NAME

Log::Abstraction - Logging Abstraction Layer

VERSION

0.39

SYNOPSIS

use Log::Abstraction;

# The default level is 'warning'; 'trace' lets every example through
my $logger = Log::Abstraction->new(logger => 'logfile.log', level => 'trace');

$logger->debug('This is a debug message');
$logger->info('This is an info message');
$logger->notice('This is a notice message');
$logger->trace('This is a trace message');
$logger->warn({ warning => 'This is a warning message' });

# Structured fields
$logger->info('User logged in', { user_id => 42 });

DESCRIPTION

The Log::Abstraction class provides a flexible logging layer on top of different types of loggers, including code references, arrays, file paths, and objects. It also supports logging to syslog if configured.

Unicode

Messages may be character strings containing any Unicode text. The file, fd and scalar-path backends write character strings as UTF-8 (unless an fd handle already has a :utf8 or :encoding layer, in which case the handle does the encoding), and format => 'json' output is UTF-8 too. journald fields are sent as UTF-8. Byte strings are written unchanged.

Structured fields

Every logging method accepts a hash reference of structured fields after the message:

$logger->info('User logged in', { user_id => 42, ip => $ip });
$logger->warn('Slow query', { ms => 1250 });

A hash reference is taken as fields only when it is the last of two or more arguments, so warn({ warning => ... }) keeps its meaning, and a lone hash reference is still a message. An empty hash reference is ignored. The fields are copied, so changing the hash afterwards doesn't change what was logged. They are kept out of the message and go to each backend as follows:

  • "messages", array and an ARRAY logger -- a fields key in the entry, alongside level and message.

  • a CODE logger -- a fields key in the hashref it is called with.

  • format => 'json' -- a nested fields object. Objects are stringified; other references are kept as JSON data.

  • journald -- journal fields. Each name is upper-cased, characters other than A-Z, 0-9 and _ become _, leading underscores are removed and it is cut to 64 characters; a field left with no name is dropped. Fields override the extra keys in the journald hash, but never MESSAGE, PRIORITY or SYSLOG_IDENTIFIER.

  • text formats (file, fd, a scalar logger), syslog, sendmail and object loggers -- appended to the message as logfmt-style key=value pairs in key order, e.g. User logged in ip=10.0.0.1 user_id=42. Characters other than [\w.-] in a key become _. A value that is empty or contains white space, ", = or \ is double-quoted, with " and \ escaped and control characters written as \n, \r, \t or \xNN. Objects are stringified and other references written as JSON. An object logger is passed the pairs as an extra argument after the message.

When logging through Log::Any, a hash reference at the end of the call, together with the proxy's context, arrives here as fields; see "structured" in Log::Any::Adapter::Abstraction.

Redaction

The redact option removes secrets before a message reaches the history or any backend. Each match of any of its patterns is replaced with [REDACTED]:

my $log = Log::Abstraction->new(
    logger => '/var/log/myapp.log',
    redact => [qr/password=\S+/, qr/\b\d{4}(?:[ -]?\d{4}){3}\b/],
);
$log->warn('Login failed: user=fred password=hunter2');
# Login failed: user=fred [REDACTED]

To keep the start of a match, end it with \K: qr/password=\K\S+/ logs password=[REDACTED]. The patterns run over the whole message, after its arguments are joined, so a match may span them, as in $log->info('password=', $password); a CODE or object logger then gets the joined message as one argument. The text given to carp or croak (see "warn" and "error") is redacted too.

"Structured fields" are redacted as well: string values, including those inside plain hashes and arrays, and objects whose stringification matches (they become the redacted string). Field names aren't redacted, nor are the format string, %env_*% values or ctx.

Redaction is applied only to messages that pass the logger's level, so it costs nothing for messages that are dropped. A pattern that can match the empty string, such as qr/x*/, is rejected by "new".

Email digests

min_interval limits the sendmail backend to one email per interval, dropping the messages in between. With digest, they are held instead, and sent ahead of the next message in the first email after the interval:

my $log = Log::Abstraction->new(
    logger => {
        file     => '/var/log/myapp.log',
        sendmail => {
            to           => 'ops@example.com',
            level        => 'error',
            min_interval => 300,
            digest       => 1,
            format       => '%level%> [%timestamp%] %message%',
        },
    },
);

Each message is one line of the email, oldest first, as its own email body would be: formatted with the sendmail format if there is one (as above, so that each line has its time), else the message. At most digest_max messages (default 100) are held; later ones are only counted, and the email says ... and N more messages after the held ones. Messages held when an email fails to send stay held, and the new one is held with them.

Nothing is sent on a timer: held messages go out with the next email, when "flush" is called, or when the logger is destroyed (which calls "flush"). A clone made with "new" starts with none held.

Per-backend level and format

Every backend can have its own level and format, as well as the logger's. syslog, journald and sendmail are hashes already, so they take them as keys. file, fd and array, inside a logger hash or at the top level, may be given as a hash holding the destination under the backend's own name:

my $log = Log::Abstraction->new(
    level  => 'debug',
    logger => {
        file   => { file => '/var/log/myapp.log', format => 'json' },
        fd     => { fd => \*STDERR, level => 'warning' },
        array  => { array => \@recent, level => 'info' },
        syslog => { level => 'error', format => '%level%: %message%' },
    },
);
  • level -- the backend only gets messages at this level or more severe. A level name or a syslog number (0-7). The logger's level is applied first, so a backend's level can only narrow it: set the logger's level to the most verbose any backend wants.

  • format -- a format string, or json, as for the logger's "format", which it overrides. For file and fd it is the line written. For the others, which by default get the message as it is, it replaces the message: the array entry's message, the text sent to syslog, the journal's MESSAGE field and the email body.

The plain forms (file => $path, fd => $handle, array => \@array) still work and have no level or format of their own. A blessed handle object is a destination, not the hash form.

