File Coverage

blib/lib/Log/Dispatchouli.pm

Criterion	Covered	Total	%
statement	265	339	78.1
branch	59	106	55.6
condition	15	25	60.0
subroutine	50	67	74.6
pod	34	40	85.0
total	423	577	73.3

use v5.20;
use warnings;
package Log::Dispatchouli 3.101;
# ABSTRACT: a simple wrapper around Log::Dispatch

# Not dangerous.  Accepted without change.
use experimental 'postderef', 'signatures';

use Carp ();
use File::Spec ();
use Log::Dispatch;
use Log::Fmt ();
use Params::Util qw(_ARRAY0 _HASH0 _CODELIKE);
use Scalar::Util qw(blessed weaken);
use String::Flogger;
use Try::Tiny 0.04;

require Log::Dispatchouli::Proxy;

our @CARP_NOT = qw(Log::Dispatchouli::Proxy);

#pod =head1 SYNOPSIS
#pod
#pod   my $logger = Log::Dispatchouli->new({
#pod     ident     => 'stuff-purger',
#pod     facility  => 'daemon',
#pod     to_stdout => $opt->{print},
#pod     debug     => $opt->{verbose}
#pod   });
#pod
#pod   $logger->log([ "There are %s items left to purge...", $stuff_left ]);
#pod
#pod   $logger->log_debug("this is extra often-ignored debugging log");
#pod
#pod   $logger->log_fatal("Now we will die!!");
#pod
#pod =head1 DESCRIPTION
#pod
#pod Log::Dispatchouli is a thin layer above L<Log::Dispatch> and meant to make it
#pod dead simple to add logging to a program without having to think much about
#pod categories, facilities, levels, or things like that.  It is meant to make
#pod logging just configurable enough that you can find the logs you want and just
#pod easy enough that you will actually log things.
#pod
#pod Log::Dispatchouli can log to syslog (if you specify a facility), standard error
#pod or standard output, to a file, or to an array in memory.  That last one is
#pod mostly useful for testing.
#pod
#pod In addition to providing as simple a way to get a handle for logging
#pod operations, Log::Dispatchouli uses L<String::Flogger> to process the things to
#pod be logged, meaning you can easily log data structures.  Basically: strings are
#pod logged as is, arrayrefs are taken as (sprintf format, args), and subroutines
#pod are called only if needed.  For more information read the L<String::Flogger>
#pod docs.
#pod
#pod =head1 LOGGER PREFIX
#pod
#pod Log messages may be prepended with information to set context.  This can be set
#pod at a logger level or per log item.  The simplest example is:
#pod
#pod   my $logger = Log::Dispatchouli->new( ... );
#pod
#pod   $logger->set_prefix("Batch 123: ");
#pod
#pod   $logger->log("begun processing");
#pod
#pod   # ...
#pod
#pod   $logger->log("finished processing");
#pod
#pod The above will log something like:
#pod
#pod   Batch 123: begun processing
#pod   Batch 123: finished processing
#pod
#pod To pass a prefix per-message:
#pod
#pod   $logger->log({ prefix => 'Sub-Item 234: ' }, 'error!')
#pod
#pod   # Logs: Batch 123: Sub-Item 234: error!
#pod
#pod If the prefix is a string, it is prepended to each line of the message.  If it
#pod is a coderef, it is called and passed the message to be logged.  The return
#pod value is logged instead.
#pod
#pod L<Proxy loggers|/METHODS FOR PROXY LOGGERS> also have their own prefix
#pod settings, which accumulate.  So:
#pod
#pod   my $proxy = $logger->proxy({ proxy_prefix => 'Subsystem 12: ' });
#pod
#pod   $proxy->set_prefix('Page 9: ');
#pod
#pod   $proxy->log({ prefix => 'Paragraph 6: ' }, 'Done.');
#pod
#pod ...will log...
#pod
#pod   Batch 123: Subsystem 12: Page 9: Paragraph 6: Done.
#pod
#pod =method new
#pod
#pod   my $logger = Log::Dispatchouli->new(\%arg);
#pod
#pod This returns a new logger, a Log::Dispatchouli object.
#pod
#pod Valid arguments are:
#pod
#pod   ident       - the name of the thing logging (mandatory)
#pod   to_self     - log to the logger object for testing; default: false
#pod   to_stdout   - log to STDOUT; default: false
#pod   to_stderr   - log to STDERR; default: false
#pod   facility    - to which syslog facility to send logs; default: none
#pod
#pod   to_file     - DEPRECATED: this option will be removed in 2025
#pod                 log to PROGRAM_NAME.YYYYMMDD in the log path; default: false
#pod   log_file    - a leaf name for the file to log to with to_file
#pod   log_path    - path in which to log to file; defaults to DISPATCHOULI_PATH
#pod                 environment variable or, failing that, to your system's tmpdir
#pod
#pod   file_format - this optional coderef is passed the message to be logged
#pod                 and returns the text to write out
#pod
#pod   log_pid     - if 1, prefix all log entries with the pid; default: true
#pod                 can also be a comma-delimited list of log targets where pid is
#pod                 logged, like "stderr,syslog"; mostly useful for logging pid in
#pod                 syslog, but not on standard I/O
#pod   fail_fatal  - a boolean; if true, failure to log is fatal; default: true
#pod   muted       - a boolean; if true, only fatals are logged; default: false
#pod   debug       - a boolean; if true, log_debug method is not a no-op
#pod                 defaults to the truth of the DISPATCHOULI_DEBUG env var
#pod   quiet_fatal - 'stderr' or 'stdout' or an arrayref of zero, one, or both
#pod                 fatal log messages will not be logged to these
#pod                 (default: stderr)
#pod   config_id   - a name for this logger's config; rarely needed!
#pod   syslog_socket - a value for Sys::Syslog's "socket" arg; default: "native"
#pod
#pod The log path is either F</tmp> or the value of the F<DISPATCHOULI_PATH> env var.
#pod
#pod If the F<DISPATCHOULI_NOSYSLOG> env var is true, we don't log to syslog.
#pod
#pod =cut

