[
https://issues.apache.org/jira/browse/NIFI-6130?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Nimrod Avni updated NIFI-6130:
------------------------------
Description:
When using record readers and writers i ran into a problem where i tried to
give it a CSV with the same header twice, i.e something like
------------------------------------------
header1,header2,header1
a,b,c
------------------------------------------
(obviously not real data)
when the reader tried to parse the flowfile as records it threw this error:
"Failed to read Header line from CSV"
Which is not very indicative, when i went ahead and debugged the source code i
could see the real error is:
java.lang.IllegalArgumentException: The header contains a duplicate name:
"header1" in [header1, header2, header1]
(Pictures are provided)
This is a case i ran into recently.
I also have another case where tried to read an avro file which has a '#'
(hashtag) character in one of the headers, which is illegal in the avro schema
defenition.
but the error i got there was : "failed to parse incoming data",
and again, when i debugged the code i found the read error
what the main point is to give a more indicative error message for record
readers / writers, which not only gives the error itself but its cause
i could probably look up many more reasons readers / writers could fail and
they will not give a good error message, but i believe these 2 are fine as an
example
was:
When using record readers and writers i ran into a problem where i tried to
give it a CSV with the same header twice, i.e something like
------------------------------------------
header1,header2,header1
a,b,c
------------------------------------------
when the reader tried to parse the flowfile as records it threw this error:
"Failed to read Header line from CSV"
Which is not very indicative, when i went ahead and debugged the source code i
could see the real error is:
java.lang.IllegalArgumentException: The header contains a duplicate name:
"header1" in [header1, header2, header1]
(Pictures are provided)
This is a case i ran into recently, i have tried to read an avro file which has
a '#' (hashtag) character in one of the headers, which is illegal in the avro
schema defenition.
but the error i got there was : "failed to parse incoming data",
and again, when i debugged the code i found the read error
what the main point is to give a more indicative error message for record
readers / writers, which not only gives the error itself but its cause
i could probably look up many more reasons readers / writers could fail and
they will not give a good error message, but i believe these 2 are fine as an
example
> Add More information about exceptions from record readers/writers
> -----------------------------------------------------------------
>
> Key: NIFI-6130
> URL: https://issues.apache.org/jira/browse/NIFI-6130
> Project: Apache NiFi
> Issue Type: Improvement
> Affects Versions: 1.9.0
> Reporter: Nimrod Avni
> Priority: Major
> Attachments: image-2019-03-18-21-46-22-468.png,
> image-2019-03-18-21-47-16-888.png
>
>
> When using record readers and writers i ran into a problem where i tried to
> give it a CSV with the same header twice, i.e something like
> ------------------------------------------
> header1,header2,header1
> a,b,c
> ------------------------------------------
> (obviously not real data)
> when the reader tried to parse the flowfile as records it threw this error:
> "Failed to read Header line from CSV"
> Which is not very indicative, when i went ahead and debugged the source code
> i could see the real error is:
> java.lang.IllegalArgumentException: The header contains a duplicate name:
> "header1" in [header1, header2, header1]
> (Pictures are provided)
> This is a case i ran into recently.
> I also have another case where tried to read an avro file which has a '#'
> (hashtag) character in one of the headers, which is illegal in the avro
> schema defenition.
> but the error i got there was : "failed to parse incoming data",
> and again, when i debugged the code i found the read error
> what the main point is to give a more indicative error message for record
> readers / writers, which not only gives the error itself but its cause
> i could probably look up many more reasons readers / writers could fail and
> they will not give a good error message, but i believe these 2 are fine as an
> example
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