File rotation

Log files written by path (a scalar logger, a file key in a logger hash, and the top-level file) can be rotated by size, by time, or both:

my $log = Log::Abstraction->new(
    file            => '/var/log/myapp.log',
    rotate_size     => '10M',
    rotate_interval => 'daily',
    rotate_keep     => 7,
);

Before each write the file is checked, and if it is due it is renamed to myapp.log.1, the old .1 to .2 and so on, the oldest beyond rotate_keep being deleted; the line then goes to a new myapp.log. A rotation that fails (e.g. for lack of permission) is ignored, and the line is still written. File handles passed as fd aren't rotated.

Rotation isn't coordinated between processes: if several processes log to the same file, use logrotate instead.

logrotate

The file is opened, appended to and closed for every message, never held open, so logrotate's default (rename the file and let the application create a new one) works without copytruncate, and without sending the process a SIGHUP: the next message is written to the new file.

METHODS

new

my $logger = Log::Abstraction->new(%args);
my $logger = Log::Abstraction->new(\%args);
my $logger = Log::Abstraction->new($file_path);

# Clone with optional overrides
my $clone = $logger->new(level => 'debug');

Creates a new Log::Abstraction instance, or clones an existing one when called on an object. It may also be called as a plain function, Log::Abstraction::new(%args), which behaves like Log::Abstraction->new(%args).

Arguments

  • carp_on_warn

    If set to 1, and no logger is given, call Carp::carp on warn(). Also causes error() to carp if croak_on_error is not set.

  • croak_on_error

    If set to 1, and no logger is given, call Carp::croak on error().

  • config_file

    Path to a configuration file (YAML, XML, INI, etc.) whose contents are merged with the constructor arguments. On non-Windows systems the class can also be configured via environment variables prefixed with "Log::Abstraction::". For example:

    export Log::Abstraction::script_name=foo
  • ctx

    Arbitrary context value passed through to CODE-ref logger callbacks as $args->{ctx}.

  • format

    Format string for the file, fd and scalar-path backends; a backend's own format (see "Per-backend level and format") overrides it for that backend. Unset or an empty string means the default, %level%> [%timestamp%] %class% %callstack% %message%. Tokens expanded at log time:

    %callstack%   caller file and line number
    %class%       blessed class of the logger object
    %level%       upper-cased level name
    %message%     the joined log message
    %timestamp%   the time of the call; YYYY-MM-DD HH:MM:SS local time by
                  default (see timestamp_format, timestamp_precision and utc)
    %env_FOO%     value of $ENV{FOO}, or empty string if unset

    Tokens are only expanded in the format string itself, never in the text of the message. Each line break in a message is followed by a tab, so a continuation line can't be mistaken for a new log entry.

    The special value "json" (not a format string but a magic keyword) switches all file and fd backends to emit one compact JSON object per log line:

    {"timestamp":"...","level":"info","message":"...","file":"...","line":42}

    This format is compatible with log aggregators such as journald, Loki, Elasticsearch, and Splunk. class is included when the logger is a subclass of Log::Abstraction, and fields when the call has "Structured fields". Keys are emitted in sorted order.

    Security note: because a format may contain %env_*% tokens, which expand to environment variables, avoid granting untrusted sources write access to config files that set format or any backend's format (see "Per-backend level and format").

  • level

    Minimum level at which to emit log entries. Defaults to "warning". Valid values (case-insensitive): trace, debug, info/informational, notice, warn/warning, error/err, crit/critical/fatal, alert, emerg/emergency/panic. trace and debug are the same threshold (see "LIMITATIONS"). It may also be an array reference, whose first element is used, as some configuration-file formats produce.

  • max_messages

    The most entries to keep in the in-memory history returned by "messages"; when it is full, the oldest entry is discarded. Must be a non-negative integer; 0 keeps no history at all. Unlimited by default, which in a long-running process means the history grows without bound.

  • logger

    One of:

    • A code reference -- called with a hashref { class, file, line, level, message, ctx, fields } (ctx and fields only when there are any)

    • An object -- method matching the level name is called on it

    • A hash reference -- may contain file, array, fd, syslog, journald, and/or sendmail keys, each of which may have its own level and format (see "Per-backend level and format")

    • An array reference -- { level, message } hashrefs are pushed onto it, with a fields key when the call has "Structured fields"

    • A scalar string -- treated as a file path to append to

    When not supplied, Log::Log4perl is initialised as the default backend.

    The sendmail sub-hash supports: host, port, to, from, subject, level, format, min_interval, digest, digest_max. to is required. With format, the email body is the formatted line rather than the message. level may be a level name or a syslog number (0-7); without it, every message is emailed. At most one email is sent per min_interval seconds per instance; the messages in between are dropped, unless digest is set, which sends them with the next email (see "Email digests"). If delivery fails, Carp::carp is called and the other backends still receive the message.

    The syslog sub-hash supports the keys below. The message is passed to syslog() through a %s format, so % sequences in it, such as %m, are logged literally.

    • facility -- the syslog facility (default: local0)

    • level -- only messages at this level or more severe are sent; a level name or a syslog number (0-7)

    • format -- format the message with this (see "format") before sending it; by default the message is sent as it is

    • host (or its alias server), and any other "setlogsock" in Sys::Syslog option -- passed to setlogsock()

    The journald sub-hash sends each message as a single datagram to the systemd journal using the journald native protocol. Supported keys:

    • socket -- path to the journald socket (default: /run/systemd/journal/socket)

    • identifier -- value for the SYSLOG_IDENTIFIER field (default: basename of $0)

    • level -- only messages at this level or more severe are sent; a level name or a syslog number (0-7)

    • format -- the MESSAGE field is the message formatted with this (see "format"); by default it is the message as it is

    • any other key -- included verbatim as an uppercase journald field name. The upper-cased name must contain only A-Z, 0-9 and _, and must not start with _; new() croaks otherwise.

    The PRIORITY field is set automatically from the log level (0=emerg...7=debug). A message too large for one datagram (about 200KB) is truncated and [truncated] appended. Delivery failures are silent apart from a single Carp::carp (repeated only after a later send has succeeded); the application is never crashed by a journald error.