sub new ($class, $arg = undef) {
  $arg ||= {};

  my $ident = $arg->{ident}
    or Carp::croak "no ident specified when using $class";

  my $config_id = defined $arg->{config_id} ? $arg->{config_id} : $ident;

  my %quiet_fatal;
  for ('quiet_fatal') {
    %quiet_fatal = map {; $_ => 1 } grep { defined }
      exists $arg->{$_}
        ? _ARRAY0($arg->{$_}) ? @{ $arg->{$_} } : $arg->{$_}
        : ('stderr');
  };

  my $log = Log::Dispatch->new;
  my $self = bless {
    dispatcher => $log,
    log_pid    => (exists $arg->{log_pid} ? $arg->{log_pid} : 1),
  } => $class;

  if ($arg->{to_file}) {
    Carp::carp("to_file argument for Log::Dispatchouli is deprecated and will be removed in late 2025");

    require Log::Dispatch::File;
    my $log_file = File::Spec->catfile(
      ($arg->{log_path} || $self->env_value('PATH') || File::Spec->tmpdir),
      $arg->{log_file} || do {
        my @time = localtime;
        sprintf('%s.%04u%02u%02u',
          $ident,
          $time[5] + 1900,
          $time[4] + 1,
          $time[3])
      }
    );

    $log->add(
      Log::Dispatch::File->new(
        name      => 'logfile',
        min_level => 'debug',
        filename  => $log_file,
        mode      => 'append',
        callbacks => do {
          my $log_pid = $self->log_pid_for('file');

          if (my $format = $arg->{file_format}) {
            sub {
              my $message = {@_}->{message};
              $message = "[$$] $message" if $log_pid;
              $format->($message)
            };
          } else {
            # The time format returned here is subject to change. -- rjbs,
            # 2008-11-21
            sub {
              my $message = {@_}->{message};
              $message = "[$$] $message" if $log_pid;
              (localtime) . " $message\n";
            };
          }
        },
      )
    );
  }

  if ($arg->{facility} and not $self->env_value('NOSYSLOG')) {
    $self->setup_syslog_output(
      facility  => $arg->{facility},
      socket    => $arg->{syslog_socket},
      ident     => $ident,
    );
  }

  if ($arg->{to_self}) {
    $self->{events} = [];
    require Log::Dispatch::Array;
    $log->add(
      Log::Dispatch::Array->new(
        name      => 'self',
        min_level => 'debug',
        array     => $self->{events},
        ($self->log_pid_for('self')
          ? (callbacks => sub { "[$$] ". {@_}->{message} })
          : ())
      ),
    );
  }

  $self->{prefix}     = $arg->{prefix};
  $self->{ident}      = $ident;
  $self->{config_id}  = $config_id;

  DEST: for my $dest (qw(err out)) {
    next DEST unless $arg->{"to_std$dest"};
    my $method = "enable_std$dest";

    $self->$method;
  }

  $self->{debug}  = exists $arg->{debug}
                  ? ($arg->{debug} ? 1 : 0)
                  : ($self->env_value('DEBUG') ? 1 : 0);
  $self->{muted}  = $arg->{muted};

  $self->{quiet_fatal} = \%quiet_fatal;
  $self->{fail_fatal}  = exists $arg->{fail_fatal} ? $arg->{fail_fatal} : 1;

  return $self;
}

sub log_pid_for ($self, $output) {
  my $log_pid = $self->{log_pid};
  return undef unless $log_pid;

  return 1 if $log_pid eq 1;

  $self->{log_pid_for} = { map {; $_ => 1 } split /,/, $log_pid };

  return $self->{log_pid_for}{$output} ? 1 : undef;
}

#pod =method enable_stdout
#pod
#pod =method enable_stderr
#pod
#pod These methods turn on logging to STDOUT or STDERR, respectively.  If that
#pod logging has already been enabled, these methods do nothing.
#pod
#pod There is not, yet, a I<disable> version of these methods.
#pod
#pod =cut

for my $dest (qw(out err)) {
  my $name = "std$dest";
  my $code = sub ($self) {
    return if $self->dispatcher->output($name);

    my $callback = $self->log_pid_for($name)
                 ? sub { "[$$] " . ({@_}->{message}) . "\n" }
                 : sub {           ({@_}->{message}) . "\n" };

    $self->dispatcher->add(
      $self->stdio_dispatcher_class->new(
        name      => "std$dest",
        min_level => 'debug',
        stderr    => ($dest eq 'err' ? 1 : 0),
        callbacks => $callback,
        ($self->{quiet_fatal}{"std$dest"} ? (max_level => 'info') : ()),
      ),
    );
  };

  no strict 'refs';
  *{"enable_std$dest"} = $code;
}

sub setup_syslog_output ($self, %arg) {
  require Log::Dispatch::Syslog;
  $self->{dispatcher}->add(
    Log::Dispatch::Syslog->new(
      name      => 'syslog',
      ident     => $arg{ident},
      facility  => $arg{facility},
      logopt    => ($self->log_pid_for('syslog') ? 'pid' : ''),
      min_level => 'debug',
      socket    => $arg{socket} || 'native',
      callbacks => sub {
        ( my $m = {@_}->{message} ) =~ s/\n/<LF>/g;
        $m
      },
    ),
  );
}

