Log::Agent::Tag::Caller - formats caller information


Log-Agent documentation Contained in the Log-Agent distribution.

Index


Code Index:

NAME

Top

Log::Agent::Tag::Caller - formats caller information

SYNOPSIS

Top

 Not intended to be used directly
 Inherits from Log::Agent::Tag.

DESCRIPTION

Top

This class handles caller information for Log::Agent services and is not meant to be used directly.

This manpage therefore only documents the creation routine parameters that can be specified at the Log::Agent level via the -caller switch in the logconfig() routine.

CALLER INFORMATION ENTITIES

Top

This class knows about four entities: package, filename, line and subroutine, which are to be understood within the context of the Log::Agent routine being called (e.g. a logwarn() routine), namely:

package

This is the package name where the call to the logwarn() routine was made. It can be specified as "pack" for short, or spelled out completely.

filename

This is the file where the call to the logwarn() routine was made. It can be specified as "file" for short, or spelled out completely.

line

This is the line number where the call to the logwarn() routine was made, in file filename. The name is short enough to be spelled out completely.

subroutine

This is the subroutine where the call to the logwarn() routine was made. If the call is made outside a subroutine, this will be empty. The name is long enough to warrant the "sub" abbreviation if you don't wish to spell it out fully.

CREATION ROUTINE PARAMETERS

Top

The purpose of those parameters is to define how caller information entities (as defined by the previous section) will be formatted within the log message.

-display => string

Specifies a string with minimal variable substitution: only the caller information entities specified above, or their abbreviation, will be interpolated. For instance:

    -display => '($package::$sub/$line)'

Don't forget to use simple quotes to avoid having Perl interpolate those as variables, or escape their leading $ sign otherwise. Using this convention was deemed to more readable (and natural in Perl) than SGML entities such as "&pack;".

Using this switch supersedes the -info and -format switches.

-format => printf format

Formatting instructions for the caller information entities listed by the -info switch. For instance:

    -format => "%s:%4d"

if you have specified two entities in -info.

The special formatting macro %a stands for all the entities specified by -info and is rendered by a string where values are separated by ":".

-info => "space separated list of parameters"

Specifies a list of caller information entities that are to be formated using the -format specification. For instance:

    -info => "pack sub line"

would only report those three entites.

-postfix => flag

Whether the string resulting from the formatting of the caller information entities should be appended to the regular log message or not (i.e. prepended, which is the default).

-separator => string

The separation string between the tag and the log message. A single space by default.

AUTHORS

Top

Raphael Manfredi <Raphael_Manfredi@pobox.com> created the module, it is currently maintained by Mark Rogaski <mrogaski@cpan.org>.

Thanks to Jeff Boes for uncovering wackiness in caller().

LICENSE

Top

Copyright (C) 1999 Raphael Manfredi. Copyright (C) 2002 Mark Rogaski; all rights reserved.

See Log::Agent(3) or the README file included with the distribution for license information.

SEE ALSO

Top

Log::Agent(3), Log::Agent::Message(3).


Log-Agent documentation Contained in the Log-Agent distribution.

###########################################################################
# $Id: Caller.pm,v 1.1 2002/03/09 16:09:45 wendigo Exp $
###########################################################################
#
# Log::Agent::Tag::Caller
#
# RCS Revision: $Revision: 1.1 $
# Date: $Date: 2002/03/09 16:09:45 $
#
# Copyright (C) 1999 Raphael Manfredi
# Copyright (C) 2002 Mark Rogaski, mrogaski@cpan.org; all rights reserved.
#
# See the README file included with the
# distribution for license information.
#
# $Log: Caller.pm,v $
# Revision 1.1  2002/03/09 16:09:45  wendigo
# Corrected  initialization
#
# Revision 0.2.1.2  2001/03/31 10:02:22  ram
# patch7: fixed =over to add explicit indent level
#
# Revision 0.2.1.1  2001/03/13 18:45:18  ram
# patch2: created
#
# Revision 0.2  2000/11/06 19:30:32  ram
# Baseline for second Alpha release.
#
###########################################################################

use strict;

########################################################################
package Log::Agent::Tag::Caller;

require Log::Agent::Tag;
use vars qw(@ISA);
@ISA = qw(Log::Agent::Tag);

#
# ->make
#
# Creation routine.
#
# Calling arguments: a hash table list.
#
# The keyed argument list may contain:
#    -OFFSET        value for the offset attribute [NOT DOCUMENTED]
#    -INFO        string of keywords like "package filename line subroutine"
#    -FORMAT        formatting instructions, like "%s:%d", used along with -INFO
#    -POSTFIX    whether to postfix log message or prefix it.
#   -DISPLAY    a string like '($subroutine/$line)', supersedes -INFO
#   -SEPARATOR  separator string to use between tag and message
#
# Attributes:
#    indices        listref of indices to select in the caller() array
#    offset        how many stack frames are between us and the caller we trace
#    format        how to format extracted caller() info
#    postfix        true if info to append to logged string
#
sub make {
    my $self = bless {}, shift;
    my (%args) = @_;