  • redact

    Patterns to remove from every message before it is logged: a qr//, a string (compiled as a regular expression; a config file can't hold a qr//), or an array reference of them. See "Redaction".

  • rotate_interval

    Rotate log files by time: hourly, daily, weekly (weeks start on Monday) or monthly, case-insensitive. Before each write, a file whose last-modified time is in an earlier period than now (in local time, or UTC with utc) is rotated, so a file not written to for a while rotates on the next write. See "File rotation".

  • rotate_keep

    How many rotated files to keep, FILE.1 to FILE.n (default 5). With 0, a file due for rotation is deleted instead.

  • rotate_size

    Rotate log files that have reached this size: a number of bytes, optionally followed by K, M or G (powers of 1024), e.g. 10M. See "File rotation".

  • script_name

    Script name reported to syslog. Auto-detected from $0 if not supplied.

  • timestamp_format

    How %timestamp%, and the timestamp key of format => 'json', are written. Either a "strftime" in POSIX pattern (default %Y-%m-%d %H:%M:%S) or one of these names (case-insensitive):

    iso8601, rfc3339   2026-10-03T20:14:23-04:00, or 2026-10-04T00:14:23Z with utc

    The pattern may also use:

    %N        fractional seconds, 9 digits (nanoseconds)
    %3N       fractional seconds, 3 digits (milliseconds); any width 1-9
    %z        UTC offset as +hhmm (on every platform, unlike some strftimes)
    %:z       UTC offset as +hh:mm, as RFC 3339 needs
    %Z        the time-zone name; "UTC" when utc is set
    %%        a literal %

    Fractional seconds come from Time::HiRes and are truncated, not rounded; digits beyond the system clock's resolution (usually microseconds) are noise. The timestamp is taken once per message, so every backend shows the same time.

  • timestamp_precision

    The number of fractional-second digits, 0-9 (default 0), added after the seconds (each %S) of whichever timestamp_format is in use:

    Log::Abstraction->new(timestamp_format => 'rfc3339', timestamp_precision => 3, utc => 1);
    # 2026-10-04T00:14:24.094Z
  • utc

    If true, timestamps are in UTC rather than local time.

  • verbose

    When using the default Log::Log4perl backend, raises the logging level to DEBUG when set to a true value.

Returns

A blessed Log::Abstraction object.

Side Effects

Loads File::Basename if syslog is configured (either at the top level or in a logger hash) and script_name is not supplied. Loads Log::Log4perl if no backend (logger, file, fd or array) is specified.

Example

my $logger = Log::Abstraction->new(
    level  => 'debug',
    logger => \@messages,
);

my $clone = $logger->new(level => 'info');

API SPECIFICATION

Input

{
    carp_on_warn   => { type => 'boolean', optional => 1 },
    config_file    => { type => 'string',  optional => 1 },
    croak_on_error => { type => 'boolean', optional => 1 },
    ctx            => { optional => 1 },
    format         => { type => 'string',  optional => 1 },
    level          => { type => 'string',  regex => qr/^(trace|debug|info(?:rmational)?|notice|warn(?:ing)?|err(?:or)?|crit(?:ical)?|fatal|alert|emerg(?:ency)?|panic)$/i, optional => 1 },
    logger         => { optional => 1 },
    max_messages   => { type => 'integer', min => 0, optional => 1 },
    redact         => { optional => 1 },    # regex, string, or arrayref of them
    rotate_interval => { type => 'string', regex => qr/^(hourly|daily|weekly|monthly)$/i, optional => 1 },
    rotate_keep    => { type => 'integer', min => 0, optional => 1 },
    rotate_size    => { type => 'string',  regex => qr/^\s*[1-9]\d*\s*[kmg]?b?\s*$/i, optional => 1 },
    script_name    => { type => 'string',  optional => 1 },
    timestamp_format    => { type => 'string', min => 1, optional => 1 },
    timestamp_precision => { type => 'integer', min => 0, max => 9, optional => 1 },
    utc            => { type => 'boolean', optional => 1 },
    verbose        => { type => 'boolean', optional => 1 },
}

Output

{ type => 'object', class => 'Log::Abstraction' }

MESSAGES

Error                                     Meaning / Action
----------------------------------------  -----------------------------------------
"<class>: <path>: File not readable"      config_file path exists but is unreadable.
                                          Check file permissions.
"<class>: Can't load configuration       Config::Abstraction could not parse the
  from <path>"                            file.  Check syntax and format.
"<class>: syslog needs to know the        syslog backend requested but script_name
  script name"                            could not be determined.  Pass it explicitly.
"<class>: attempt to encapsulate          logger => Log::Abstraction would create
  Log::Abstraction as a logging class,    a needless forwarding loop.  Use a
  that would add a needless indirection"  different backend.
"<class>: invalid syslog level '<l>'"     level value is not a recognised syslog
                                          level name.  Use trace/debug/info/notice/
                                          warn/warning/error.
"<class>: max_messages must be a          max_messages is negative or not a number.
  non-negative integer, not '<v>'"
"<class>: rotate_size must be a           rotate_size is not, e.g., 1048576, 512K,
  positive number of bytes, optionally    10M or 1G.
  with K, M or G, not '<v>'"
"<class>: rotate_interval must be         rotate_interval is not one of those names.
  hourly, daily, weekly or monthly,
  not '<v>'"
"<class>: rotate_keep must be a           rotate_keep is negative or not a number.
  non-negative integer, not '<v>'"
"<class>: timestamp_format must be a      timestamp_format is undef, empty or a
  non-empty string"                       reference.
"<class>: timestamp_precision must be     timestamp_precision is not a whole
  an integer from 0 to 9, not '<v>'"      number of digits from 0 to 9.
"<class>: redact patterns must be         A redact entry is a reference other than
  regular expressions or non-empty        a qr//, or an empty string.
  strings"
"<class>: invalid redact pattern '<p>':   A redact string is not a valid regular
  <error>"                                expression.
"<class>: redact pattern <p> matches the  The pattern can match nothing at all
  empty string"                           (e.g. qr/x*/), which would put a marker
                                          between every character.
"<class>: invalid <backend> level '<l>'"  A backend's 'level' (file, fd, array,
                                          sendmail, journald) is neither a level
                                          name nor 0-7.  (A bad syslog 'level'
                                          gives "invalid syslog level", as above.)
"<class>: the <backend> format must be   A backend's 'format' is undef, empty or
  a non-empty string"                     a reference.
"<class>: the <backend> hash needs a      The hash form of file, fd or array has
  '<backend>' key"                        no destination (e.g. file => { level =>
                                          'info' } without a 'file' key).
"<class>: the sendmail backend needs      The sendmail sub-hash has no 'to' key.
  a 'to' address"
"<class>: sendmail digest_max must be a   digest_max is zero, negative or not a
  positive integer, not '<v>'"            number.
"<class>: invalid journald field name     An extra journald key, upper-cased, is not
  '<k>'"                                  [A-Z0-9_] or starts with '_'.

