[
https://issues.apache.org/jira/browse/AVRO-1521?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17035340#comment-17035340
]
Martin Spevak commented on AVRO-1521:
-------------------------------------
Main issue is with default boolean, because when you have in schema:
{code:java}
{
"type" : "record",
"name" : "Some",
"namespace" : "some.namespace",
"fields" : [ {
"name" : "fieldname",
"type" : "boolean",
"default" : false
}, {
....
}
{code}
Schema is parsed from string (Avro::Schema) as json, so all booleans are
transformed into JSON::PP::Boolean objects:
{code:java}
28 sub parse {
29 my $schema = shift;
30 my $json_string = shift;
31 my $names = shift || {};
32 my $namespace = shift || "";
33
34 my $struct = try {
35 $json->decode($json_string);
36 }
37 catch {
38 throw Avro::Schema::Error::Parse(
39 "Cannot parse json string: $_"
40 );
41 };
42 return $schema->parse_struct($struct, $names, $namespace);
43 }
{code}
finally validation in method is_data_valid for all booleans failings, because
ref of $data is JSON::PP::BooleanÂ
{code:java}
313 if ($type eq 'boolean') {
314 return 0 if ref $data; # sometimes risky
315 return 1 if $data =~ m{yes|no|y|n|t|f|true|false}i;
316 return 0;
317 } {code}
quick fix is to prepend test before line 314:
{code:java}
return 1 if ref $data eq 'JSON::PP::Boolean';{code}
> Inconsistent behavior of Perl API with 'boolean' type
> -----------------------------------------------------
>
> Key: AVRO-1521
> URL: https://issues.apache.org/jira/browse/AVRO-1521
> Project: Apache Avro
> Issue Type: Bug
> Components: perl
> Reporter: John Karp
> Assignee: John Karp
> Priority: Major
> Fix For: 1.10.0
>
>
> The perl boolean serialization code in BinaryEncoder.pm encodes anything
> false to perl, such as 0, '0', '', () and undef, as false, and anything true
> to perl, which is literally everything else, as true.
> Inconsistent with the above serialization, the code used in Schema.pm to
> determine which union branch to use, is checking for boolean-ness with:
> {noformat}
> m{yes|no|y|n|t|f|true|false}i
> {noformat}
> meaning only those particular strings are considered booleans.
> So all those values, including 'no' 'n' 'f' and 'false', still get serialized
> to true.
> We could just standardize on one of the two and use it consistently. But
> neither works that well in unions, because unless you put the boolean type
> last in the union definition, a wide variety of data will be downcast to
> boolean type.
> Perl has no built-in or standardized boolean type, so there's no solution
> like we have in the other language Avro APIs. But we could do as the perl
> JSON module does, and define objects for true and false.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
