=encoding utf8

=head1 NAME

Log::Report - report a problem, with exceptions and translation support

=head1 INHERITANCE

 Log::Report
   is a Exporter

=head1 SYNOPSIS

 # Invocation with 'mode' to get trace and verbose messages
 use Log::Report mode => 'DEBUG';

 # Usually invoked with a domain, which groups packages for translation
 use Log::Report 'my-domain', %options;

 # Interpolation syntax via String::Print
 # First step to translations, once you need it.
 print __x"my name is {name}", name => $n;  # print, so no exception
 print __"Hello World\n";     # no interpolation, optional translation
 print __x'Hello World';      # SYNTAX ERROR!!  ' is alternative for ::

 # Functions replacing die/warn/carp, casting exceptions.
 error "oops";                # exception like die(), no translation
 -f $config or panic "Help!"; # alert/error/fault/info/...more

 # Combined exception, interpolation, and optional translation
 error __x"Help!";            # __x() creates ::Message object
 error __x('gettext msgid', param => $value, ...)
     if $condition;

 # Also non fatal "exceptions" find their way to dispatchers
 info __x"started {pid}", pid => $$;   # translatable
 debug "$i was here!";        # you probably do not want to translate debug
 panic "arrghhh";             # like Carp::Confess

 # Many destinations for an exception message (may exist in parallel)
 dispatcher PERL => 'default' # see Log::Report::Dispatcher: use die/warn
   , reasons => 'NOTICE-';    # this dispatcher is already present at start

 dispatcher SYSLOG => 'syslog'# also send to syslog
   , charset => 'iso-8859-1'  # explicit character conversions
   , locale => 'en_US';       # overrule user's locale

 dispatcher close => 'default';  # stop default die/warn dispatcher

 # Fill-in values, like Locale::TextDomain and gettext
 # See Log::Report::Message section DETAILS
 fault __x"cannot allocate {size} bytes", size => $size;
 fault "cannot allocate $size bytes";     # no translation, ok
 fault __x"cannot allocate $size bytes";  # not translatable, wrong

 # Translation depending on count
 # Leading and trailing whitespace stay magically outside translation
 # tables.  @files in scalar context.  Special parameter with _
 print __xn"found one file\n", "found {_count} files", @files;

 # Borrow from an other text-domain (see Log::Report::Message)
 print __x(+"errors in {line}", _domain => 'global', line => $line);

 # catch errors (implements hidden eval/die)
 try { error };
 if($@) {...}      # $@ isa Log::Report::Dispatcher::Try
 if(my $exception = $@->wasFatal)         # ::Exception object

 # Language translations at the output component
 # Translation management via Log::Report::Lexicon
 use POSIX::1003::Locale qw/setlocale LC_ALL/;
 setlocale(LC_ALL, 'nl_NL');
 info __"Hello World!";      # in Dutch, if translation table found

 # Exception classes, see Log::Report::Exception
 try { error __x"something", _class => 'parsing,schema' };
 if($@->wasFatal->inClass('parsing')) ...

=head1 DESCRIPTION

Get messages to users and logs.  C<Log::Report> combines three tasks
which are closely related in one:

=over 4

=item . logging (like L<Log::Log4Perl> and syslog), and

=item . exceptions (like error and info), with

=item . translations (like C<gettext> and L<Locale::TextDomain>)

=back

You B<do not need> to use this module for all three reasons: pick what
you need now, maybe extend the usage later.  Read more about how and
why in the L</DETAILS> section, below.  Especially, you should B<read
about the REASON parameter>.

Also, you can study this module swiftly via the article published in
the German Perl C<$foo-magazine>.  English version:
F<http://perl.overmeer.net/log-report/papers/201306-PerlMagazine-article-en.html>

=head1 FUNCTIONS

=head2 Report Production and Configuration

=over 4

=item B<dispatcher>( <$type, $name, %options>|<$command, @names> )

The C<dispatcher> function controls access to dispatchers: the back-ends
which process messages, do the logging.  Dispatchers are global entities,
addressed by a symbolic $name.  Please read L<Log::Report::Dispatcher|Log::Report::Dispatcher> as
well.

The C<Log::Report> suite has its own dispatcher @types, but also connects
to external dispatching frameworks.  Each need some (minor) conversions,
especially with respect to translation of REASONS of the reports
into log-levels as the back-end understands.

[1.10] When you open a dispatcher with a $name which is already in use,
that existing dispatcher gets closed.  Except when you have given an
'dispatcher "do-not-reopen"' earlier, in which case the first object
stays alive, and the second attempt ignored. [1.11] The automatically
created default dispatcher will get replaced, even when this option
is given, by another dispatcher which is named 'default'.

The %options are a mixture of parameters needed for the
Log::Report dispatcher wrapper and the settings of the back-end.
See L<Log::Report::Dispatcher|Log::Report::Dispatcher>, the documentation for the back-end
specific wrappers, and the back-ends for more details.