The following are not raised by new() but later, by the logging methods (trace, debug, info, notice, warn, error, fatal, critical, alert, emergency), when a message that passes the level threshold reaches the backend concerned. Croaks are configuration errors; delivery failures only carp, because a logging failure must never crash the application.

Croak                                     Meaning / Action
----------------------------------------  -----------------------------------------
"<class>: Invalid file name: <path>"      A file path (logger string, 'file' key or
                                          logger hash 'file') contains one of
                                          < > | * ? ; ! ` $ " or a control
                                          character, or contains '..'.
"<class>: Invalid SMTP host: <host>"      The sendmail 'host' contains characters
                                          other than A-Z a-z 0-9 . -
"<class>: Invalid SMTP port: <port>"      The sendmail 'port' is not an integer in
                                          1-65535.
"<class>: Don't know how to deal with     A logger hash has none of the keys file,
  the <level> message"                    array, fd, syslog, journald or sendmail.
"<class>: <object class> doesn't know     An object logger has no method for this
  how to deal with the <level> message"   level.  (notice falls back to info.)
"<class>: configuration error, no         logger is a reference of an unsupported
  handler written for the <level>         type, e.g. a SCALAR or GLOB reference.
  message"

Carp                                      Meaning / Action
----------------------------------------  -----------------------------------------
"Failed to send email: <error>"           SMTP delivery failed.  The other backends
                                          still receive the message.
"<class>: syslog failed: <error>"         Sys::Syslog::syslog() died.
"<class>: journald send failed: <error>"  The journald socket could not be reached.
                                          Given once, then not again until a send
                                          succeeds.

PSEUDOCODE

FUNCTION new(class_or_obj, args...)

  Parse args:
    IF single non-hash scalar
    THEN store as logger shorthand
    ELSE extract named params via Params::Get

  IF config_file present:
    CROAK if file is not readable
    Load via Config::Abstraction, merge into args (constructor args win)
    Restore caller-supplied array ref that config merge would have dropped

  IF called on a blessed instance (clone form):
    CROAK on an invalid timestamp_format or timestamp_precision
    CROAK on an invalid rotate_size, rotate_interval or rotate_keep, and
      normalise rotate_size to bytes
    CROAK on an invalid redact pattern; compile redact to one regex
    shallow-clone self merged with override args
    validate and store new level integer if level given in args
    copy message history list
    start the clone with no held email digest
    count the clone as a user of an open syslog connection
    RETURN clone

  IF syslog requested (top level or in a logger hash) and script_name
  not supplied:
    auto-detect script name via File::Basename
    CROAK if still undefined

  IF logger arg is a Log::Abstraction object:
    CROAK (would create a needless forwarding loop)

  IF no logger AND no file AND no fd AND no array:
    load Log::Log4perl, easy_init at DEBUG or ERROR per verbose flag
    store Log4perl logger as the backend

  Normalise and validate level:
    IF level is an arrayref, take first element
    lc() the level string
    CROAK if not in syslog_values lookup
    default to $DEFAULT_LEVEL if not supplied

  CROAK if max_messages is given and is not a non-negative integer
  CROAK if timestamp_format is empty or not a string, or
    timestamp_precision is not an integer 0-9
  CROAK if rotate_size is not a positive size, rotate_interval is not
    hourly/daily/weekly/monthly, or rotate_keep is not a non-negative
    integer; normalise rotate_size to bytes
  CROAK if a redact pattern is not a regex or non-empty string, doesn't
    compile, or matches the empty string; compile redact to one regex

  FOR each backend (top-level file/fd/array, and the logger hash's
  file/fd/array/syslog/sendmail/journald) given as a hash:
    CROAK if a file/fd/array hash lacks its destination key (its own name)
    CROAK if its 'level' is not a level name or 0-7
    CROAK if its 'format' is undef, empty or not a string

  IF logger is a hash:
    CROAK if a sendmail sub-hash has no 'to' address, or a digest_max
      that isn't a positive integer
    CROAK if an extra journald key is not a valid journald field name

  RETURN bless { messages => [], merged args, level => numeric } as class

END FUNCTION

level

my $current = $logger->level();
$logger->level('debug');

Get or set the minimum logging level. When setting, returns $self to allow method chaining. When getting, returns the current level as an integer (per the syslog numeric scale; lower numbers are higher priority).

Arguments

  • $level (optional)

    A level name string: trace, debug, info, notice, warn/warning, or error. Case-insensitive. Omit to perform a pure get.

Returns

In getter mode: an integer in the range 0 (emergency) to 7 (debug/trace).

In setter mode: $self (to allow chaining), or undef, after a Carp::carp, if the level name is not recognised; the level is then unchanged. A false argument (undef, '' or 0) is a get, not a set, so levels are set by name.

Side Effects

When setting, updates $self->{level}.

