Data::Sah::Coerce - Coercion rules for Data::Sah


This document describes version 0.033 of Data::Sah::Coerce (from Perl distribution Data-Sah-Coerce), released on 2019-01-26.


 use Data::Sah::Coerce qw(gen_coercer);

 # a utility routine: gen_coercer
 my $c = gen_coercer(
     type               => 'date',
     coerce_to          => 'DateTime',
     coerce_rules       => ['str_alami'],  # explicitly enable a rule, etc
     # return_type      => 'str+val',      # default is 'val'

 my $val = $c->(123);          # unchanged, 123
 my $val = $c->(1463307881);   # becomes a DateTime object
 my $val = $c->("2016-05-15"); # becomes a DateTime object
 my $val = $c->("2016foo");    # unchanged, "2016foo"


This distribution contains a standard set of coercion rules for Data::Sah. It is separated from the Data-Sah distribution and can be used independently.

A coercion rule is put in Data::Sah::Coerce::$COMPILER::$TARGET_TYPE::$SOURCE_TYPE_AND_EXTRA_DESCRIPTION module, for example: Data::Sah::Coerce::perl::date::float_epoch for converting date from integer (Unix epoch) or Data::Sah::Coerce::perl::date::str_iso8601 for converting date from ISO8601 strings like "2016-05-15".

Basically, a coercion rule will provide an expression (expr_match) that evaluates to true when data can be coerced, and an expression (expr_coerce) to actually coerce/convert data to the target type. This rule can be combined with other rules to form the final coercion code.

The module must contain meta subroutine which must return a hashref that has the following keys (* marks that the key is required):

  • v* => int (default: 1)

    Metadata specification version. Currently at 4.

    History: bumped from 3 to 4 to remove enable_by_default property. Now the list of standard (enabled-by-default) coercion rules is maintained in Data::Sah::Coerce itself. This allows us to skip scanning all Data::Sah::Coerce::* coercion modules installed on the system. Data::Sah::Coerce still accepts version 3; it just ignores the enable_by_default property.

    History: bumped from 2 to 3 to allow coercion expression to return error message explaining why coercion fails. The might_die metadata property is replaced with might_fail. When might_fail is set to true, expr_coerce must return array containing error message and coerced data, instead of just coerced data.

    History: Bumped from 1 to 2 to exclude old module names.

  • might_fail => bool (default: 0)

    Whether coercion might fail, e.g. because of invalid input. If set to 1, expr_coerce key that the coerce() routine returns must be an expression that returns an array (envelope) of (error_msg, data) instead of just coerced data. Error message should be a string that is set when coercion fails and explains why. Otherwise, if coercion succeeds, the string should be set to undefined value.

    An example of a rule like this is coercing from string in the form of "YYYY-MM-DD" to a DateTime object. The rule might match any string in the form of /\A(\d{4})-(\d{2})-(\d{2})\z/ while it might not be a valid date.

  • prio => int (0-100, default: 50)

    This is to regulate the ordering of rules. The higher the number, the lower the priority (meaning the rule will be put further back). Rules that are computationally more expensive and/or match more broadly in general should be put further back (lower priority, higher number).

  • precludes => array of (str|re)

    List the other rules or rule patterns that are precluded by this rule. Rules that are mutually exclusive or pure alternatives to one another (e.g. date coercien rules str_alami vs str_alami both parse natural language date string; there is usually little to none of usefulness in using both; besides, both rules match all string and dies when failing to parse the string. So in str_natural rule, you'll find this metadata:

     precludes => [qr/\Astr_alami(_.+)?\z/]

    and in str_alami rule you'll find this metadata:

     precludes => [qr/\Astr_alami(_.+)?\z/, 'str_natural']

    Note that the str_alami rule also precludes other str_alami_* rules (like str_alami_en and str_alami_id).

    Also note that rules which are specifically requested to be used (e.g. using x.perl.coerce_rules attribute in Sah schema) will still be precluded.

The module must also contain coerce subroutine which must generate the code for coercion. The subroutine must accept a hash of arguments (* indicates required arguments):

  • data_term => str

  • coerce_to => str

    Some Sah types are "abstract" and can be represented using a choice of several actual types in the target programming language. For example, "date" can be represented in Perl as an integer (Unix epoch value), or a DateTime object, or a Time::Moment object.

    Not all target Sah types will need this argument.

The coerce subroutine must return a hashref with the following keys (* indicates required keys):

  • expr_match => str

    Expression in the target language to test whether the data can be coerced. For example, in Data::Sah::Coerce::perl::date::float_epoch, only integers ranging from 10^8 to 2^31 are converted into date. Non-integers or integers outside this range are not coerced.

  • expr_coerce => str

    Expression in the target language to actually convert data to the target type.

  • modules => hash

    A list of modules required by the expressions.

Basically, the coerce subroutine must generates a code that accepts a non-undef data and must convert this data to the desired type/format under the right condition. The code to match the right condition must be put in expr_match and the code to convert data must be put in expr_coerce.

Program/library that uses Data::Sah::Coerce can collect rules from the rule modules then compose them into the final code, something like (in pseudocode):

 if (data is undef) {
   return undef;
 } elsif (data matches expr-match-from-rule1) {
   return expr-coerce-from-rule1;
 } elsif (data matches expr-match-from-rule2) {
   return expr-coerce-from-rule1;
 } else {
   # does not match any expr-match
   return original data;


$Log_Coercer_Code => bool (default: from ENV or 0)

If set to true, will log the generated coercer code (currently using Log::ger at trace level). To see the log message, e.g. to the screen, you can use something like:

 % TRACE=1 perl -MLog::ger::LevelFromEnv -MLog::ger::Output=Screen \
     -MData::Sah::Coerce=gen_coercer -E'my $c = gen_coercer(...)'




 gen_coercer() -> any

Generate coercer code.

This is mostly for testing. Normally the coercion rules will be used from Data::Sah.

This function is not exported by default, but exportable.

No arguments.

Return value: (any)



Set default for $Log_Coercer_Code.


Please visit the project's homepage at


Source repository is at


Please report any bugs or feature requests on the bugtracker website

When submitting a bug or request, please include a test-file or a patch to an existing test-file that illustrates the bug or desired feature.




App::SahUtils, including coerce-with-sah to conveniently test coercion from the command-line.


perlancar <>


This software is copyright (c) 2019, 2018, 2017, 2016 by

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