Implemented COMMANDs are C<close>, C<find>, C<list>, C<disable>,
C<enable>, C<mode>, C<filter>, C<needs>, C<active-try>, and C<do-not-reopen>.

Most commands are followed by a LIST of dispatcher @names to be addressed.
For C<mode> see section L</Run modes>; it requires a MODE argument
before the LIST of NAMEs.  Non-existing names will be ignored. When
C<ALL> is specified, then all existing dispatchers will get addressed.
For C<filter> see L<Log::Report::Dispatcher/Filters>; it requires a CODE
reference before the @names of the dispatchers which will have the it
applied (defaults to all).

With C<needs>, you only provide a REASON: it will return the list of
dispatchers which need to be called in case of a message with the REASON
is triggered.  The C<active-try> [1.09] returns the closest surrounding
exception catcher, a L<Log::Report::Dispatcher::Try|Log::Report::Dispatcher::Try> object.

For both the creation as COMMANDs version of this method, all objects
involved are returned as LIST, non-existing ones skipped.  In SCALAR
context with only one name, the one object is returned.

example: play with dispatchers

 dispatcher Log::Dispatcher::File => mylog =>
   , accept   => 'MISTAKE-'              # for wrapper
   , locale   => 'pt_BR'                 # other language
   , filename => 'logfile';              # for back-end

 dispatcher close => 'mylog';            # cleanup
 my $obj = dispatcher find => 'mylog'; 
 my @obj = dispatcher 'list';
 dispatcher disable => 'syslog';
 dispatcher enable => 'mylog', 'syslog'; # more at a time
 dispatcher mode => 'DEBUG', 'mylog';
 dispatcher mode => 'DEBUG', 'ALL';
 my $catcher = dispatcher 'active-try';
 dispatcher 'do-not-reopen';

 my @need_info = dispatcher needs => 'INFO';
 if(dispatcher needs => 'INFO') ...      # anyone needs INFO

 # Getopt::Long integration: see Log::Report::Dispatcher::mode()
 dispatcher PERL => 'default', mode => 'DEBUG', accept => 'ALL'
     if $debug;

=item B<report>( [%options], $reason, $message|<STRING,$params>, )

The C<report> function is sending (for some $reason) a $message to be
displayed or logged (by a `dispatcher').  This function is the core
for L<error()|Log::Report/"Abbreviations for report()">, L<info()|Log::Report/"Abbreviations for report()"> etc functions, which are nicer names for this
exception throwing: better use those short names.

The $reason is a string like 'ERROR' (for function C<error()>).
The $message is a L<Log::Report::Message|Log::Report::Message> object (which are created with
the special translation syntax like L<__x()|Log::Report/"Messages (optionally translatable)">).  The $message may also
be a plain string, or an L<Log::Report::Exception|Log::Report::Exception> object. The optional
first parameter is a HASH which can be used to influence the dispatchers.

The optional %options are listed below.  Quite differently from other
functions and methods, they have to be passed in a HASH as first parameter.

This function returns the LIST of dispatchers which accepted the $message.
When empty, no back-end has accepted it so the $message was "lost".
Even when no back-end needs the message, the program will still exit
when there is a $reason to C<die()>.

 -Option  --Default
  errno     $! or 1
  is_fatal  <depends on reason>
  locale    undef
  location  undef
  stack     undef
  to        undef

=over 2

=item errno => INTEGER

When the $reason includes the error text (See L</Run modes>), you can
overrule the error code kept in C<$!>.  In other cases, the return code
defaults to C<1> (historical UNIX behavior). When the message $reason
(combined with the run-mode) is severe enough to stop the program,
this value as return code of the program.  The use of this option itself
will not trigger an C<die()>.

=item is_fatal => BOOLEAN

Some logged exceptions are fatal, other aren't.  The default usually
is correct. However, you may want an error to be caught (usually with
L<try()|Log::Report/"Report Production and Configuration">), redispatch it to syslog, but without it killing the main
program.

=item locale => LOCALE

Use this specific locale, in stead of the user's preference.

=item location => STRING

When defined, this location is used in the display.  Otherwise, it
is determined automatically if needed.  An empty string will disable
any attempt to display this line.

=item stack => ARRAY

When defined, that data is used to display the call stack.  Otherwise,
it is collected via C<caller()> if needed.

=item to => NAME|ARRAY-of-NAMEs

Sent the $message only to the NAMEd dispatchers.  Ignore unknown NAMEs.
Still, the dispatcher needs to be enabled and accept the REASONs.

=back

example: for use of L<report()|Log::Report/"Report Production and Configuration">

 # long syntax example
 report TRACE => "start processing now";
 report INFO  => '500: ' . __'Internal Server Error';

 # explicit dispatcher, no translation
 report {to => 'syslog'}, NOTICE => "started process $$";
 notice "started process $$", _to => 'syslog';   # same

 # short syntax examples
 trace "start processing now";
 warning  __x'Disk {percent%.2f}% full', percent => $p
     if $p > 97;

 # error message, overruled to be printed in Brazilian
 report {locale => 'pt_BR'}
    , WARNING => "do this at home!";

=item B<try>(CODE, %options)

Execute the CODE while blocking all dispatchers as long as it is running.
The exceptions which occur while running the CODE are caught until it
has finished.  When there where no fatal errors, the result of the CODE
execution is returned.

