The London Perl and Raku Workshop takes place on 26th Oct 2024. If your company depends on Perl, please consider sponsoring and/or attending.


Carp::Datum::Cfg - Dynamic Debug Configuration Setting for Datum


 # In application's main
 use Carp::Datum qw(:all on);      # turns Datum "on" or "off"
 DLOAD_CONFIG(-file => "./", -config => "config string");


By using the DLOAD_CONFIG function in an application's main file, a debugging configuration can be dynamically loaded to define a particular level of debug/trace flags for a specific sub-part of code.

For instance, the tracing can be turned off when entering a routine of a designated package. That is very useful for concentrating the debugging onto the area that is presently developed and/or to filter some verbose parts of code (recursive function call), when they don't need to be monitored to fix the problem.


Before the obscure explaination of the grammar, here is an example of what can be specified by dynamic configuration:

   * flags definition: macro that can be used in further configuration
   * settings
  flags common {
      trace(yes): all;

  flags silent {

   * default setting to use when there is no specific setting 
   * for the area
  default common;

   * specific settings for specific areas
  routine "context", "cleanup"                 { use silent; }
  routine "validate", "is_num", "is_greater"   { use silent; }

  file ""                         { use silent; }
  file "" {
      use silent; 
      trace(yes): emergency, alert, critical;

  cluster "CGI::MxScreen" {
      use silent; 

   * aliasing to reduce the trace output line length

  alias "/home/dehaudtc/usr/perl/lib/site_perl/5.6.0/CGI" => "<PM>";


The only user interface is the DLOAD_CONFIG routine, which expects the following optional named parameters:

-config => string

Give an inlined configuration string that is appended to the one defined by -file, if any.

-file => filename

Specifies the configuration file to load to initialize the debugging and tracing flags to be used for this run.


Main Configuration Directives

The following main directives can appear at a nesting level of 0. The syntax unit known as BLOCK is a list of semi-colon terminated directives held within curly braces.

alias large_path => short_path

Defines an alias to be used during tracing. The large_path string is replaced by the short_path in the logs.

For instance, given:

  alias "/home/dehaudtc/lib/CGI" => "<CGI>";

then a trace for file /home/dehaudtc/lib/CGI/ would be traced as coming from file <CGI>/, which is nicer to read.

cluster name1, name2 BLOCK

The BLOCK defines the flags to be applied to all named clusters. A cluster is a set of classes under a given name scope. Cluster names are given by strings within double quotes, as in:

    cluster "CGI::MxScreen", "Net::MsgLink" { use silent; }

This would apply to all classes under the "CGI::MxScreen" or "Net::MsgLink" name scopes, i.e. CGI::MxScreen::Screen would be affected.

An exact match is attempted first, i.e. saying:

    cluster "CGI::MxScreen"         { use verbose; }
    cluster "CGI::MxScreen::Screen" { use silent; }

would apply the silent flags for CGI::MxScreen::Screen but the verbose ones to CGI::MxScreen::Tie::Stdout.

default name|BLOCK.

Specifies the default flags that should apply. The default flags can be given by providing the name of flags, defined by the flags directive, or by expansing them in the following BLOCK.

For instance:

    default silent;

would say that the flags to apply by default are the ones defined by an earlier flags silent directive. Not expanding defaults allows for quick switching by replacing silent with verbose. It is up to the module user to define what is meant by that though.

file name1, name2 BLOCK

The BLOCK defines the flags to be applied to all named files. File names are given by strings withing double quotes, as in:

    file "", "" { use silent; }

This would apply to all files named "" or "", whatever their directory, i.e. it would apply to /tmp/ as well as ../

An exact match is attempted first, i.e. saying:

    file ""      { use verbose; }
    file "/tmp/" { use silent; }

would apply the silent flags for /tmp/ but the verbose ones to ./

flags name BLOCK

Define a symbol name whose flags are described by the following BLOCK. This name can then be used in default and use directives.

For instance:

    flags common {
        trace(yes): all;

would define the flags known as common, which can then be re-used, as in:

    flags other {
        use common;         # reuses definiton of common flags
        panic(no);          # but switches off panic, enabled in common

A flag symbol must be defined prior being used.

routine name1, name2 BLOCK

The BLOCK defines the flags to be applied to all named routines. Routine names are given by strings within double quotes, as in:

    routine "foo", "bar" { use silent; }

This would apply to all routines named "foo" or "bar", whatever their package, for instance main::foo and x::bar.

Debugging and Tracing Flags

Debugging (and tracing) flags can be specified only within syntactic BLOCK items, as expected by main directives such as flags or file.

Following is a list of debugging flags that can be specified in the configuration. The order in which they are given in the file is significant: the yes/no settings are applied sequentially.

use name

Uses flags defined by a flags directive under name. It acts as a recursive macro expansion (since use can also be specified in flags). The symbol name must have been defined earlier.


Whether to print out the entering/exiting of routines. That implies the invocation of the DFEATURE function in the routines.


Whether to print out the returned when using the return DVAL and DARY routines.


Whether to print out traces specified by the DTRACE function. By default all trace levels are affected. It may be followed by a list of trace levels affected by the directive, as in.

    trace(yes): emergency, alert, critical;

Trace levels are purely conventional, and have a strict one-to-one mapping with DTM_TRC_ levels given at the DTRACE call. They are further described in "Trace Levels" below. There is one bit per defined trace level, contrary to the convention established by syslog(), for better tuning.


Whether to evaluate the pre-condition given by DREQUIRE. But see "Assertion Evaluation Note" below.


Whether to evaluate the assertion given by DASSERT. But see "Assertion Evaluation Note" below.


Whether to evaluate the post-condition given by DENSURE. But see "Assertion Evaluation Note" below.


Whether to panic upon an assertion failure (pre/post condition or assertion). If not enabled, a simple warning is issued, tracing the assertion failure.


Whether to print out a stack trace upon assertion failure.


Enable or disables all the previously described items.

Assertion Evaluation Note

When Carp::Datum is switched off, the assertions are always monitored, and any failure is fatal. This is because a failing assertion is a Bad Thing in production mode. Also, since DREQUIRE and friends are not C macros but routines, the assertion expression is evaluated anyway, so it might as well be tested.

Therefore, a directive like:


will only turn off monitoring of pre-conditions in debugging mode (e.g. because the interface is not finalized, or the clients do not behave properly yet).

Trace Levels

Here is the list of trace flags that can be specified by the configuration:

    Configuration    DTRACE flag
    -------------    -------------
              all    TRC_ALL
        emergency    TRC_EMERGENCY
            alert    TRC_ALERT
         critical    TRC_CRITICAL
            error    TRC_ERROR
          warning    TRC_WARNING
           notice    TRC_NOTICE
             info    TRC_INFO
            debug    TRC_DEBUG

A user could say something like:

    trace(no): all;
    trace(yes): emergency, alert, critical, error;

Since flags are applied in sequence, the first directive turns all tracing flags to off, the second enables only the listed ones.


Some things are not fully documented.


Christophe Dehaudt and Raphael Manfredi are the original authors.

Send bug reports, hints, tips, suggestions to Dave Hoover at <>.



1 POD Error

The following errors were encountered while parsing the POD:

Around line 605:

You forgot a '=back' before '=head2'