Hello Isha, i just tried that now and it works and is perfect for my use case, so problem solved.
That said, I think the type coercion in the reader and then the correct formatting in the writer as I described in the first mail, should also work based on the documentation. So should this be filed as a bug or maybe the documentation should be updated? Thank you! On 2024/12/05 14:02:55 Isha Lamboo wrote: > Hi Bill, > > Have you tried setting the JsonRecordSetWriter to use your schema instead of the JsonTreeReader? > I think that should cause fields to be converted when the reader infers a different schema than the writer uses. > > Regards, > > Isha > > Van: Bill Tsek <bi...@gmail.com> > Verzonden: donderdag 5 december 2024 14:12 > Aan: users@nifi.apache.org > Onderwerp: RE: Re: Nifi 2 type coercion on JsonTreeReader doesn't work > > You don't often get email from billtse...@gmail.com<ma...@gmail.com>. Learn why this is important<https://aka.ms/LearnAboutSenderIdentification> > Hello David thank you for your answer. > > What you described works, but is not what I'm trying to achieve. > I don't want to change the schema, because I don't know the type of the field in the input flowfile. > I used int as an example but it could be some other type and I want to convert it to string nevertheless. > > To my understanding, when I configure the schema on the reader with string, the field should be coerced to string. > This is described in the documentation of JsonTreeReader, in the additional details section. > Then as you say the writer will inherit the schema and also format the field as a string. > > Correct me if I'm wrong or if this should be achieved some other way. > > As noted this used to work in nifi 1.14 and i didn't see any difference in that matter in the documentation. > > Thanks again for your time. > > On 2024/12/05 11:49:07 Dávid Szabó wrote: > > Hi Bill, > > > > If you want to change the schema of your record you have to set a different > > schema on your reader and writer, i.e. the reader should have the field as > > int and the writer as string. > > If ConvertRecord sees that the schema does not change it will just progress > > the record as is without actually interpreting the schema. > > You mentioned you did not change any default setting on > > JsonRecordSetWriter. It means it is set to the "Inherit Record Schema" > > strategy, which in your case just uses the schema set in the reader, so > > ConvertRecord will not see a change in the schema. > > > > Regards, > > David > > > > > > Bill Tsek <bi...@gmail.com>> ezt írta (időpont: 2024. dec. 4., Sze, > > 17:47): > > > > > Trying to convert a record's field from int to string using a schema, the > > > ConvertRecord processor, a JsonTreeReader and a JsonRecordSetWriter, > > > doesn't seem to work. > > > > > > After the processor runs the field is still an int. > > > > > > The same example in nifi 1.14 works. > > > > > > Bellow I attached screenshots in order to reproduce. > > > > > > The flow: > > > > > > [image: image.png] > > > > > > The GenerateFlowFile properties: > > > [image: image.png] > > > > > > The ConvertRecord properties: > > > [image: image.png] > > > The JsonTreeReader properties: > > > [image: image.png] > > > The schema.text: > > > { > > > "type" : "record", > > > "namespace" : "schemas", > > > "name" : "foo", > > > "fields" : [ > > > { "name" : "foo" , "type" : "string" } > > > ] > > > } > > > > > > The JsonRecordSetWriter properties has not been changed. > > > > > > The ConvertRecord processor's output: > > > > > > [ { > > > "foo" : 5 > > > } ] > > > > > > Thanks > > > > > >
