Hi Chris, We are making a review of pending open topics on the PR, we think most important issues are resolved by now.. However, we still have some doubts regarding the actual mapping between PlcValues (from readResponse), AvroSchema (to build the record schema) and the Java types. At the present there is a mapping that is working, but we would like you to take a look to give your opinion, as we think there could be a better way to make the casting there. The code to be reviewed is org.apache.plc4x.nifi.util.Plc4xCommon (https://github.com/zylklab/plc4x/blob/c535aa3698515375792fde3bc424800466b35cab/plc4j/integrations/apache-nifi/nifi-plc4x-processors/src/main/java/org/apache/plc4x/nifi/util/Plc4xCommon.java)
thank you in advance! --gustavo ----------------------------------------- Gustavo Fernández ZYLK.net :: consultoría.openSource Ribera de Axpe, 11 Edificio A, modulo 201-203 48950 Erandio (Bizkaia) movil: +34 637546184 ofic.: +34 944272119 email: gus at zylk.net http://www.zylk.net/web/guest/web-2-0/blog ----------------------------------------- ----- Mensaje original ----- > De: "Christofer Dutz" <[email protected]> > Para: "dev" <[email protected]>, "Gustavo Fernández" <[email protected]> > Enviados: Viernes, 22 de Octubre 2021 9:17:38 > Asunto: AW: AW: AW: Nifi integration record oriented processor for reading > Ping ... > > Just wanted to check-in on you foks here ... is there anything I can do to > help > finish this? > > Chris > > -----Ursprüngliche Nachricht----- > Von: Gustavo Fernandez <[email protected]> > Gesendet: Mittwoch, 25. August 2021 12:10 > An: dev <[email protected]> > Betreff: Re: AW: AW: Nifi integration record oriented processor for reading > > Hi, > > thanks Otto > > --gustavo > > > ----- Mensaje original ----- >> De: "Otto Fowler" <[email protected]> >> Para: "dev" <[email protected]> >> Enviados: Martes, 24 de Agosto 2021 19:22:49 >> Asunto: Re: AW: AW: Nifi integration record oriented processor for >> reading > >> ok, new review up >> >> From: Otto Fowler <[email protected]> <[email protected]> >> Reply: Otto Fowler <[email protected]> <[email protected]> >> Date: August 24, 2021 at 12:30:23 >> To: [email protected] <[email protected]> <[email protected]> >> Subject: Re: AW: AW: Nifi integration record oriented processor for >> reading >> >> Sorry, I’ve been crushed by IRL work, I’ll review this as soon as >> possible >> >> From: Iñigo Angulo <[email protected]> >> <[email protected]> >> Reply: [email protected] <[email protected]> >> <[email protected]> >> Date: August 19, 2021 at 02:42:02 >> To: dev <[email protected]> <[email protected]> >> Cc: Gustavo Fernandez <[email protected]> <[email protected]> >> Subject: Re: AW: AW: Nifi integration record oriented processor for >> reading >> >> Hi Chris, >> >> I also had the confirmation from Apache that my ICLA is filed. My >> username is inigoao. >> >> Our collegue Alvaro has also sent it, I think in few days he will >> receive the confirmation. His username is asellart. >> >> regards, >> iñigo >> >> ----------------------------------------- r/>Iñigo Annngulo r/> < >> ZYLK.net :: consultoría.openSource r/>telf.: 74741233337 r/>Ribera de >> Axpe, >> 11 r/>Edificio A, modulo 201-203 r/>48950 Erandio (Bizkaia) >> r/>----------------------------------------- < >> >> ----- Mensaje original ----- >> De: "Christofer Dutz" <[email protected]> >> Para: "dev" <[email protected]>, "Gustavo Fernandez" <[email protected]> >> Enviados: Miércoles, 18 de Agosto 2021 14:28:00 >> Asunto: AW: AW: Nifi integration record oriented processor for reading >> >> Hi Gustavo, >> >> great ... I had something in the back of my mind, but I wasn't sure >> anymore ... then we're safe on that side ;-) >> >> Chris >> >> -----Ursprüngliche