Example

$logger->level('debug');
my $n = $logger->level();   # e.g. 7

# Method chaining
$logger->level('info')->info('Now at info level');

API SPECIFICATION

Input

{
    level => { type => 'string', regex => qr/^(trace|debug|info(?:rmational)?|notice|warn(?:ing)?|err(?:or)?|crit(?:ical)?|fatal|alert|emerg(?:ency)?|panic)$/i, optional => 1 },
}

Output

Getter: { type => 'integer', min => 0, max => 7 }
Setter: { type => 'object', class => 'Log::Abstraction' }

MESSAGES

Warning                                   Meaning / Action
----------------------------------------  ------------------------------------------
"<class>: invalid syslog level '<l>'"     The supplied level name is not recognised.
                                          Use trace/debug/info/notice/warn/error.

PSEUDOCODE

FUNCTION level(self, level?)

  IF level argument supplied:
    CARP and RETURN undef if level is not a recognised syslog name
    Store syslog_values{level} in self->{'level'}
    RETURN self  (allows method chaining)

  ELSE (getter mode):
    RETURN self->{'level'}  (current numeric threshold)

END FUNCTION

Level detection methods

is_trace
is_debug
is_info
is_notice
is_warn
is_error
is_critical
is_alert
is_emergency
if($logger->is_debug()) { ... }

Each returns a true value when a message logged with the method of the same name (is_warn for warn()) would pass the logger's level threshold, so that expensive message-building can be skipped. They follow the current threshold, including changes made with "level". As with the levels themselves, is_trace equals is_debug. Provided for compatibility with Log::Any.

Arguments

None.

Returns

1 if messages at that level would be emitted; 0 otherwise.

Example

if($logger->is_debug()) {
    $logger->debug('Expensive diagnostic: ' . Dumper(\%state));
}

$logger->level('warning');
$logger->is_warn();    # 1
$logger->is_info();    # 0

API SPECIFICATION

Input

{} (no arguments)

Output

{ type => 'boolean' }

messages

my $aref = $logger->messages();

Returns a reference to a shallow copy of all messages emitted through this logger since it was created (or since the last clone).

Arguments

None.

Returns

An array reference of hashrefs, each with keys level (string) and message (string), and fields (hashref) when the message was logged with "Structured fields".

Side Effects

None. The returned array is a copy; modifying it does not affect the internal history.

Example

$logger->info('hello');
my $msgs = $logger->messages();
# $msgs->[0] = { level => 'info', message => 'hello' }

API SPECIFICATION

Input

{} (no arguments)

Output

{ type => 'arrayref', element_type => { level => 'string', message => 'string', fields => 'hashref?' } }

flush

$logger->flush();

Sends, now, the messages that a sendmail backend with digest is holding back because of min_interval (see "Email digests"), whether or not the interval has passed. Does nothing if none are held. Called automatically when the logger is destroyed.

Arguments

None.

Returns

The logger, for method chaining.

Side Effects

May send an email, which starts the min_interval interval again. A delivery failure is carped and the messages stay held. Croaks if the sendmail host or port is invalid.

Example

$logger->error('disk full');    # emailed
$logger->error('disk still full');    # held: within min_interval
$logger->flush();                # emailed now

API SPECIFICATION

Input

{} (no arguments)

Output

{ type => 'object', class => 'Log::Abstraction' }

trace

$logger->trace(@messages);
$logger->trace(\@messages);

Logs a message at trace level. syslog has no priority below debug, so trace shares debug's threshold: trace messages are emitted whenever debug messages are, and are sent to syslog and journald as debug. The message is dropped silently when the configured level is above debug.

Arguments

  • @messages

    One or more strings, or a single array reference. All elements are joined without a separator before storage. May be followed by a hashref of "Structured fields".

Returns

$self, to allow method chaining.

Side Effects

Appends to the internal message history and dispatches to configured backends.

Example

$logger->trace('entering sub foo, args=', join(',', @args));

# Chaining
$logger->trace('start')->debug('details')->info('summary');

API SPECIFICATION

Input

{ messages => { type => [ 'arrayref', 'scalar' ] } }

Output

{ type => 'object', class => 'Log::Abstraction' }

MESSAGES

Croaks if the configured backend is misconfigured, and carps if delivery fails; see the second table under "new"'s MESSAGES.

debug

$logger->debug(@messages);
$logger->debug(\@messages);

Logs a message at debug level.

Arguments

  • @messages

    One or more strings, or a single array reference, optionally followed by a hashref of "Structured fields".

Returns

$self, to allow method chaining.

Side Effects

Appends to the internal message history and dispatches to configured backends.

Example

$logger->debug('Query took ', $elapsed, 'ms');

API SPECIFICATION

Input

{ messages => { type => [ 'arrayref', 'scalar' ] } }

Output

{ type => 'object', class => 'Log::Abstraction' }

MESSAGES

Croaks if the configured backend is misconfigured, and carps if delivery fails; see the second table under "new"'s MESSAGES.

info

$logger->info(@messages);
$logger->info(\@messages);

Logs a message at info level.

Arguments

  • @messages

    One or more strings, or a single array reference, optionally followed by a hashref of "Structured fields".

Returns

$self, to allow method chaining.

Side Effects

Appends to the internal message history and dispatches to configured backends.

Example

$logger->info('Server started on port ', $port);

API SPECIFICATION

Input

{ messages => { type => [ 'arrayref', 'scalar' ] } }

Output

{ type => 'object', class => 'Log::Abstraction' }

MESSAGES

Croaks if the configured backend is misconfigured, and carps if delivery fails; see the second table under "new"'s MESSAGES.

notice

$logger->notice(@messages);
$logger->notice(\@messages);

Logs a message at notice level (higher priority than info, lower than warn).

Arguments

  • @messages

    One or more strings, or a single array reference, optionally followed by a hashref of "Structured fields".

Returns

$self, to allow method chaining.

Side Effects

Appends to the internal message history and dispatches to configured backends.