#pod =method log
#pod
#pod   $logger->log(@messages);
#pod
#pod   $logger->log(\%arg, @messages);
#pod
#pod This method uses L<String::Flogger> on the input, then I<unconditionally> logs
#pod the result.  Each message is flogged individually, then joined with spaces.
#pod
#pod If the first argument is a hashref, it will be used as extra arguments to
#pod logging.  It may include a C<prefix> entry to preprocess the message by
#pod prepending a string (if the prefix is a string) or calling a subroutine to
#pod generate a new message (if the prefix is a coderef).
#pod
#pod =cut

sub _flog_messages ($self, $arg, $rest) {
  my $flogger = $self->string_flogger;
  my @flogged = map {; $flogger->flog($_) } @$rest;
  my $message = @flogged > 1 ? join(q{ }, @flogged) : $flogged[0];

  my @prefix  = _ARRAY0($arg->{prefix})
              ? @{ $arg->{prefix} }
              : $arg->{prefix};

  for (reverse grep { defined } $self->get_prefix, @prefix) {
    if (_CODELIKE( $_ )) {
      $message = $_->($message);
    } else {
      $message =~ s/^/$_/gm;
    }
  }

  return $message;
}

sub flog_messages ($self, @rest) {
  my $arg = _HASH0($rest[0]) ? shift(@rest) : {};
  return $self->_flog_messages($arg, \@rest);
}

sub log ($self, @rest) {
  my $arg = _HASH0($rest[0]) ? shift(@rest) : {};

  my $message;

  if ($arg->{fatal} or ! $self->get_muted) {
    try {
      $message = $self->flog_messages($arg, @rest);

      $self->dispatcher->log(
        level   => $arg->{level} || 'info',
        message => $message,
      );
    } catch {
      $message = '(no message could be logged)' unless defined $message;
      die $_ if $self->{fail_fatal};
    };
  }

  Carp::croak $message if $arg->{fatal};

  return;
}

#pod =method log_fatal
#pod
#pod This behaves like the C<log> method, but will throw the logged string as an
#pod exception after logging.
#pod
#pod This method can also be called as C<fatal>, to match other popular logging
#pod interfaces.  B<If you want to override this method, you must override
#pod C<log_fatal> and not C<fatal>>.
#pod
#pod =cut

sub log_fatal ($self, @rest) {
  my $arg = _HASH0($rest[0]) ? shift(@rest) : {}; # for future expansion

  local $arg->{level} = defined $arg->{level} ? $arg->{level} : 'error';
  local $arg->{fatal} = defined $arg->{fatal} ? $arg->{fatal} : 1;

  $self->log($arg, @rest);
}

#pod =method log_debug
#pod
#pod This behaves like the C<log> method, but will only log (at the debug level) if
#pod the logger object has its debug property set to true.
#pod
#pod This method can also be called as C<debug>, to match other popular logging
#pod interfaces.  B<If you want to override this method, you must override
#pod C<log_debug> and not C<debug>>.
#pod
#pod =cut

sub log_debug ($self, @rest) {
  return unless $self->is_debug;

  my $arg = _HASH0($rest[0]) ? shift(@rest) : {}; # for future expansion

  local $arg->{level} = defined $arg->{level} ? $arg->{level} : 'debug';

  $self->log($arg, @rest);
}

#pod =method log_event
#pod
#pod This method is like C<log>, but is used for structured logging instead of free
#pod form text.  It's invoked like this:
#pod
#pod   $logger->log($event_type => $data_ref);
#pod
#pod C<$event_type> should be a simple string, probably a valid identifier, that
#pod identifies the kind of event being logged.  It is suggested, but not required,
#pod that all events of the same type have the same kind of structured data in them.
#pod
#pod C<$data_ref> is a set of key/value pairs of data to log in this event.  It can
#pod be an arrayref (in which case the ordering of pairs is preserved) or a hashref
#pod (in which case they are sorted by key).
#pod
#pod The logged string will be in logfmt format, meaning a series of key=value
#pod pairs separated by spaces and following these rules:
#pod
#pod =for :list
#pod * an "identifier" is a string of printable ASCII characters between C<!> and
#pod   C<~>, excluding C<\> and C<=>
#pod * keys must be valid identifiers
#pod * if a key is empty, C<~> is used instead
#pod * if a key contains characters not permitted in an identifier, they are
#pod   replaced by C<?>
#pod * values must I<either> be valid identifiers, or be quoted
#pod * quoted value start and end with C<">
#pod * in a quoted value, C<"> becomes C<\">, C<\> becomes C<\\>, newline and
#pod   carriage return become C<\n> and C<\r> respectively, and other control
#pod   characters are replaced with C<\u{....}> where the contents of the braces are
#pod   the hex value of the control character
#pod
#pod When values are undef, they are represented as C<~>.
#pod
#pod When values are array references, the index/values are mapped over, so that:
#pod
#pod   key => [ 'a', 'b' ]
#pod
#pod becomes
#pod
#pod   key.0=a key.1=b
#pod
#pod When values are hash references, the key/values are mapped, with keys sorted,
#pod so that:
#pod
#pod   key => { b => 2, a => 1 }
#pod
#pod becomes
#pod
#pod   key.a=1 key.b=2
#pod
#pod This expansion is performed recursively.  If a value itself recurses,
#pod appearances of a reference after the first time will be replaced with a string
#pod like C<&foo.bar>, pointing to the first occurrence.  I<This is not meant to be
#pod a robust serialization mechanism.>  It's just here to help you be a little
#pod lazy.  Don't push the limits.
#pod
#pod If the value in C<$data_ref> is a code reference, it will be called and its
#pod result logged.  If its result is also a code reference, you get whatever
#pod garbage that code reference stringifies to.
#pod
#pod If the value in C<$data_ref> is a reference reference, then the referenced
#pod scalar will be passed to String::Flogger, and the resulting string will be used
#pod as the value to log.  That string will be quoted as described above, if needed.
#pod
#pod =cut