    $self->{'offset'} = 0;

    my $info;
    my $postfix = 0;
    my $separator;

    my %set = (
        -offset        => \$self->{'offset'},
        -info        => \$info,
        -format        => \$self->{'format'},
        -postfix    => \$postfix,
        -display    => \$self->{'display'},
        -separator    => \$separator,
    );

    while (my ($arg, $val) = each %args) {
        my $vset = $set{lc($arg)};
        next unless ref $vset;
        $$vset = $val;
    }

    $self->_init("caller", $postfix, $separator);

    return $self if $self->display;        # A display string takes precedence

    #
    # pre-process info to compute the indices
    #

    my $i = 0;
    my %indices = map { $_ => $i++ } qw(pac fil lin sub);    # abbrevs
    my @indices = ();

    foreach my $token (split(' ', $info)) {
        my $abbr = substr($token, 0, 3);
        push(@indices, $indices{$abbr}) if exists $indices{$abbr};
    }

    $self->{'indices'} = \@indices;

    return $self;
}

#
# Attribute access
#

sub offset        { $_[0]->{'offset'} }
sub indices        { $_[0]->{'indices'} }
sub format        { $_[0]->{'format'} }
sub display        { $_[0]->{'display'} }
sub postfix        { $_[0]->{'postfix'} }

#
# expand_a
#
# Expand the %a macro and return new string.
#
if ($] >= 5.005) { eval q{                # if VERSION >= 5.005

# 5.005 and later version grok /(?<!)/
sub expand_a {
        my ($str, $aref) = @_;
        $str =~ s/((?<!%)(?:%%)*)%a/join(':', @$aref)/ge;
        return $str;
}

}} else { eval q{                        # else /* VERSION < 5.005 */

# pre-5.005 does not grok /(?<!)/
sub expand_a {
        my ($str, $aref) = @_;
        $str =~ s/%%/\01/g;
        $str =~ s/%a/join(':', @$aref)/ge;
        $str =~ s/\01/%%/g;
        return $str;
}

}}                                        # endif /* VERSION >= 5.005 */

#
# ->string        -- defined
#
# Compute string with properly formatted caller info
#
sub string {
    my $self = shift;

    #
    # The following code:
    #
    #    sub foo {
    #        my ($pack, $file, $line, $sub) = caller(0);
    #        print "excuting $sub called at $file/$line in $pack";
    #    }
    #
    # will report who called us, except that $sub will be US, not our CALLER!
    # This is an "anomaly" somehow, and therefore to get the routine name
    # that called us, we need to move one frame above the ->offset value.
    #

    my @caller = caller($self->offset);
    
    # Kludge for anomalies in caller()
    # Thanks to Jeff Boes for finding the second one!
    $caller[3] = (caller($self->offset + 1))[3] || '(main)';

    my ($package, $filename, $line, $subroutine) = @caller;

    #
    # If there is a display, it takes precedence and is formatted accordingly,
    # with limited variable substitution. The variables that are recognized
    # are:
    #
    #        $package or $pack        package name of caller
    #        $filename or $file        filename of caller
    #        $line                    line number of caller
    #        $subroutine or $sub        routine name of caller
    #
    # We recognize both $line and ${line}, the difference being that the
    # first needs to be at a word boundary (i.e. $lineage would not result
    # in any expansion).
    #
    # Otherwise, the necessary information is gathered from the caller()
    # output, and formatted via sprintf, along with the special %a macro
    # which stands for all the information, separated by ':'.
    #
    # NB: The default format is "[%a]" for postfixed info, "(%a)" otherwise.
    #

    my $display = $self->display;
    if ($display) {
        $display =~ s/\$pack(?:age)?\b/$package/g;
        $display =~ s/\${pack(?:age)?}/$package/g;
        $display =~ s/\$file(?:name)?\b/$filename/g;
        $display =~ s/\${file(?:name)?}/$filename/g;
        $display =~ s/\$line\b/$line/g;
        $display =~ s/\${line}/$line/g;
        $display =~ s/\$sub(?:routine)?\b/$subroutine/g;
        $display =~ s/\${sub(?:routine)?}/$subroutine/g;
    } else {
        my @show = map { $caller[$_] } @{$self->indices};
        my $format = $self->format || ($self->postfix ? "[%a]" : "(%a)");
        $format = expand_a($format, \@show);    # depends on Perl's version
        $display = sprintf $format, @show;
    }

    return $display;
}

1;            # for "require"
__END__