[
https://issues.apache.org/jira/browse/AVRO-1454?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13897343#comment-13897343
]
Antonio Piccolboni commented on AVRO-1454:
------------------------------------------
Ooops, that will have to wait a few more minutes. On the other subjects, yes I
doubted that I was building or patching the right thing, I added some print
statements after applying the patch to make sure I wasn't doing something very
naive. I am running hadoop in standalone to simplify things. I build with
build.sh test (build dist fails on some svn stuff, as reported elsewhere) but
it seems to build everything. I do a build.sh clean before to be absolutely
sure I get fresh .class and .jar.
I can see that it calls GenericData.toString in AvroAsTextRecordReader, but it
doesn't seem to continue to the GenericData.toString method.
> AvroAsTextInput format creates invalid JSON for NaN values
> ----------------------------------------------------------
>
> Key: AVRO-1454
> URL: https://issues.apache.org/jira/browse/AVRO-1454
> Project: Avro
> Issue Type: Bug
> Components: java
> Affects Versions: 1.7.5
> Reporter: Antonio Piccolboni
> Assignee: Doug Cutting
> Fix For: 1.7.7
>
> Attachments: AVRO-1454.patch, AVRO-1454.patch, AVRO-1454.patch
>
>
> The AvroAsTextInput format generates a record containing
> [omitted] , "someName": NaN}
> A json parser chokes on that and it is not legal JSON AFAIK. In related
> issues AVRO-972 and AVRO-1290 it is resolved, for lack of a better option, to
> map the value NaN to the string "NaN", which needs to be quoted in JSON. Our
> parser can handle that with no modifications, Java will parse such a string
> to the correct value using standard libraries.
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