sub _compute_proxy_ctx_kvstr_aref ($) {
  return [];
}

my $LOG_FMT_PACKAGE;
sub _log_fmt_package { $LOG_FMT_PACKAGE }

BEGIN {
  $LOG_FMT_PACKAGE = 'Log::Fmt';
  my $ok = eval { require Log::Fmt::XS; };
  if ($ok && ! $ENV{LOG_FMT_NO_XS}) {
    $LOG_FMT_PACKAGE = 'Log::Fmt::XS';
  }
}

sub fmt_event ($self, $type, $data) {
  my $kv_aref = $self->_log_fmt_package->_pairs_to_kvstr_aref([
    event => $type,
    (_ARRAY0($data) ? @$data : $data->%{ sort keys %$data })
  ]);

  return join q{ }, @$kv_aref;
}

sub log_event ($self, $type, $data) {
  return if $self->get_muted;

  my $message = $self->fmt_event($type, $data);

  $self->dispatcher->log(
    level   => 'info',
    message => $message,
  );

  return;
}

#pod =method log_debug_event
#pod
#pod This method is just like C<log_event>, but will log nothing unless the logger
#pod has its C<debug> property set to true.
#pod
#pod =cut

sub log_debug_event ($self, $type, $data) {
  return unless $self->get_debug;

  $self->log_event($type, $data);
}

#pod =method set_debug
#pod
#pod   $logger->set_debug($bool);
#pod
#pod This sets the logger's debug property, which affects the behavior of
#pod C<log_debug>.
#pod
#pod =cut

sub set_debug ($self, $bool) { $self->{debug} = $bool ? 1 : 0 }

#pod =method get_debug
#pod
#pod This gets the logger's debug property, which affects the behavior of
#pod C<log_debug>.
#pod
#pod =cut

sub get_debug ($self) { return $self->{debug} }

#pod =method clear_debug
#pod
#pod This method does nothing, and is only useful for L<Log::Dispatchouli::Proxy>
#pod objects.  See L<Methods for Proxy Loggers|/METHODS FOR PROXY LOGGERS>, below.
#pod
#pod =cut

sub clear_debug ($) { }

sub mute   ($self) { $self->{muted} = 1 }
sub unmute ($self) { $self->{muted} = 0 }

#pod =method set_muted
#pod
#pod   $logger->set_muted($bool);
#pod
#pod This sets the logger's muted property, which affects the behavior of
#pod C<log>.
#pod
#pod =cut

sub set_muted ($self, $bool) {
  return ($self->{muted} = $bool ? 1 : 0);
}

#pod =method get_muted
#pod
#pod This gets the logger's muted property, which affects the behavior of
#pod C<log>.
#pod
#pod =cut

sub get_muted ($self) { return $self->{muted} }

#pod =method clear_muted
#pod
#pod This method does nothing, and is only useful for L<Log::Dispatchouli::Proxy>
#pod objects.  See L<Methods for Proxy Loggers|/METHODS FOR PROXY LOGGERS>, below.
#pod
#pod =cut

sub clear_muted ($) { }

#pod =method get_prefix
#pod
#pod   my $prefix = $logger->get_prefix;
#pod
#pod This method returns the currently-set prefix for the logger, which may be a
#pod string or code reference or undef.  See L<Logger Prefix|/LOGGER PREFIX>.
#pod
#pod =method set_prefix
#pod
#pod   $logger->set_prefix( $new_prefix );
#pod
#pod This method changes the prefix.  See L<Logger Prefix|/LOGGER PREFIX>.
#pod
#pod =method clear_prefix
#pod
#pod This method clears any set logger prefix.  (It can also be called as
#pod C<unset_prefix>, but this is deprecated.  See L<Logger Prefix|/LOGGER PREFIX>.
#pod
#pod =cut

sub get_prefix   ($self)          { return $self->{prefix}    }
sub set_prefix   ($self, $prefix) { $self->{prefix} = $prefix }
sub clear_prefix ($self)          { $self->unset_prefix       }
sub unset_prefix ($self)          { undef $self->{prefix}     }