After the CODE was tried, the C<$@> will contain a
L<Log::Report::Dispatcher::Try|Log::Report::Dispatcher::Try> object, which contains the collected
messages.

Run-time errors from Perl and die's, croak's and confess's within the
program (which shouldn't appear, but you never know) are collected into an
L<Log::Report::Message|Log::Report::Message> object, using L<Log::Report::Die|Log::Report::Die>.

The %options are passed to the constructor of the try-dispatcher, see
L<Log::Report::Dispatcher::Try::new()|Log::Report::Dispatcher::Try/"Constructors">.  For instance, you may like to
add C<< mode => 'DEBUG' >>, or C<< accept => 'ERROR-' >>.

B<Be warned> that the parameter to C<try> is a CODE reference.  This means
that you shall not use a comma after the block when there are %options
specified.  On the other hand, you shall use a semi-colon after the
block if there are no arguments.

B<Be warned> that the {} are interpreted as subroutine, which means that,
for instance, it has its own C<@_>.  The manual-page of Try::Tiny
lists a few more side-effects of this.

example: 

 my $x = try { 3/$x };  # mind the ';' !!
 if($@) {               # signals something went wrong

 if(try {...}) {        # block ended normally, returns bool

 try { ... }            # no comma!!
    mode => 'DEBUG', accept => 'ERROR-';

 try sub { ... },       # with comma, also \&function
    mode => 'DEBUG', accept => 'ALL';

 my $response = try { $ua->request($request) };
 if(my $e = $@->wasFatal) ...

=back

=head2 Abbreviations for report()

The following functions are all wrappers for calls to L<report()|Log::Report/"Report Production and Configuration">,
and available when "syntax is SHORT" (by default, see L<import()|Log::Report/"Configuration">).
You cannot specify additional options to influence the behavior of
C<report()>, which are usually not needed anyway.

=over 4

=item B<alert>($message)

Short for C<< report ALERT => $message >>

=item B<assert>($message)

Short for C<< report ASSERT => $message >>

=item B<error>($message)

Short for C<< report ERROR => $message >>

=item B<failure>($message)

Short for C<< report FAILURE => $message >>

=item B<fault>($message)

Short for C<< report FAULT => $message >>

=item B<info>($message)

Short for C<< report INFO => $message >>

=item B<mistake>($message)

Short for C<< report MISTAKE => $message >>

=item B<notice>($message)

Short for C<< report NOTICE => $message >>

=item B<panic>($message)

Short for C<< report PANIC => $message >>

=item B<trace>($message)

Short for C<< report TRACE => $message >>

=item B<warning>($message)

Short for C<< report WARNING => $message >>

=back

=head2 Messages (optionally translatable)

Even when you do not support translations (yet) you may want to use
message objects to improve the logging feature. For instance,
you get very powerful interpolation from L<String::Print|String::Print>.

The language translations are initiate by limited set of functions
which contain B<two under-scores> (C<__>) in their name.  Most
of them return a L<Log::Report::Message|Log::Report::Message> object.

B<Be warned(1)> that -in general- its considered very bad practice to
combine multiple translations into one message: translating may also
affect the order of the translated components. Besides, when the person
which translates only sees smaller parts of the text, his (or her) job
becomes more complex.  So:

 print __"Hello" . ', ' . __"World!";  # works, but to be avoided
 print __"Hello, World!";              # preferred, complete sentence

The the former case, tricks with overloading used by the
L<Log::Report::Message|Log::Report::Message> objects will still make delayed translations
work.

In normal situations, it is not a problem to translate interpolated
values:

 print __"the color is {c}", c => __"red";

B<Be warned(2)> that using C<< __'Hello' >> will produce a syntax error like
"String found where operator expected at .... Can't find string terminator
"'" anywhere before EOF".  The first quote is the cause of the complaint,
but the second generates the error.  In the early days of Perl, the single
quote was used to separate package name from function name, a role which
was later replaced by a double-colon.  So C<< __'Hello' >> gets interpreted
as C<< __::Hello ' >>.  Then, there is a trailing single quote which has
no counterpart.

=over 4

=item B<N__>($msgid)

Label to indicate that the string is a text which will be translated
later.  The function itself does nothing.  See also L<N__w()|Log::Report/"Messages (optionally translatable)">.

This no-op function is used as label to the xgettext program to build the
translation tables.

example: how to use N__()

 # add three msgids to the translation table
 my @colors = (N__"red", N__"green", N__"blue");
 my @colors = N__w "red green blue";   # same
 print __ $colors[1];                  # translate green

 # using __(), would work as well
 my @colors = (__"red", __"green", __"blue");
 print $colors[1];
 # however: this will always create all Log::Report::Message objects,
 # where maybe only one is used.