Example

$logger->notice('Configuration reloaded');

API SPECIFICATION

Input

{ messages => { type => [ 'arrayref', 'scalar' ] } }

Output

{ type => 'object', class => 'Log::Abstraction' }

MESSAGES

Croaks if the configured backend is misconfigured, and carps if delivery fails; see the second table under "new"'s MESSAGES.

warn

$logger->warn(@messages);
$logger->warn(\@messages);
$logger->warn(warning => $text);
$logger->warn({ warning => $text });
$logger->warn(warning => \@parts);
$logger->warn($text, \%fields);

Logs a warning message. Also dispatches to syslog and/or email backends when those are configured. Falls back to Carp::carp when no backend (logger, array, file or fd) is set. The Carp::carp (whether from carp_on_warn or the fallback) only happens when the message passes the level threshold.

Called as a class method (Log::Abstraction->warn(...), or on a subclass), it calls Carp::carp directly.

A warn() call with an empty or all-undef argument list is a silent no-op.

Arguments

  • @messages

    A plain list of strings joined without separator, or a named warning parameter whose value may be a string or an array reference of strings. Either form may be followed by a hashref of "Structured fields", e.g. warn('Slow query', { ms => 1250 }).

Returns

$self, to allow method chaining.

Side Effects

Appends to internal message history. Writes to all configured backends. May call Carp::carp if carp_on_warn is set or no backend is active.

Example

$logger->warn('Disk usage is high');
$logger->warn(warning => 'Connection reset', ' retrying');
$logger->warn({ warning => ['Part A', 'Part B'] });

API SPECIFICATION

Input

# Named form
{ warning => { type => [ 'scalar', 'arrayref' ] } }
# Plain-list form
{ messages => { type => 'arrayref' } }

Output

{ type => 'object', class => 'Log::Abstraction' }

MESSAGES

(the warning text itself)                 Carped if carp_on_warn is set, or if no
                                          backend (logger, array, file or fd) is
                                          configured, provided the warning passes
                                          the level threshold.  Also carped when
                                          called as a class method.

Backend misconfiguration and delivery failures are reported as described in the second table under "new"'s MESSAGES.

error

$logger->error(@messages);
$logger->error(warning => $text);
$logger->error($text, \%fields);

Logs an error-level message. Behaves identically to warn() but at the error level, which triggers Carp::croak if croak_on_error is set or no backend (logger, array, file or fd) is set. Called as a class method, it calls Carp::croak directly.

Arguments

Same argument forms as warn().

Returns

$self, to allow method chaining. Note: if croak_on_error is set, the method never returns -- execution unwinds via Carp::croak.

Side Effects

Same as warn() plus optional Carp::croak escalation.

Example

$logger->error('Fatal: database unavailable');

API SPECIFICATION

Input

{ warning => { type => [ 'scalar', 'arrayref' ], optional => 1 } }

Output

{ type => 'object', class => 'Log::Abstraction' }

MESSAGES

Croak                                     Meaning / Action
----------------------------------------  ------------------------------------------
(the error message text itself)           croak_on_error is set, or no backend
                                          (logger, array, file or fd) is
                                          configured, or error() was called as a
                                          class method.  The call stack is unwound.
(the error message text itself), as a     carp_on_warn is set and croak_on_error
  carp                                    is not.

Backend misconfiguration and delivery failures are reported as described in the second table under "new"'s MESSAGES.

fatal

$logger->fatal(@messages);

Synonym for error(). Provided for compatibility with logging frameworks that use fatal as the highest-severity level name.

Arguments

Same as error().

Returns

$self.

Side Effects

Same as error().

Example

$logger->fatal('Unrecoverable state; aborting');

API SPECIFICATION

Input

{ warning => { type => [ 'scalar', 'arrayref' ], optional => 1 } }

Output

{ type => 'object', class => 'Log::Abstraction' }

MESSAGES

Same as error().

Methods above error

critical
alert
emergency
$logger->critical(@messages);
$logger->alert(warning => $text);
$logger->emergency($text, \%fields);

Log a message at a level more severe than error:

Method      Level       syslog   Priority
----------  ----------  -------  --------
critical    critical    crit     2
alert       alert       alert    1
emergency   emergency   emerg    0

Arguments

critical, alert and emergency take the same argument forms as warn().

Returns

$self, to allow method chaining (unless they croak; see below).

Side Effects

These behave like error(), at a more severe level: croak_on_error, or having no backend, makes them Carp::croak, and carp_on_warn makes them Carp::carp. The level string passed to backends is the method name (critical, alert or emergency, upper-cased in text formats); syslog gets crit, alert or emerg, and journald PRIORITY 2, 1 or 0. An object logger without the method (such as Log::Log4perl) is called with fatal, or error if it has no fatal either.

Example

$logger->critical('Disk 95% full', { mount => '/var' });
$logger->alert('Primary database unreachable');
$logger->emergency('Data corruption detected; shutting down');

API SPECIFICATION

Input

{ warning => { type => [ 'scalar', 'arrayref' ], optional => 1 } }

Output

{ type => 'object', class => 'Log::Abstraction' }

MESSAGES

Same as error().

EXAMPLES

CSV file logging for BI import

The code-reference backend gives you full control over the output format. The example below writes every message at trace level and above as a CSV row to a file, producing output that can be loaded directly into a spreadsheet or BI tool (Tableau, Power BI, Metabase, etc.).

Each row contains: timestamp, level, class, file, line, message.

use Log::Abstraction;

my $csv_file = 'app_events.csv';

# Write the header row once (skip if the file already exists and has data).
unless (-s $csv_file) {
    open my $fh, '>', $csv_file or die "Cannot open $csv_file: $!";
    print $fh qq{timestamp,level,class,file,line,message\n};
    close $fh;
}

# Helper: quote a single CSV field (escapes embedded double-quotes).
my $csv_field = sub {
    my $v = defined $_[0] ? $_[0] : '';
    $v =~ s/"/""/g;
    return qq{"$v"};
};

my $logger = Log::Abstraction->new(
    level  => 'trace',        # capture everything from trace upwards
    logger => sub {
        my $args = $_[0];

        my $timestamp = POSIX::strftime('%Y-%m-%dT%H:%M:%SZ', gmtime);
        my $message  = join(' ', @{ $args->{message} // [] });

        open my $fh, '>>', $csv_file or return;
        print $fh join(',',
            $csv_field->($timestamp),
            $csv_field->($args->{level}),
            $csv_field->($args->{class}),
            $csv_field->($args->{file}),
            $csv_field->($args->{line}),
            $csv_field->($message),
        ), "\n";
        close $fh;
    },
);