#pod =method ident
#pod
#pod This method returns the logger's ident.
#pod
#pod =cut

sub ident ($self) { $self->{ident} }

#pod =method config_id
#pod
#pod This method returns the logger's configuration id, which defaults to its ident.
#pod This can be used to make two loggers equivalent in Log::Dispatchouli::Global so
#pod that trying to reinitialize with a new logger with the same C<config_id> as the
#pod current logger will not throw an exception, and will simply do no thing.
#pod
#pod =cut

sub config_id ($self) { $self->{config_id} }

#pod =head1 METHODS FOR SUBCLASSING
#pod
#pod =head2 string_flogger
#pod
#pod This method returns the thing on which F<flog> will be called to format log
#pod messages.  By default, it just returns C<String::Flogger>
#pod
#pod =cut

sub string_flogger ($) { 'String::Flogger' }

#pod =head2 env_prefix
#pod
#pod This method should return a string used as a prefix to find environment
#pod variables that affect the logger's behavior.  For example, if this method
#pod returns C<XYZZY> then when checking the environment for a default value for the
#pod C<debug> parameter, Log::Dispatchouli will first check C<XYZZY_DEBUG>, then
#pod C<DISPATCHOULI_DEBUG>.
#pod
#pod By default, this method returns C<()>, which means no extra environment
#pod variable is checked.
#pod
#pod =cut

sub env_prefix ($) { return; }

#pod =head2 env_value
#pod
#pod   my $value = $logger->env_value('DEBUG');
#pod
#pod This method returns the value for the environment variable suffix given.  For
#pod example, the example given, calling with C<DEBUG> will check
#pod C<DISPATCHOULI_DEBUG>.
#pod
#pod =cut

sub env_value ($self, $suffix) {
  my @path = grep { defined } ($self->env_prefix, 'DISPATCHOULI');

  for my $prefix (@path) {
    my $name = join q{_}, $prefix, $suffix;
    return $ENV{ $name } if defined $ENV{ $name };
  }

  return;
}

#pod =method flog_messages
#pod
#pod   my $str = $logger->flog_messages($m1, $m2, ...);
#pod
#pod This returns the string that would have been logged if the given arguments had
#pod been passed to C<< $logger->log(...) >>, without regard for log level,
#pod debugging, or the like.
#pod
#pod Unlike using the logger's string flogger, this will include any relevant prefix
#pod strings.
#pod
#pod =method fmt_event
#pod
#pod   my $str = $logger->fmt_event($event_type => $data_ref);
#pod
#pod This method is equivalent to C<flog_messages>, but for an event.  It returns
#pod the string format of the event, including all relevant prefixes.
#pod
#pod =head1 METHODS FOR TESTING
#pod
#pod =head2 new_tester
#pod
#pod   my $logger = Log::Dispatchouli->new_tester( \%arg );
#pod
#pod This returns a new logger that logs only C<to_self>.  It's useful in testing.
#pod If no C<ident> arg is provided, one will be generated.  C<log_pid> is off by
#pod default, but can be overridden.
#pod
#pod C<\%arg> is optional.
#pod
#pod =cut

sub new_tester ($class, $arg = undef) {
  $arg ||= {};

  return $class->new({
    ident     => "$$:$0",
    log_pid   => 0,
    %$arg,
    to_stderr => 0,
    to_stdout => 0,
    to_file   => 0,
    to_self   => 1,
    facility  => undef,
  });
}

#pod =head2 events
#pod
#pod This method returns the arrayref of events logged to an array in memory (in the
#pod logger).  If the logger is not logging C<to_self> this raises an exception.
#pod
#pod =cut

sub events ($self) {
  Carp::confess "->events called on a logger not logging to self"
    unless $self->{events};

  return $self->{events};
}

#pod =head2 clear_events
#pod
#pod This method empties the current sequence of events logged into an array in
#pod memory.  If the logger is not logging C<to_self> this raises an exception.
#pod
#pod =cut

sub clear_events ($self) {
  Carp::confess "->events called on a logger not logging to self"
    unless $self->{events};

  $self->{events}->@* = ();
  return;
}

#pod =head1 METHODS FOR PROXY LOGGERS
#pod
#pod =head2 proxy
#pod
#pod   my $proxy_logger = $logger->proxy( \%arg );
#pod
#pod This method returns a new proxy logger -- an instance of
#pod L<Log::Dispatchouli::Proxy> -- which will log through the given logger, but
#pod which may have some settings localized.
#pod
#pod C<%arg> is optional.  It may contain the following entries:
#pod
#pod =for :list
#pod = proxy_prefix
#pod This is a prefix that will be applied to anything the proxy logger logs, and
#pod cannot be changed.
#pod = proxy_ctx
#pod This is data to be inserted in front of event data logged through the proxy.
#pod It will appear I<after> the C<event> key but before the logged event data.  It
#pod can be in the same format as the C<$data_ref> argument to C<log_event>.
#pod = debug
#pod This can be set to true or false to change the proxy's "am I in debug mode?"
#pod setting.  It can be changed or cleared later on the proxy.
#pod
#pod =cut

sub proxy_class ($) {
  return 'Log::Dispatchouli::Proxy';
}

sub proxy ($self, $arg = undef) {
  $arg ||= {};

  my $proxy = $self->proxy_class->_new({
    parent => $self,
    logger => $self,
    proxy_prefix => $arg->{proxy_prefix},
    (exists $arg->{debug} ? (debug => ($arg->{debug} ? 1 : 0)) : ()),
  });

  if (my $ctx = $arg->{proxy_ctx}) {
    $proxy->{proxy_ctx} = _ARRAY0($ctx)
                        ? [ @$ctx ]
                        : [ $ctx->%{ sort keys %$ctx } ];
  }

  return $proxy;
}

#pod =head2 parent
#pod
#pod =head2 logger
#pod
#pod These methods return the logger itself.  (They're more useful when called on
#pod proxy loggers.)
#pod
#pod =cut

sub parent ($self) { $self }
sub logger ($self) { $self }

#pod =method dispatcher
#pod
#pod This returns the underlying Log::Dispatch object.  This is not the method
#pod you're looking for.  Move along.
#pod
#pod =cut

sub dispatcher ($self) { $self->{dispatcher} }

#pod =method stdio_dispatcher_class
#pod
#pod This method is an experimental feature to allow you to pick an alternate
#pod dispatch class for stderr and stdio.  By default, Log::Dispatch::Screen is
#pod used.  B<This feature may go away at any time.>
#pod
#pod =cut

sub stdio_dispatcher_class ($self) {
  require Log::Dispatch::Screen;
  return 'Log::Dispatch::Screen';
}