=item B<N__n>($single_msgid, $plural_msgid)

Label to indicate that the two MSGIDs are related, the first as
single, the seconds as its plural.  Only used to find the text
fragments to be translated.  The function itself does nothing.

example: how to use L<N__n()|Log::Report/"Messages (optionally translatable)">

 my @save = N__n "save file", "save files";
 my @save = (N__n "save file", "save files");
 my @save = N__n("save file", "save files");

 # be warned about SCALARs in prototype!
 print __n @save, $nr_files;  # wrong!
 print __n $save[0], $save[1], @files, %vars;

=item B<N__w>(STRING)

This extension to the Locale::TextDomain syntax, is a combined
C<qw> (list of quoted words) and L<N__()|Log::Report/"Messages (optionally translatable)"> into a list of translatable
words.

example: of L<N__w()|Log::Report/"Messages (optionally translatable)">

  my @colors = (N__"red", N__"green", N__"blue");
  my @colors = N__w"red green blue";  # same
  print __ $colors[1];

=item B<__>($msgid)

This function (name is B<two> under-score characters) will cause the $msgid
to be replaced by the translations when doing the actual output.  Returned
is a L<Log::Report::Message|Log::Report::Message> object, which will be used in translation
later.  Translating is invoked when the object gets stringified.  When
you have no translation tables, the $msgid will be shown untranslated.

If you need options for L<Log::Report::Message::new()|Log::Report::Message/"Constructors"> then use L<__x()|Log::Report/"Messages (optionally translatable)">;
the prototype of this function does not permit parameters: it is a
prefix operator!

example: how to use __()

 print __"Hello World";      # translated into user's language
 print __'Hello World';      # syntax error!
 print __('Hello World');    # ok, translated
 print __"Hello", " World";  # World not translated

 my $s = __"Hello World";    # creates object, not yet translated
 print ref $s;               # Log::Report::Message
 print $s;                   # ok, translated
 print $s->toString('fr');   # ok, forced into French

=item B<__n>($msgid, $plural_msgid, $count, PAIRS)

It depends on the value of $count (and the selected language) which
text will be displayed.  When translations can not be performed, then
$msgid will be used when $count is 1, and PLURAL_MSGSID in other cases.
However, some languages have more complex schemes than English.

The PAIRS are options for L<Log::Report::Message::new()|Log::Report::Message/"Constructors"> and variables
to be filled in.

example: how to use __n()

 print __n "one", "more", $a;
 print __n("one", "more", $a), "\n";
 print +(__n "one", "more", $a), "\n";

 # new-lines are ignore at lookup, but printed.
 print __n "one\n", "more\n", $a;

 # count is in scalar context
 # the value is also available as _count
 print __n "found one\n", "found {_count}\n", @r;

 # ARRAYs and HASHes are counted
 print __n "one", "more", \@r;

=item B<__nx>($msgid, $plural_msgid, $count, PAIRS)

It depends on the value of $count (and the selected language) which
text will be displayed.  See details in L<__n()|Log::Report/"Messages (optionally translatable)">.  After translation,
the VARIABLES will be filled-in.

The PAIRS are options for L<Log::Report::Message::new()|Log::Report::Message/"Constructors"> and variables
to be filled in.

example: how to use __nx()

 print __nx "one file", "{_count} files", $nr_files;
 print __nx "one file", "{_count} files", @files;

 local $" = ', ';
 print __nx "one file: {f}", "{_count} files: {f}", @files, f => \@files;

=item B<__x>($msgid, PAIRS)

Translate the $msgid and then interpolate the VARIABLES in that string.
Of course, translation and interpolation is delayed as long as possible.
Both OPTIONS and VARIABLES are key-value pairs.

The PAIRS are options for L<Log::Report::Message::new()|Log::Report::Message/"Constructors"> and variables
to be filled in.

=item B<__xn>($single_msgid, $plural_msgid, $count, $paurs)

Same as L<__nx()|Log::Report/"Messages (optionally translatable)">, because we have no preferred order for 'x' and 'n'.

=back

=head3 Messages with msgctxt

In Log::Report, the message context (mgsctxt in the PO-files --in the
translation tables) can be used in a very powerful way.  Read all about
it in L<Log::Report::Translator::Context|Log::Report::Translator::Context>

The msgctxt versions of the tranditional gettext infrastructure are far
less useful for Log::Report, because we can easily work with different
text domains within the same program.  That should avoid most of the
accidental translation conflicts between components of the code.

Just for compatibility with Locale::TextDomain and completeness, the
'p' versions of above methods are supported.  See examples for these
functions in Locale::TextDomain.

B<Warnings:> Functions C<N__p()> and C<N__np()> seem not to be usable in
reality, hence not implemented.  The script xgettext-perl and
L<Log::Report::Extract::PerlPPI|Log::Report::Extract::PerlPPI> (both in the L<Log::Report::Lexicon|Log::Report::Lexicon>
distribution) do not yet support these functions.