$logger->trace('application started');
$logger->info('user logged in', { user => 'alice' });
$logger->warn({ warning => 'disk usage above 80%' });

The resulting app_events.csv looks like:

timestamp,level,class,file,line,message
"2026-05-27T14:00:00Z","trace","Log::Abstraction","app.pl","42","application started"
"2026-05-27T14:00:01Z","info","Log::Abstraction","app.pl","43","user logged in"
"2026-05-27T14:00:02Z","warn","Log::Abstraction","Log/Abstraction.pm","820","disk usage above 80%"

Note: class is always Log::Abstraction (or the subclass name if you subclass the module). For trace, debug, info, and notice calls, file and line resolve to the caller's source location. For warn and error calls the extra _high_priority stack frame shifts the resolution one level inward, so file and line point into the module rather than the calling script.

For production use, consider replacing the manual $csv_field quoting with Text::CSV for correct handling of embedded newlines and other edge cases.

If you also want real-time alerting on critical events, add the email logic directly inside the code-ref callback -- test $args->{level} and call your mailer for warn / error messages while still writing the CSV row for every message.

Alternatively, use the sendmail hash-ref backend on its own (without the code-ref) and add a level key to restrict emails to warn-and-above:

my $logger = Log::Abstraction->new(
    level  => 'warn',
    logger => {
        sendmail => {
            host         => 'smtp.example.com',
            to           => 'ops@example.com',
            from         => 'logger@example.com',
            subject      => 'Application alert',
            level        => 'warn',   # only email at warn level and above
            min_interval => 300,      # at most one alert email per 5 minutes
        },
    },
);

Note: the sendmail backend writes the module's standard text format, not CSV. To produce CSV rows and send email alerts from the same logger, embed both the CSV-write and the mail-send logic inside a single code-ref callback as described above.

LIMITATIONS

Syslog hash mutation

The syslog sub-hash passed to new() is mutated in-place on the first log call: facility, level and format are temporarily removed before setlogsock() is called, then restored; server is permanently renamed to host. Sharing a syslog hashref between two Log::Abstraction instances is not supported and produces undefined behaviour on the second instance.

trace is the same threshold as debug

syslog has no priority below debug, so trace and debug share one threshold: a logger at debug level also emits trace messages, and trace can't be filtered separately.

Unbounded message history by default

Every logged message is kept in the history returned by "messages". In a long-running process (a daemon, or under mod_perl) set max_messages to stop it growing without bound.

syslog connection is shared

openlog() and closelog() act on the whole process, so every instance logging to syslog shares one connection, opened with the script_name of the first. It is closed when the last such instance is destroyed.

Structured fields are text in most backends

Only the history, array, CODE-ref, JSON and journald backends keep "Structured fields" as data. Text formats, syslog, email and object loggers get them as key=value text appended to the message, and a custom format has no token for them on their own.

Single-threaded email throttle

The min_interval throttle and digest for the sendmail backend and the _syslog_opened first-open flag are stored on the object without mutex protection. Under Perl ithreads or other concurrency models, objects shared between threads are not safe.

OpenTelemetry not yet supported

The OTel Logs SDK for Perl is incomplete; see the TODO block at the top of lib/Log/Abstraction.pm for a full status report and the list of blockers. Monitor https://metacpan.org/pod/OpenTelemetry::SDK for progress.

Log::Log4perl is a de-facto required dependency

When no logger, file, fd or array backend is configured, new() loads Log::Log4perl and uses it as the default backend, so it is a required dependency even for applications that never use it.

AUTHOR

Nigel Horne njh@nigelhorne.com

SEE ALSO

SUPPORT

This module is provided as-is without any warranty.

Please report any bugs or feature requests to bug-log-abstraction at rt.cpan.org, or through the web interface at http://rt.cpan.org/NoAuth/ReportBug.html?Queue=Log-Abstraction. I will be notified, and then you'll automatically be notified of progress on your bug as I make changes.

You can find documentation for this module with the perldoc command.

perldoc Log::Abstraction

You can also look for information at:

FORMAL SPECIFICATION

new

FIELDS == STRING ⇸ VALUE          structured fields (see Structured fields)
ENTRY  == { level : STRING; message : STRING; fields : FIELDS }

entry(l, m, f) == {level ↦ l, message ↦ ρ(m)} ∪ (if f = ∅ then ∅ else {fields ↦ ρ(f)})

ρ(x) == if redact = ∅ then x
        else x with each match of redact replaced by '[REDACTED]'
             (in strings, recursively in plain hashes and arrays)

┌─ LogState ──────────────────────────────────────────────────
│ level        : ℤ
│ messages     : seq ENTRY
│ max_messages : ℕ ∪ {∞}
│ logger       : LOGGER
│ redact       : REGEX ∪ {∅}
├─────────────────────────────────────────────────────────────
│ 0 ≤ level ≤ 7
│ #messages ≤ max_messages
└─────────────────────────────────────────────────────────────

┌─ New ───────────────────────────────────────────────────────
│ args? : Args
│ result! : LogState
├─────────────────────────────────────────────────────────────
│ result!.level = syslog_values(args?.level ∨ 'warning')
│ result!.messages = ⟨⟩
│ result!.max_messages = args?.max_messages ∨ ∞
│ result!.redact = ⋃ args?.redact   {one regex matching any; ∅ if none}
│ args?.redact ≠ ∅ ⟹ ¬('' ∈ L(result!.redact))
│ args?.logger ≠ ∅ ⟹ result!.logger = args?.logger
│ args?.logger = ∅ ∧ args?.file = ∅ ∧ args?.fd = ∅ ∧ args?.array = ∅
│   ⟹ result!.logger = Log4perl
└─────────────────────────────────────────────────────────────

Clone operation (called on an existing object):