#pod =head1 METHODS FOR API COMPATIBILITY
#pod
#pod To provide compatibility with some other loggers, most specifically
#pod L<Log::Contextual>, the following methods are provided.  You should not use
#pod these methods without a good reason, and you should never subclass them.
#pod Instead, subclass the methods they call.
#pod
#pod =begin :list
#pod
#pod = is_debug
#pod
#pod This method calls C<get_debug>.
#pod
#pod = is_info
#pod
#pod = is_fatal
#pod
#pod These methods return true.
#pod
#pod = info
#pod
#pod = fatal
#pod
#pod = debug
#pod
#pod These methods redispatch to C<log>, C<log_fatal>, and C<log_debug>
#pod respectively.
#pod
#pod =end :list
#pod
#pod =cut

sub is_debug ($self) { $self->get_debug }
sub is_info  ($) { 1 }
sub is_fatal ($) { 1 }

sub info  ($self, @rest) { $self->log(@rest); }
sub fatal ($self, @rest) { $self->log_fatal(@rest); }
sub debug ($self, @rest) { $self->log_debug(@rest); }

use overload
  '&{}'    => sub { my ($self) = @_; sub { $self->log(@_) } },
  fallback => 1,
;

#pod =head1 SEE ALSO
#pod
#pod =for :list
#pod * L<Log::Dispatch>
#pod * L<String::Flogger>
#pod
#pod =cut

1;

__END__

=pod

=encoding UTF-8

=head1 NAME

Log::Dispatchouli - a simple wrapper around Log::Dispatch

=head1 VERSION

version 3.101

=head1 SYNOPSIS

  my $logger = Log::Dispatchouli->new({
    ident     => 'stuff-purger',
    facility  => 'daemon',
    to_stdout => $opt->{print},
    debug     => $opt->{verbose}
  });

  $logger->log([ "There are %s items left to purge...", $stuff_left ]);

  $logger->log_debug("this is extra often-ignored debugging log");

  $logger->log_fatal("Now we will die!!");

=head1 DESCRIPTION

Log::Dispatchouli is a thin layer above L<Log::Dispatch> and meant to make it
dead simple to add logging to a program without having to think much about
categories, facilities, levels, or things like that.  It is meant to make
logging just configurable enough that you can find the logs you want and just
easy enough that you will actually log things.

Log::Dispatchouli can log to syslog (if you specify a facility), standard error
or standard output, to a file, or to an array in memory.  That last one is
mostly useful for testing.

In addition to providing as simple a way to get a handle for logging
operations, Log::Dispatchouli uses L<String::Flogger> to process the things to
be logged, meaning you can easily log data structures.  Basically: strings are
logged as is, arrayrefs are taken as (sprintf format, args), and subroutines
are called only if needed.  For more information read the L<String::Flogger>
docs.

=head1 PERL VERSION

This library should run on perls released even a long time ago.  It should
work on any version of perl released in the last five years.

Although it may work on older versions of perl, no guarantee is made that the
minimum required version will not be increased.  The version may be increased
for any reason, and there is no promise that patches will be accepted to
lower the minimum required perl.

=head1 METHODS

=head2 new

  my $logger = Log::Dispatchouli->new(\%arg);

This returns a new logger, a Log::Dispatchouli object.

Valid arguments are:

  ident       - the name of the thing logging (mandatory)
  to_self     - log to the logger object for testing; default: false
  to_stdout   - log to STDOUT; default: false
  to_stderr   - log to STDERR; default: false
  facility    - to which syslog facility to send logs; default: none

  to_file     - DEPRECATED: this option will be removed in 2025
                log to PROGRAM_NAME.YYYYMMDD in the log path; default: false
  log_file    - a leaf name for the file to log to with to_file
  log_path    - path in which to log to file; defaults to DISPATCHOULI_PATH
                environment variable or, failing that, to your system's tmpdir

  file_format - this optional coderef is passed the message to be logged
                and returns the text to write out

  log_pid     - if 1, prefix all log entries with the pid; default: true
                can also be a comma-delimited list of log targets where pid is
                logged, like "stderr,syslog"; mostly useful for logging pid in
                syslog, but not on standard I/O
  fail_fatal  - a boolean; if true, failure to log is fatal; default: true
  muted       - a boolean; if true, only fatals are logged; default: false
  debug       - a boolean; if true, log_debug method is not a no-op
                defaults to the truth of the DISPATCHOULI_DEBUG env var
  quiet_fatal - 'stderr' or 'stdout' or an arrayref of zero, one, or both
                fatal log messages will not be logged to these
                (default: stderr)
  config_id   - a name for this logger's config; rarely needed!
  syslog_socket - a value for Sys::Syslog's "socket" arg; default: "native"

The log path is either F</tmp> or the value of the F<DISPATCHOULI_PATH> env var.

If the F<DISPATCHOULI_NOSYSLOG> env var is true, we don't log to syslog.

=head2 enable_stdout

=head2 enable_stderr

These methods turn on logging to STDOUT or STDERR, respectively.  If that
logging has already been enabled, these methods do nothing.

There is not, yet, a I<disable> version of these methods.