=over 4

=item B<__np>($msgctxt, $msgid, $plural, count)

=item B<__npx>($msgctxt, $msgid, $plural, count, PAIRS)

=item B<__p>($msgctxt, $msgid)

=item B<__px>($msgctxt, $msgid, PAIRS)

=back

=head2 Configuration

=over 4

=item $obj-E<gt>B<import>( [$level,][$domain,] %options )

The import is automatically called when the package is compiled.  For all
packages but one in your distribution, it will only contain the name of
the $domain.

For one package, the import list may additionally contain textdomain
configuration %options.  These %options are used for all packages which
use the same $domain.  These are alternatives:

  # Do not use variables in the %*config!  They are not yet initialized
  # when Log::Report->import is run!!!
  use Log::Report 'my-domain', %config, %domain_config;

  use Log::Report 'my-domain', %config;
  textdomain 'my-domain', %domain_config;   # vars allowed

The latter syntax has major advantages, when the configuration of the
domain is determined at run-time.  It is probably also easier to understand.

See L<Log::Report::Domain::configure()|Log::Report::Domain/"Attributes">, for the B<list of %options>
for the domain configuration.  Here, we only list the options which are
related to the normal import behavior.

The export $level is a plus (+) followed by a number, for instance C<+1>,
to indicate to on which caller level we need to work.  This is used
in L<Log::Report::Optional|Log::Report::Optional>.  It defaults to '0': my direct caller.

 -Option       --Default
  import         undef
  message_class  Log::Report::Message
  mode           'NORMAL'
  syntax         'SHORT'

=over 2

=item import => FUNCTION|ARRAY

[0.998] When not specified, the C<syntax> option determines the list
of functions which are being exported.  With this option, the C<syntax>
option is ignored and only the specified FUNCTION(s) are imported.

=item message_class => CLASS

[1.08] Use a more powerful message object class, for instance because
your messages need extra attributes.  The provided CLASS must extend
L<Log::Report::Message|Log::Report::Message>

=item mode => LEVEL

This sets the default mode for all created dispatchers.  You can
also selectively change the output mode, like
 dispatcher PERL => 'default', mode => 3

=item syntax => 'REPORT'|'SHORT'|'LONG'

The SHORT syntax will add the report abbreviations (like function
L<error()|Log::Report/"Abbreviations for report()">) to your name-space.  Otherwise, each message must be produced
with L<report()|Log::Report/"Report Production and Configuration">. C<LONG> is an alternative to C<REPORT>: both do not
pollute your namespace with the useful abbrev functions.

=back

example: of import

 use Log::Report mode => 3;     # '3' or 'DEBUG'

 use Log::Report 'my-domain';   # in each package producing messages

 use Log::Report 'my-domain'    # in one package, top of distr
  , mode            => 'VERBOSE'
  , syntax          => 'REPORT' # report ERROR, not error()
  , translator      => Log::Report::Translator::POT->new
     ( lexicon => '/home/mine/locale'  # translation tables
     )
  , native_language => 'nl_NL'; # untranslated msgs are Dutch

 use Log::Report import => 'try';      # or ARRAY of functions

=item B<textdomain>( <[$name],$config>|<$name, 'DELETE'|'EXISTS'>|$domain )

[1.00] Without CONFIGuration, this returns the L<Log::Report::Domain|Log::Report::Domain> object
which administers the $domain, by default the domain effective in the scope
of the package.

A very special case is "DELETE", which will remove the domain
configuration. [1.20] "EXISTS" will check for existence: when it exists,
it will be returned, but a domain will not be automagically created.

[1.20] You may also pass a pre-configured domain.

=back

=head2 Reasons

=over 4

=item Log::Report-E<gt>B<needs>( $reason, [$reasons] )

Returns true when the reporter needs any of the $reasons, when any of
the active dispatchers is collecting messages in the specified level.
This is useful when the processing of data for the message is relatively
expensive, but for instance only required in debug mode.

example: 

  if(Log::Report->needs('TRACE'))
  {   my @args = ...expensive calculation...;
      trace "your options are: @args";
  }

=back

=head1 DETAILS

=head2 Introduction

Getting messages to users and logs. The distincting concept of this module,
is that three tasks which are strongly related are merged into one simple
syntax.  The three tasks:

=over 4

=item produce some text on a certain condition,

=item translate it to the proper language, and

=item deliver it in some way to a user.

=back

Text messages in Perl are produced by commands like C<print>, C<die>,
C<warn>, C<carp>, or C<croak>.  But where is that output directed to?
Translations is hard.  There is no clean exception mechanism.

Besides, the C<print>/C<warn>/C<die> together produce only three different
output "levels" with a message.  Think of the variation syslog offers:
more than 7 levels.  Many people manually implement their own tricks to
get additional levels, like verbose and debug flags.  Log::Report offers
that variety.

The (optional) translations use the beautiful syntax defined by
Locale::TextDomain, with some own extensions (of course).  A very
important difference is that translations are delayed till the delivery
step: until a dispatcher actually writes your message into a file, sends
it to syslog, or shows it on the screen.  This means that the pop-up in
the graphical interface of the user may show the text in the language
of the user --say Chinese in utf8--, but at the same time syslog may
write the latin1 English version of the same message.