Nachricht----- >> Von: Gustavo Fernandez <[email protected]> r/>GGGesendet: Mittwoch, 18. >> August 2021 12:32 >> An: [email protected] >> Betreff: Re: AW: Nifi integration record oriented processor for >> reading >> >> Hi Chris >> >> I sent my ICLA a month ago. The response from ASF was: >> >> [...] >> This message acknowledges receipt of your ICLA, which has been filed >> in the Apache Software Foundation records. >> With this message, the PLC4X PMC has been notified that your ICLA has >> been filed. >> [...] >> >> My apache user id is zgus >> >> un saludo >> --gustavo >> >> ----------------------------------------- >> Gustavo Fernández >> >> ZYLK.net :: consultoría.openSource >> Ribera de Axpe, 11 >> Edificio A, modulo 201-203 >> 48950 Erandio (Bizkaia) >> >> movil: +34 637546184 >> ofic.: +34 944272119 >> email: gus at zylk.net >> skype: gus.zylk >> >> http://twitter.com/zylknet >> http://www.zylk.net/web/guest/web-2-0/blog >> ----------------------------------------- >> >> ----- Mensaje original ----- >>> De: "Christofer Dutz" <[email protected]> >>> Para: [email protected] >>> Enviados: Miércoles, 18 de Agosto 2021 10:49:43 >>> Asunto: AW: Nifi integration record oriented processor for reading >> >>> HI Inigo, >>> r/>> great news ... I hope Otto can spare some cycleees to review the >>> PR >> from r/>> the NiFFFi side ... Happy to assist with merging things back. >>> r/>> As it's been quite some time ... I am pretty suuure we have an >>> ICLA >> on r/>> file from your side, correct??? >>> r/>> Chris < >>> r/>> r/>> -----Ursprüngliche Nachricht----- < >>> Von: Iñigo Angulo <[email protected]> >>> Gesendet: Mittwoch, 18. August 2021 10:46 >>> An: dev <[email protected]> >>> Betreff: Re: Nifi integration record oriented processor for reading >>> r/>> Hi Otto, Chris, < r/>> We have gone over the comments on the >>> Nifi PR and made a commit r/>> >> including revisions for all (or most) topppics. >>> r/>> the last changes we did were related to reforciiing the methods >>> for >> r/>> datatype conversion (for the record schheema construction), and >> r/>> extending the tests. About this lasst one, we changed the r/>> >> org.apache.plc4x.nifi.BasePlc4xProceessoor class to protected. This >> led r/>> to renaming the test claassess package to match >> BasePlc4xProcessor's one r/>> >> ("""org.apache.plc4x.processors.plc4x4nifi" package was renamed to >> r/>>t; ""org.apache.plc4x.nifi"). This last commit caused a conflict on the >> PR, but we think it can be easily corrected when doing the merge. >>> r/>> At the moment, we think this nifi feature is prrretty much >> completed. Of r/>> course there are things that can bbee extended yet, >> which we will continue working on. >>> So please, if you could take a look and check if you miss something >>> r/>> >> would be great. Hopefully this could be soon tested by users in the >> r/>> community too and get some feedback :) < >>> r/>> thank you < >>> iñigo >>> r/>> ----------------------------------------- < Iñigo Angulo r/>> >>> ZYLK.net :: consultoría.openSource < >>> telf.: 747412337 >>> Ribera de Axpe, 11 >>> Edificio A, modulo 201-203 >>> 48950 Erandio (Bizkaia) >>> ----------------------------------------- >>> r/>> ----- Mensaje original ----- < >>> De: "Iñigo Angulo" <[email protected]> >>> Para: "dev" <[email protected]> >>> Enviados: Lunes, 5 de Julio 2021 12:38:14 >>> Asunto: Re: Nifi integration record oriented processor for reading >>> r/>> Hi Otto, < r/>> thank you for the review of the PR and your >>> cooomments. >>> r/>> We have answered them, and will review the sectttions you suggested. >> I r/>> would like to know if there are any tthings you want us to priorize. >> r/>> Maybe you have some deadliinees for the next release or >> something, and