=head2 log

  $logger->log(@messages);

  $logger->log(\%arg, @messages);

This method uses L<String::Flogger> on the input, then I<unconditionally> logs
the result.  Each message is flogged individually, then joined with spaces.

If the first argument is a hashref, it will be used as extra arguments to
logging.  It may include a C<prefix> entry to preprocess the message by
prepending a string (if the prefix is a string) or calling a subroutine to
generate a new message (if the prefix is a coderef).

=head2 log_fatal

This behaves like the C<log> method, but will throw the logged string as an
exception after logging.

This method can also be called as C<fatal>, to match other popular logging
interfaces.  B<If you want to override this method, you must override
C<log_fatal> and not C<fatal>>.

=head2 log_debug

This behaves like the C<log> method, but will only log (at the debug level) if
the logger object has its debug property set to true.

This method can also be called as C<debug>, to match other popular logging
interfaces.  B<If you want to override this method, you must override
C<log_debug> and not C<debug>>.

=head2 log_event

This method is like C<log>, but is used for structured logging instead of free
form text.  It's invoked like this:

  $logger->log($event_type => $data_ref);

C<$event_type> should be a simple string, probably a valid identifier, that
identifies the kind of event being logged.  It is suggested, but not required,
that all events of the same type have the same kind of structured data in them.

C<$data_ref> is a set of key/value pairs of data to log in this event.  It can
be an arrayref (in which case the ordering of pairs is preserved) or a hashref
(in which case they are sorted by key).

The logged string will be in logfmt format, meaning a series of key=value
pairs separated by spaces and following these rules:

=over 4

=item *

an "identifier" is a string of printable ASCII characters between C<!> and C<~>, excluding C<\> and C<=>

=item *

keys must be valid identifiers

=item *

if a key is empty, C<~> is used instead

=item *

if a key contains characters not permitted in an identifier, they are replaced by C<?>

=item *

values must I<either> be valid identifiers, or be quoted

=item *

quoted value start and end with C<">

=item *

in a quoted value, C<"> becomes C<\">, C<\> becomes C<\\>, newline and carriage return become C<\n> and C<\r> respectively, and other control characters are replaced with C<\u{....}> where the contents of the braces are the hex value of the control character

=back

When values are undef, they are represented as C<~>.

When values are array references, the index/values are mapped over, so that:

  key => [ 'a', 'b' ]

becomes

  key.0=a key.1=b

When values are hash references, the key/values are mapped, with keys sorted,
so that:

  key => { b => 2, a => 1 }

becomes

  key.a=1 key.b=2

This expansion is performed recursively.  If a value itself recurses,
appearances of a reference after the first time will be replaced with a string
like C<&foo.bar>, pointing to the first occurrence.  I<This is not meant to be
a robust serialization mechanism.>  It's just here to help you be a little
lazy.  Don't push the limits.

If the value in C<$data_ref> is a code reference, it will be called and its
result logged.  If its result is also a code reference, you get whatever
garbage that code reference stringifies to.

If the value in C<$data_ref> is a reference reference, then the referenced
scalar will be passed to String::Flogger, and the resulting string will be used
as the value to log.  That string will be quoted as described above, if needed.

=head2 log_debug_event

This method is just like C<log_event>, but will log nothing unless the logger
has its C<debug> property set to true.

=head2 set_debug

  $logger->set_debug($bool);

This sets the logger's debug property, which affects the behavior of
C<log_debug>.

=head2 get_debug

This gets the logger's debug property, which affects the behavior of
C<log_debug>.

=head2 clear_debug

This method does nothing, and is only useful for L<Log::Dispatchouli::Proxy>
objects.  See L<Methods for Proxy Loggers|/METHODS FOR PROXY LOGGERS>, below.

=head2 set_muted

  $logger->set_muted($bool);

This sets the logger's muted property, which affects the behavior of
C<log>.

=head2 get_muted

This gets the logger's muted property, which affects the behavior of
C<log>.

=head2 clear_muted

This method does nothing, and is only useful for L<Log::Dispatchouli::Proxy>
objects.  See L<Methods for Proxy Loggers|/METHODS FOR PROXY LOGGERS>, below.

=head2 get_prefix

  my $prefix = $logger->get_prefix;

This method returns the currently-set prefix for the logger, which may be a
string or code reference or undef.  See L<Logger Prefix|/LOGGER PREFIX>.

=head2 set_prefix

  $logger->set_prefix( $new_prefix );

This method changes the prefix.  See L<Logger Prefix|/LOGGER PREFIX>.

=head2 clear_prefix

This method clears any set logger prefix.  (It can also be called as
C<unset_prefix>, but this is deprecated.  See L<Logger Prefix|/LOGGER PREFIX>.

=head2 ident

This method returns the logger's ident.

=head2 config_id

This method returns the logger's configuration id, which defaults to its ident.
This can be used to make two loggers equivalent in Log::Dispatchouli::Global so
that trying to reinitialize with a new logger with the same C<config_id> as the
current logger will not throw an exception, and will simply do no thing.

=head2 flog_messages

  my $str = $logger->flog_messages($m1, $m2, ...);

This returns the string that would have been logged if the given arguments had
been passed to C<< $logger->log(...) >>, without regard for log level,
debugging, or the like.

Unlike using the logger's string flogger, this will include any relevant prefix
strings.

=head2 fmt_event

  my $str = $logger->fmt_event($event_type => $data_ref);

This method is equivalent to C<flog_messages>, but for an event.  It returns
the string format of the event, including all relevant prefixes.

=head2 dispatcher

This returns the underlying Log::Dispatch object.  This is not the method
you're looking for.  Move along.