=head2 Background ideas

The following ideas are the base of this implementation:

=over 4

=item . simplification

Handling errors and warnings is probably the most labor-intensive
task for a programmer: when programs are written correctly, up-to
three-quarters of the code is related to testing, reporting, and
handling (problem) conditions.  Simplifying the way to create reports,
simplifies programming and maintenance.

=item . multiple dispatchers

It is not the location where the (for instance) error occurs which
determines what will happen with the text, but the main application which
uses the the complaining module has control.  Messages have a reason.
Based on the `reason' classification, they can get ignored, send to one
or multiple dispatchers, like Log::Dispatch, Log::Log4perl,
or UNIX syslog.

=item . delayed translations

The background ideas are that of Locale::TextDomain, based
on C<gettext()>.  However, in the C<Log::Report> infrastructure,
translations are postponed until the text is dispatched to a screen
or log-file; the same report can be sent to syslog in (for instance)
English and to the user interface in Dutch.

=item . context sensitive

Using contexts, you can set-up how to translate or rewrite messages,
to improve messages.  A typical problem is whether to use gender in
text (use 'his' or 'her'): you can set a gender in a context, and the
use translation tables to pick the right one.

=back

=head2 Error handling models

There are two approaches to handling errors and warnings.  In the first
approach, as produced by C<die>, C<warn> and the C<carp> family of
commands, the program handles the problem immediately on the location
where the problem appears.  In the second approach, an I<exception>
is thrown on the spot where the problem is created, and then somewhere
else in the program the condition is handled.

The implementation of exceptions in Perl5 is done with a eval-die pair:
on the spot where the problem occurs, C<die> is called.  But, because of
the execution of that routine is placed within an C<eval>, the program
as a whole will not die, just the execution of a part of the program
will seize.  However, what if the condition which caused the routine to die
is solvable on a higher level?  Or what if the user of the code doesn't
bother that a part fails, because it has implemented alternatives for
that situation?  Exception handling is quite clumsy in Perl5.

The C<Log::Report> set of distributions let modules concentrate on the
program flow, and let the main program decide on the report handling
model.  The infrastructure to translate messages into multiple languages,
whether to create exceptions or carp/die, to collect longer explanations
with the messages, to log to mail or syslog, and so on, is decided in
pluggable back-ends.

=head3 The Reason for the report

Traditionally, perl has a very simple view on error reports: you
either have a warning or an error.  However, it would be much clearer
for user's and module-using applications, when a distinction is made
between various causes.  For instance, a configuration error is quite
different from a disk-full situation.  In C<Log::Report>, the produced
reports in the code tell I<what> is wrong.  The main application defines
loggers, which interpret the cause into (syslog) levels.

Defined by C<Log::Report> are

=over 4

=item . trace (debug, program)

The message will be used when some logger has debugging enabled.  The
messages show steps taken by the program, which are of interest by the
developers and maintainers of the code, but not for end-users.

=item . assert (program)

Shows an unexpected condition, but continues to run.  When you want the
program to abort in such situation, that use C<panic>.

=item . info (verbose, program)

These messages show larger steps in the execution of the program.
Experienced users of the program usually do not want to see all these
intermediate steps.  Most programs will display info messages (and
higher) when some C<verbose> flag is given on the command-line.

=item . notice (program)

An user may need to be aware of the program's accidental smart behavior,
for instance, that it initializes a lasting C<Desktop> directory in your
home directory.  Notices should be sparse.

=item . warning (program)

The program encountered some problems, but was able to work around it
by smart behavior.  For instance, the program does not understand a
line from a log-file, but simply skips the line.

=item . mistake (user)

When a user does something wrong, but what is correctable by smart
behavior of the program.  For instance, in some configuration file,
you can fill-in "yes" or "no", but the user wrote "yeah".  The program
interprets this as "yes", producing a mistake message as warning.

It is much nicer to tell someone that he/she made a mistake, than
to call that an error.

=item . error (user)

The user did something wrong, which is not automatically correctable
or the program is not willing to correct it automatically for reasons
of code quality.  For instance, an unknown option flag is given on the
command-line.  These are configuration issues, and have no useful
value in C<$!>.  The program will be stopped, usually before taken off.

=item . fault (system)

The program encountered a situation where it has no work-around.  For
instance, a file cannot be opened to be written.  The cause of that
problem can be some user error (i.e. wrong filename), or external
(you accidentally removed a directory yesterday).  In any case, the
C<$!> (C<$ERRNO>) variable is set here.

=item . alert (system)

Some external cause disturbs the execution of the program, but the
program stays alive and will try to continue operation.  For instance,
the connection to the database is lost.  After a few attempts, the
database can be reached and the program continues as if nothing happened.
The cause is external, so C<$!> is set.  Usually, a system administrator
needs to be informed about the problem.

=item . failure (system)

Some external cause makes it impossible for this program to continue.
C<$!> is set, and usually the system administrator wants to be
informed.  The program will die.