r/>> we could try too geet things done by this time. If >> you miss some task we r/>> coould do here before doing the merge, please let >> us >> know. >>> r/>> iñigo < >>> r/>> ----------------------------------------- < Iñigo Angulo r/>> >>> ZYLK.net :: consultoría.openSource < >>> telf.: 747412337 >>> Ribera de Axpe, 11 >>> Edificio A, modulo 201-203 >>> 48950 Erandio (Bizkaia) >>> ----------------------------------------- >>> r/>> ----- Mensaje original ----- < >>> De: "ottobackwards" <[email protected]> >>> Para: "dev" <[email protected]> >>> Enviados: Jueves, 24 de Junio 2021 14:40:12 >>> Asunto: Re: Nifi integration record oriented processor for reading >>> r/>> Great, < I’m doing the review, so you should mark it ready :) >>> r/>> r/>> r/>> FFFrom: Iñigo Angulo <[email protected]> r/>> >> <iangulo@@@zylk.net.invalid> >>> Reply: [email protected] <[email protected]> r/>> <dev@@@ >> plc4x.apache.org> >>> Date: June 24, 2021 at 04:52:19 >>> To: dev <[email protected]> <[email protected]> >>> Subject: Re: Nifi integration record oriented processor for reading >>> r/>> Hi Otto, < r/>> I have opened the new pull request, from a >>> dedddicated branch in our >> r/>> fork, and a commit per contributor, thhhe team members that have >> been working on this. >>> r/>> Please take a look, and let me know if you wanttt me to mark it >>> as >> ready r/>> to review. < >>> r/>> iñigo < >>> r/>> ----------------------------------------- br/&&>Iñigo Anngulo >>> br/> < >> ZYLK.net :: >>> consultoría.openSource br/>telf.: 7474123337 br/>Ribera de Axpe, 11 >>> r/>> >> br/>Edificio A, modulo 201-203 br/&gggt;48950 Erandio (Bizkaia) >>> br/>----------------------------------------- < r/>> ----- Mensaje >>> original ----- < >>> De: "ottobackwards" <[email protected]> >>> Para: "dev" <[email protected]> >>> Enviados: Jueves, 17 de Junio 2021 16:38:13 >>> Asunto: Re: Nifi integration record oriented processor for reading >>> r/>> That sounds great. < r/>>> On Jun 17, 2021, at 10:14, Iñigo >>> Annngulo <[email protected]> >> wrote: >>>> br/>> Hi Otto, < >>>> br/>> So we did this workaround, we re-forked the reepo and added >>>> the >>> changes of the nifi-record feature on a dedicated branch. I closed >>> the >> r/>> previous PR, will upload the new one on the follooowing days (we >> are r/>> doing some last test just in case we forrggot to copy >> something...) >>>> br/>> I will keep you informed. < >>>> br/>> regards, < >>>> iñigo >>>> br/>> ----------------------------------------- br/>> Iñigo Angulo >>>> r/>>> >> br/>> < >>> br/>> ZYLK.net :: consultoría.openSource br/>> telf.: 747412337 br/>> >> r/>> Ribera de Axpe, 11 br/&gggt;> Edificio A, modulo 201-203 br/>> >> 48950 r/>> Erandiooo >>> (Bizkaia) br/>> ----------------------------------------- < >>>> br/>> ----- Mensaje original ----- < >>>> De: "ottobackwards" <[email protected]> >>>> Para: "dev" <[email protected]> >>>> Enviados: Jueves, 10 de Junio 2021 16:14:54 >>>> Asunto: Re: Nifi integration record oriented processor for reading >>>> r/>>> >> br/>> Yes, that sounds great. < Thannnk you br/>>> On Jun 10, 2021, at >> r/>>> 03:43, I&ntiillde;igo Anggulo <[email protected]> >>> wrote: >>>>> br/>>> Hi Otto, < >>>>> br/>>> I have tried to do the rebase of our ccommits, but I am >>>>> r/>>>> >> having < >>> difficulties at it... br/>>> br/>>> The issue is: we forked the repo >>> r/>> >> on september 2020, and started mmmaking tests and commits to our fork >> (on 'develop' >>> branch). Now I am trying to do a rebase (using git rebase -i ID) r/>> >> specifying the ID of our first commit. But when the fillle is open for >> r/>> interactive mode, it gets < >>> ~900 commits on the develop branch (belonging to members of the PLC4X >> r/>> community). I think