┌─ Clone ─────────────────────────────────────────────────────
│ ΔLogState
│ overrides? : Args
├─────────────────────────────────────────────────────────────
│ result!.level    = syslog_values(overrides?.level ∨ level)
│ result!.messages = messages   {new sequence; entries shared}
│ result!.logger   = overrides?.logger ∨ logger
└─────────────────────────────────────────────────────────────

level

┌─ LevelGet ─────────────────────────────────────────────────
│ ΞLogState
│ result! : ℤ
├─────────────────────────────────────────────────────────────
│ result! = level
│ 0 ≤ result! ∧ result! ≤ 7
└─────────────────────────────────────────────────────────────

┌─ LevelSet ─────────────────────────────────────────────────
│ ΔLogState
│ new_level? : STRING
├─────────────────────────────────────────────────────────────
│ new_level? ∈ dom(syslog_values)
│ level' = syslog_values(new_level?)
└─────────────────────────────────────────────────────────────

┌─ LevelSetInvalid ──────────────────────────────────────────
│ ΞLogState
│ new_level? : STRING
│ result! : undef
├─────────────────────────────────────────────────────────────
│ new_level? ≠ ''
│ new_level? ¬in; dom(syslog_values)
│ carp("invalid syslog level")
└─────────────────────────────────────────────────────────────

level(new_level?) ≡ LevelSet ∨ LevelSetInvalid

is_trace, is_debug, is_info, is_notice, is_warn, is_error, is_critical, is_alert, is_emergency

┌─ IsLevel ──────────────────────────────────────────────────
│ ΞLogState
│ lvl? : LEVEL
│ result! : BOOLEAN
├─────────────────────────────────────────────────────────────
│ result! = (level ≥ syslog_values(lvl?))
└─────────────────────────────────────────────────────────────

is_<lvl> ≡ IsLevel[lvl? := lvl]

messages

┌─ Messages ─────────────────────────────────────────────────
│ ΞLogState
│ result! : seq ENTRY
├─────────────────────────────────────────────────────────────
│ result! = messages
└─────────────────────────────────────────────────────────────

flush

┌─ Flush ────────────────────────────────────────────────────
│ ΔLogState
│ result! : LogState
├─────────────────────────────────────────────────────────────
│ digest ≠ ⟨⟩ ∧ sent(digest) ⟹ digest' = ⟨⟩ ∧ last_email_sent' = now
│ digest ≠ ⟨⟩ ∧ ¬sent(digest) ⟹ digest' = digest
│ digest = ⟨⟩ ⟹ digest' = digest
│ messages' = messages
│ result! = self
└─────────────────────────────────────────────────────────────

trace

┌─ Trace ────────────────────────────────────────────────────
│ ΔLogState
│ msg? : seq STRING
│ fields? : FIELDS
├─────────────────────────────────────────────────────────────
│ syslog_values('trace') ≤ level
│ messages' = messages ⌢ ⟨entry('trace', ⊕(msg?), fields?)⟩
└─────────────────────────────────────────────────────────────

debug

┌─ Debug ────────────────────────────────────────────────────
│ ΔLogState
│ msg? : seq STRING
│ fields? : FIELDS
├─────────────────────────────────────────────────────────────
│ syslog_values('debug') ≤ level
│ messages' = messages ⌢ ⟨entry('debug', ⊕(msg?), fields?)⟩
└─────────────────────────────────────────────────────────────

info

┌─ Info ─────────────────────────────────────────────────────
│ ΔLogState
│ msg? : seq STRING
│ fields? : FIELDS
├─────────────────────────────────────────────────────────────
│ syslog_values('info') ≤ level
│ messages' = messages ⌢ ⟨entry('info', ⊕(msg?), fields?)⟩
└─────────────────────────────────────────────────────────────

notice

┌─ Notice ───────────────────────────────────────────────────
│ ΔLogState
│ msg? : seq STRING
│ fields? : FIELDS
├─────────────────────────────────────────────────────────────
│ syslog_values('notice') ≤ level
│ messages' = messages ⌢ ⟨entry('notice', ⊕(msg?), fields?)⟩
└─────────────────────────────────────────────────────────────

warn

┌─ Warn ─────────────────────────────────────────────────────
│ ΔLogState
│ msg? : seq STRING | { warning : STRING | seq STRING }
│ fields? : FIELDS
├─────────────────────────────────────────────────────────────
│ msg? ≠ ∅ ∧ join(msg?) ≠ ''
│ syslog_values('warn') ≤ level
│ messages' = messages ⌢ ⟨entry('warn', join(msg?), fields?)⟩
│ (carp_on_warn ∨ no_backend) ⟹ carp(join(msg?))
└─────────────────────────────────────────────────────────────

no_backend ≡ logger = ∅ ∧ array = ∅ ∧ file = ∅ ∧ fd = ∅

Called as a class method (no LogState): carp(join(msg?)), and
messages is not touched.

error

┌─ Error ────────────────────────────────────────────────────
│ ΔLogState
│ msg? : seq STRING | { warning : STRING | seq STRING }
│ fields? : FIELDS
├─────────────────────────────────────────────────────────────
│ msg? ≠ ∅ ∧ join(msg?) ≠ ''
│ syslog_values('error') ≤ level
│ messages' = messages ⌢ ⟨entry('error', join(msg?), fields?)⟩
│ (croak_on_error ∨ no_backend) ⟹ execution_continues = false
└─────────────────────────────────────────────────────────────

Called as a class method (no LogState): croak(join(msg?)).

fatal

fatal ≡ error   (identical operation schema)

critical, alert, emergency

The Error schema, with 'error' replaced by 'critical', 'alert' or
'emergency' respectively.

In every logging schema, when #messages' would exceed max_messages
the oldest entries are dropped: messages' = the last max_messages
entries.  fields? is a hashref given after the message (see
Structured fields); fields? = ∅ when there is none.

COPYRIGHT AND LICENSE

Copyright (C) 2025-2026 Nigel Horne

Usage is subject to the GPL2 licence terms. If you use it, please let me know.