details.
Regular expression indicating which warnings to ignore when loading components. Any warning that is not ignored will prevent the component from being loaded and executed. For example:
ignore_warnings_expr =>
'Global symbol.*requires explicit package'
If set to undef, all warnings are heeded. If set to '.', warnings are turned off completely as a specially optimized case.
By default, this is set to 'Subroutine .* redefined'. This allows you to declare global subroutines inside <%once> sections and not receive an error when the component is reloaded.
This is the package in which a component's code is executed. For historical reasons, this defaults to HTML::Mason::Commands.
The class to use when creating a interpreter. Defaults to .
The class to use when creating a lexer. Defaults to .
The maximum recursion depth for the component stack, for the request stack, and for the inheritance stack. An error is signalled if the maximum is exceeded. Default is 32.
When compiling a component, use uniquely named subroutines for the a component's body, subcomponents, and methods. Doing this allows you to effectively profile Mason components. Without this, all components simply show up as __ANON__ or something similar in the profiler.
Extension to add to the end of object files. Default is ".obj".
Indicates where to send output. If out_method is a reference to a scalar, output is appended to the scalar. If out_method is a reference to a subroutine, the subroutine is called with each output string. For example, to send output to a file called "mason.out":
my $fh = new IO::File ">mason.out";
...
out_method => sub { $fh->print($_[0]) }
By default, out_method prints to standard output. Under , standard output is redirected to $r->print.
An array of plugins that will be called at various stages of request processing. Please see for details.
Text given for this parameter is placed at the end of each component. See also . The request will be available as $m in postamble code.
Sub reference that is called to postprocess the Perl portion of a compiled component, just before it is assembled into its final subroutine form. The sub is called with a single parameter, a scalar reference to the Perl portion of the component. The sub is expected to process the string in-place. See also and .
Sub reference that is called to postprocess the text portion of a compiled component, just before it is assembled into its final subroutine form. The sub is called with a single parameter, a scalar reference to the text portion of the component. The sub is expected to process the string in-place. See also and .
Text given for this parameter is placed at the beginning of each component, but after the execution of any <%once> block. See also . The request will be available as $m in preamble code.
A list of component paths, optionally with glob wildcards, to load when the interpreter initializes. e.g.
preloads => ['/foo/index.html','/bar/*.pl']
Default is the empty list. For maximum performance, this should only be used for components that are frequently viewed and rarely updated. See the section of the administrator's manual for further details.
As mentioned in the developer's manual, a component's <%once> section is executed when it is loaded. For preloaded components, this means that this section will be executed before a Mason or Apache request exist, so preloading a component that uses $m or $r in a <%once> section will fail.
Sub reference that is called to preprocess each component before the compiler does it's magic. The sub is called with a single parameter, a scalar reference to the script. The sub is expected to process the script in-place. This is one way to extend the HTML::Mason syntax with new tags, etc., although a much more flexible way is to subclass the Lexer or Compiler class. See also and .
The class to use when creating requests. Defaults to .
The class to use when creating a resolver. Defaults to .
True or false, default is false. When false, Mason checks the timestamp of the component source file each time the component is used to see if it has changed. This provides the instant feedback for source changes that is expected for development. However it does entail a file stat for each component executed.
When true, Mason assumes that the component source tree is unchanging: it will not check component source files to determine if the memory cache or object file has expired. This can save many file stats per request. However, in order to get Mason to recognize a component source change, you must flush the memory cache and remove object files. See for one easy way to arrange this.
We recommend turning this mode on in your production sites if possible, if performance is of any concern.
Specifies a filename that Mason will check once at the beginning of of every request. When the file timestamp changes, Mason will (1) clear its in-memory component cache, and (2) remove object files if they have not already been deleted by another process.
This provides a convenient way to implement mode. All you need to do is make sure that a single file gets touched whenever components change. For Mason's part, checking a single file at the beginning of a request is much cheaper than checking every component file when static_source=0.
The class into which subcomponent objects are blessed. This defaults to .
True or false, default is true. Specifies whether Mason creates object files to save the results of component parsing. You may want to turn off object files for disk space reasons, but otherwise this should be left alone.
True or false, default is true. Indicates whether component line numbers that appear in error messages, stack traces, etc. are in terms of the source file instead of the object file. Mason does this by inserting '#line' directives into compiled components. While source line numbers are more immediately helpful, object file line numbers may be more appropriate for in-depth debugging sessions.
True or false, default is true. Indicates whether or not a given component should use strict.