The difference with C<fault> is subtile and not always clear.  A fault
reports an error returned by an operating system call, where the failure
would report an operational problem, like a failing mount.

=item . panic (program)

All above report classes are expected: some predictable situation
is encountered, and therefore a message is produced.  However, programs
often do some internal checking.  Of course, these conditions should
never be triggered, but if they do... then we can only stop.

For instance, in an OO perl module, the base class requires all
sub-classes to implement a certain method.  The base class will produce
a stub method with triggers a panic when called.  The non-dieing version
of this test C<assert>.

=back

I<Debugging> or being C<verbose> are run-time behaviors, and have nothing
directly to do with the type of message which is produced.  These two
are B<modes> which can be set on the dispatchers: one dispatcher may
be more verbose that some other.

On purpose, we do not use the terms C<die> or C<fatal>, because the
dispatcher can be configured what to do in cause of which condition.
For instance, it may decide to stop execution on warnings as well.

The terms C<carp> and C<croak> are avoided, because the program cause
versus user cause distinction (warn vs carp) is reflected in the use
of different reasons.  There is no need for C<confess> and C<croak>
either, because the dispatcher can be configured to produce stack-trace
information (for a limited sub-set of dispatchers)

=head3 Report levels

Various frameworks used with perl programs define different labels
to indicate the reason for the message to be produced.

 Perl5 Log::Dispatch Syslog Log4Perl Log::Report
 print   0,debug     debug  debug    trace
 print   0,debug     debug  debug    assert
 print   1,info      info   info     info
 warn\n  2,notice    notice info     notice
 warn    3,warning   warn   warn     mistake
 carp    3,warning   warn   warn     warning
 die\n   4,error     err    error    error
 die     5,critical  crit   fatal    fault
 croak   6,alert     alert  fatal    alert  
 croak   7,emergency emerg  fatal    failure
 confess 7,emergency emerg  fatal    panic

=head3 Run modes

The run-mode change which messages are passed to a dispatcher, but
from a different angle than the dispatch filters; the mode changes
behavioral aspects of the messages, which are described in detail in
L<Log::Report::Dispatcher/Processing the message>.  However, it should
behave as you expect: the DEBUG mode shows more than the VERBOSE mode,
and both show more than the NORMAL mode.

B<. Example: extract run mode from Getopt::Long>

The C<GetOptions()> function will count the number of C<v> options
on the command-line when a C<+> is after the option name.

 use Log::Report;
 use Getopt::Long qw(:config no_ignore_case bundling);

 my $mode;    # defaults to NORMAL
 GetOptions 'v+'        => \$mode
          , 'verbose=i' => \$mode
          , 'mode=s'    => \$mode
     or exit 1;

 dispatcher 'PERL', 'default', mode => $mode;

Now, C<-vv> will set C<$mode> to C<2>, as will C<--verbose 2> and
C<--verbose=2> and C<--mode=ASSERT>.  Of course, you do not need to
provide all these options to the user: make a choice.

B<. Example: the mode of a dispatcher>

 my $mode = dispatcher(find => 'myname')->mode;

B<. Example: run-time change mode of a dispatcher>

To change the running mode of the dispatcher, you can do
  dispatcher mode => DEBUG => 'myname';

However, be warned that this does not change the types of messages
accepted by the dispatcher!  So: probably you will not receive
the trace, assert, and info messages after all.  So, probably you
need to replace the dispatcher with a new one with the same name:
  dispatcher FILE => 'myname', to => ..., mode => 'DEBUG';

This may reopen connections (depends on the actual dispatcher), which
might be not what you wish to happened.  In that case, you must take
the following approach:

  # at the start of your program
  dispatcher FILE => 'myname', to => ...
     , accept => 'ALL';    # overrule the default 'NOTICE-' !!

  # now it works
  dispatcher mode => DEBUG => 'myname';    # debugging on
  ...
  dispatcher mode => NORMAL => 'myname';   # debugging off

Of course, this comes with a small overall performance penalty.

=head3 Exceptions

The simple view on live says: you 're dead when you die.  However,
more complex situations try to revive the dead.  Typically, the "die"
is considered a terminating exception, but not terminating the whole
program, but only some logical block.  Of course, a wrapper round
that block must decide what to do with these emerging problems.

Java-like languages do not "die" but throw exceptions which contain the
information about what went wrong.  Perl modules like C<Exception::Class>
simulate this.  It's a hassle to create exception class objects for each
emerging problem, and the same amount of work to walk through all the
options.

Log::Report follows a simpler scheme.  Fatal messages will "die", which is
caught with "eval", just the Perl way (used invisible to you).  However,
the wrapper gets its hands on the message as the user has specified it:
untranslated, with all unprocessed parameters still at hand.

 try { fault __x "cannot open file {file}", file => $fn };
 if($@)                         # is Log::Report::Dispatcher::Try
 {   my $cause = $@->wasFatal;  # is Log::Report::Exception
     $cause->throw if $cause->message->msgid =~ m/ open /;
     # all other problems ignored
 }

See L<Log::Report::Dispatcher::Try|Log::Report::Dispatcher::Try> and L<Log::Report::Exception|Log::Report::Exception>.