=head2 stdio_dispatcher_class

This method is an experimental feature to allow you to pick an alternate
dispatch class for stderr and stdio.  By default, Log::Dispatch::Screen is
used.  B<This feature may go away at any time.>

=head1 LOGGER PREFIX

Log messages may be prepended with information to set context.  This can be set
at a logger level or per log item.  The simplest example is:

  my $logger = Log::Dispatchouli->new( ... );

  $logger->set_prefix("Batch 123: ");

  $logger->log("begun processing");

  # ...

  $logger->log("finished processing");

The above will log something like:

  Batch 123: begun processing
  Batch 123: finished processing

To pass a prefix per-message:

  $logger->log({ prefix => 'Sub-Item 234: ' }, 'error!')

  # Logs: Batch 123: Sub-Item 234: error!

If the prefix is a string, it is prepended to each line of the message.  If it
is a coderef, it is called and passed the message to be logged.  The return
value is logged instead.

L<Proxy loggers|/METHODS FOR PROXY LOGGERS> also have their own prefix
settings, which accumulate.  So:

  my $proxy = $logger->proxy({ proxy_prefix => 'Subsystem 12: ' });

  $proxy->set_prefix('Page 9: ');

  $proxy->log({ prefix => 'Paragraph 6: ' }, 'Done.');

...will log...

  Batch 123: Subsystem 12: Page 9: Paragraph 6: Done.

=head1 METHODS FOR SUBCLASSING

=head2 string_flogger

This method returns the thing on which F<flog> will be called to format log
messages.  By default, it just returns C<String::Flogger>

=head2 env_prefix

This method should return a string used as a prefix to find environment
variables that affect the logger's behavior.  For example, if this method
returns C<XYZZY> then when checking the environment for a default value for the
C<debug> parameter, Log::Dispatchouli will first check C<XYZZY_DEBUG>, then
C<DISPATCHOULI_DEBUG>.

By default, this method returns C<()>, which means no extra environment
variable is checked.

=head2 env_value

  my $value = $logger->env_value('DEBUG');

This method returns the value for the environment variable suffix given.  For
example, the example given, calling with C<DEBUG> will check
C<DISPATCHOULI_DEBUG>.

=head1 METHODS FOR TESTING

=head2 new_tester

  my $logger = Log::Dispatchouli->new_tester( \%arg );

This returns a new logger that logs only C<to_self>.  It's useful in testing.
If no C<ident> arg is provided, one will be generated.  C<log_pid> is off by
default, but can be overridden.

C<\%arg> is optional.

=head2 events

This method returns the arrayref of events logged to an array in memory (in the
logger).  If the logger is not logging C<to_self> this raises an exception.

=head2 clear_events

This method empties the current sequence of events logged into an array in
memory.  If the logger is not logging C<to_self> this raises an exception.

=head1 METHODS FOR PROXY LOGGERS

=head2 proxy

  my $proxy_logger = $logger->proxy( \%arg );

This method returns a new proxy logger -- an instance of
L<Log::Dispatchouli::Proxy> -- which will log through the given logger, but
which may have some settings localized.

C<%arg> is optional.  It may contain the following entries:

=over 4

=item proxy_prefix

This is a prefix that will be applied to anything the proxy logger logs, and
cannot be changed.

=item proxy_ctx

This is data to be inserted in front of event data logged through the proxy.
It will appear I<after> the C<event> key but before the logged event data.  It
can be in the same format as the C<$data_ref> argument to C<log_event>.

=item debug

This can be set to true or false to change the proxy's "am I in debug mode?"
setting.  It can be changed or cleared later on the proxy.

=back

=head2 parent

=head2 logger

These methods return the logger itself.  (They're more useful when called on
proxy loggers.)

=head1 METHODS FOR API COMPATIBILITY

To provide compatibility with some other loggers, most specifically
L<Log::Contextual>, the following methods are provided.  You should not use
these methods without a good reason, and you should never subclass them.
Instead, subclass the methods they call.

=over 4

=item is_debug

This method calls C<get_debug>.

=item is_info

=item is_fatal

These methods return true.

=item info

=item fatal

=item debug

These methods redispatch to C<log>, C<log_fatal>, and C<log_debug>
respectively.

=back

=head1 SEE ALSO

=over 4

=item *

L<Log::Dispatch>

=item *

L<String::Flogger>

=back

=head1 AUTHOR

Ricardo SIGNES <[e-mail]>

=head1 CONTRIBUTORS

=for stopwords Charlie Garrison Christopher J. Madsen Dagfinn Ilmari Mannsåker Dan Book George Hartzell Jon Stuart Matt Phillips Olivier Mengué Randy Stauner Ricardo Signes Sawyer X

=over 4

=item *

Charlie Garrison <[e-mail]>

=item *

Christopher J. Madsen <[e-mail]>

=item *

Dagfinn Ilmari Mannsåker <[e-mail]>

=item *

Dan Book <[e-mail]>

=item *

George Hartzell <[e-mail]>

=item *

Jon Stuart <[e-mail]>

=item *

Matt Phillips <[e-mail]>

=item *

Olivier Mengué <[e-mail]>

=item *

Randy Stauner <[e-mail]>

=item *

Ricardo Signes <[e-mail]>

=item *

Sawyer X <[e-mail]>

=back

=head1 COPYRIGHT AND LICENSE

This software is copyright (c) 2026 by Ricardo SIGNES.

This is free software; you can redistribute it and/or modify it under
the same terms as the Perl 5 programming language system itself.

=cut