[
https://issues.apache.org/jira/browse/NIFI-12118?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Mark Payne updated NIFI-12118:
------------------------------
Description:
The Processor compiles all Record Paths on the first call to onTrigger, in the
{{process}} method. To do this, it evaluates all dynamic properties against
expression language. It then caches these compiled RecordPaths, and it never
evaluates FlowFile attributes again. So if I configure the processor to remove
/fields/${fieldName} and I then pass in a FlowFile where attribute
{{fieldName}} is {{hello}} it will remove {{{}/fields/hello{}}}. But if the
next FlowFile has the attribute value {{greeting}} it will still remove
{{/fields/hello}} instead of removing {{/fields/greeting}}
Additionally, the processor is not thread-safe. It uses a {{volatile List}} to
store these RecordPaths. But if more than one Concurrent Task is set, any of
the threads could win the race and set the value.
was:
The Processor compiles all Record Paths on the first call to onTrigger, in the
{{process}} method. To do this, it evaluates all dynamic properties against
expression language. It then caches these compiled RecordPaths, and it never
evaluates FlowFile attributes again. So if I configure the processor to remove
/fields/${fieldName} and I then pass in a FlowFile where attribute
{{fieldName}} is {{hello}} it will remove {{{}/fields/hello{}}}. But if the
next FlowFile has the attribute value {{greeting}} it will still remove
{{/fields/hello}} instead of removing {{/fields/greeting}}
Additionally, the processor is not thread-safe. It uses a {{volatile List}} to
store these RecordPaths. But if more than one Concurrent Task is set, any of
the threads could win the race and set the value.
It also appears that not all RecordPath expressions work. I created the
following JSON:
{code:json}
{
"name": "MyCo, Inc.",
"addresses": [
{
"name": "work",
"street": "1234 My Street",
"city": "My City",
"state": "MS",
"zipcode": "90210"
}, {
"name": "home",
"street": "1234 Your Street",
"city": "Your City",
"state": "YS",
"zipcode": "44112"
}
],
"industry": "IT",
"ceo": {
"name": "Mr. CEO",
"ssn": "444-33-2211",
"city": "Your City"
}
}{code}
I then configured it to remove
{code:java}
/addresses[*][./city = /ceo/city] {code}
Which should have removed the second element in the array, but it did not
remove any elements.
> RemoveRecordField does not work if specifying FlowFile attributes
> -----------------------------------------------------------------
>
> Key: NIFI-12118
> URL: https://issues.apache.org/jira/browse/NIFI-12118
> Project: Apache NiFi
> Issue Type: Bug
> Components: Extensions
> Reporter: Mark Payne
> Assignee: Mark Payne
> Priority: Major
>
> The Processor compiles all Record Paths on the first call to onTrigger, in
> the {{process}} method. To do this, it evaluates all dynamic properties
> against expression language. It then caches these compiled RecordPaths, and
> it never evaluates FlowFile attributes again. So if I configure the processor
> to remove /fields/${fieldName} and I then pass in a FlowFile where attribute
> {{fieldName}} is {{hello}} it will remove {{{}/fields/hello{}}}. But if the
> next FlowFile has the attribute value {{greeting}} it will still remove
> {{/fields/hello}} instead of removing {{/fields/greeting}}
> Additionally, the processor is not thread-safe. It uses a {{volatile List}}
> to store these RecordPaths. But if more than one Concurrent Task is set, any
> of the threads could win the race and set the value.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