=head2 Comparison

Some notes on differences between the Log::Report approach and other
Perl concepts.

=head3 die/warn/Carp

Perl's built-in exception system is very primitive: "die" and "warn".
Most programming languages provide a much more detailed exception
mechanism.

A typical perl program can look like this:

 my $dir = '/etc';

 File::Spec->file_name is_absolute($dir)
     or die "ERROR: directory name must be absolute.\n";

 -d $dir
     or die "ERROR: what platform are you on?";

 until(opendir DIR, $dir)
 {   warn "ERROR: cannot read system directory $dir: $!";
     sleep 60;
 }

 print "Processing directory $dir\n"
     if $verbose;

 while(defined(my $file = readdir DIR))
 {   if($file =~ m/\.bak$/)
     {   warn "WARNING: found backup file $dir/$f\n";
         next;
     }

     die "ERROR: file $dir/$file is binary"
         if $debug && -B "$dir/$file";

     print "DEBUG: processing file $dir/$file\n"
         if $debug;

     open FILE, "<", "$dir/$file"
         or die "ERROR: cannot read from $dir/$f: $!";

     close FILE
         or croak "ERROR: read errors in $dir/$file: $!";
 }

Where C<die>, C<warn>, and C<print> are used for various tasks.  With
C<Log::Report>, you would write

 use Log::Report;

 # can be left-out when there is no debug/verbose
 dispatcher PERL => 'default', mode => 'DEBUG';

 my $dir = '/etc';

 File::Spec->file_name is_absolute($dir)
     or mistake "directory name must be absolute";

 -d $dir
     or panic "what platform are you on?";

 until(opendir DIR, $dir)
 {   alert "cannot read system directory $dir";
     sleep 60;
 }

 info "Processing directory $dir";

 while(defined(my $file = readdir DIR))
 {   if($file =~ m/\.bak$/)
     {   notice "found backup file $dir/$f";
         next;
     }

     assert "file $dir/$file is binary"
         if -B "$dir/$file";

     trace "processing file $dir/$file";

     unless(open FILE, "<", "$dir/$file")
     {   error "no permission to read from $dir/$f"
             if $!==ENOPERM;
         fault "unable to read from $dir/$f";
     }

     close FILE
         or failure "read errors in $dir/$file";
 }

A lot of things are quite visibly different, and there are a few smaller
changes.  There is no need for a new-line after the text of the message.
When applicable (error about system problem), then the C<$!> is added
automatically.

=head3 Log::Dispatch and Log::Log4perl

The two major logging frameworks for Perl are Log::Dispatch and
Log::Log4perl; both provide a pluggable logging interface.

Both frameworks do not have (gettext or maketext) language translation
support, which has various consequences.  When you wish for to report
in some other language, it must be translated before the logging
function is called.   This may mean that an error message is produced
in Chinese, and therefore also ends-up in the syslog file in Chinese.
When this is not your language, you have a problem.

Log::Report translates only in the back-end, which means that the user may
get the message in Chinese, but you get your report in your beloved Dutch.
When no dispatcher needs to report the message, then no time is lost in
translating.

With both logging frameworks, you use terminology comparable to
syslog: the module programmer determines the seriousness of the
error message, not the application which integrates multiple modules.
This is the way perl programs usually work, but often the cause for
inconsequent user interaction.

=head3 Locale::gettext and Locate::TextDomain

Both on GNU gettext based implementations can be used as translation
frameworks.  Locale::TextDomain syntax is supported, with quite some
extensions. Read the excellent documentation of Locale::Textdomain.
Only the tried access via C<$__> and C<%__> are not supported.

The main difference with these modules is the moment when the translation
takes place.  In Locale::TextDomain, an C<__x()> will result in an
immediate translation request via C<gettext()>.  C<Log::Report>'s version
of C<__x()> will only capture what needs to be translated in an object.
When the object is used in a print statement, only then the translation
will take place.  This is needed to offer ways to send different
translations of the message to different destinations.

To be able to postpone translation, objects are returned which stringify
into the translated text.

=head1 DIAGNOSTICS

=over 4

=item Error: in SCALAR context, only one dispatcher name accepted

The L<dispatcher()|Log::Report/"Report Production and Configuration"> method returns the L<Log::Report::Dispatcher|Log::Report::Dispatcher>
objects which it has accessed.  When multiple names where given, it
wishes to return a LIST of objects, not the count of them.

=back

=head1 SEE ALSO

This module is part of Log-Report distribution version 1.32,
built on January 26, 2021. Website: F<http://perl.overmeer.net/CPAN/>

=head1 LICENSE

Copyrights 2007-2021 by [Mark Overmeer <markov@cpan.org>]. For other contributors see ChangeLog.

This program is free software; you can redistribute it and/or modify it
under the same terms as Perl itself.
See F<http://dev.perl.org/licenses/>