this happens because beforeee opening the >> r/>> PullRequest, I did a 'merge upstream' with theee actual PLC4X >> repo, to get the updates of the code. >>> So, I understand that in the interactive mode file, I have to leave >>> r/>> >> all commits to 'pick' (by default), and change our cccommits of the >> nifi feature to 'squash' >>> except from the first one (which also remains as 'pick'). However, >>> r/>> >> when I tried this, many conflicts appear, almost one per each commit >> (comunity members' >>> commits)... >>>>> I may be doing something wrong (never did a rebase before...) and I >>> prefered just to ask, as I dont want to break or cause any conflict >>> to >> r/>> the repo code.. If you see anything Im missing plllease let me know. >>>>> br/>>> As a workaround, I was thinking we coulld close the PR, >>>>> re-do >> r/>>>> the < >>> fork of the PLC4X repo, and add the changes to the code on a >>> dedicated >> r/>> 'feature-nifi-record' branch. Maybe this could mmmake things >> clearer... >>>>> What do you think? >>>>> br/>>> thank you, < >>>>> iñigo >>>>> br/>>> ------------------------------------------ br/>>> Iñigo >>>>> r/>>>> >> Angulo < >>> br/>>> br/>>> ZYLK.net :: consultoría.openSource br/>>> telf.: r/>> >> 747412337 br/>>&&> Ribera de Axpe, 11 br/>>> Edificio A, modulo >> 201-203 rrr/>> br/>>> 48950 Erandio >>> (Bizkaia) br/>>> ----------------------------------------- < >>>>> br/>>> ----- Mensaje original ----- < >>>>> De: "ottobackwards" <[email protected]> >>>>> Para: "dev" <[email protected]> >>>>> Enviados: Jueves, 27 de Mayo 2021 16:10:19 >>>>> Asunto: Re: Nifi integration record oriented processor for reading >> r/>>>> br/>>> Awesome. < br/>;;>> If you can, can I ask you to: < br/>>> 1. >>>>> Mark the PR as ready to review in github 2. rebase or squash it to >>>>> a >> r/>>>> single commit and force push to youuur branch >>> to clean it up >>>>> br/>>> br/>>> br/>>>> On May 27, 2021, at 06:45, Iñigo Angulo >>> <[email protected]> wrote: >>>>>> br/>>>> Hi Otto, Chris < >>>>>> br/>>>> we have finally commited the uupdates on the Nifi >>>>>> Processor >> r/>>>>> to < >>> the Pull Request. The changes we have done are the following: >>>>>> - deducing avro datatypes from PlcResponse. Here we may check the >>> method (org.apache.plc4x.nifi.util.Plc4xCommon.createSchema()), in >>> r/>> >> order to see if it is the best way to do it. < >>>>>> - we have added in the plc4j/pom.xml an "ignoredDependency" for >>> org.apache.nifi:nifi-standard-nar, as it is used on runtime and was >>> r/>> >> rising errors during compilation. < >>>>>> - we have changed onScheduled method in Plc4xSourceRecordProcessor >>> (comparing to BaseProcessor), as we have included the posibility to >>> r/>> >> have an input connection into the processor, and indddicate the target >> r/>> addressMap through flowfile attributes. TThhe addressMap is now >> created in the onTrigger method. >>>>>> - we have tested the performance with S7 and Modbus protocols >>>>>> r/>>>>> >> (using a < >>> Siemens S7-1200 and Schneider M221). We will upload an updated nifi >>> r/>> >> template for both protocols, but regarding this, dooo you have any >> r/>> testing environment to simulate PLCs??? If that is the case, we >> could r/>> prepare the Processors configurratiion to match these ones >> (connection Strings and addressMaps). >>>>>> br/>>>> Please take a look at the code,, any suggestion will be >> r/>>>>> very < >>> welcome. >>>>>> br/>>>> iñigo < >>>>>> br/>>>> ------------------------------------------ br/>>>> Iñigo >> r/>>>>> Angulo >>> br/>>>> br/>>>> ZYLK.net :: consultoría.openSource br/>>>> telf.: >>> r/>> >> 747412337 bbbr/>>>> Ribera de Axpe, 11 br/>>>> Edificio A, modulo r/>> >> 201-203 br/>>>> 48950 Erandio (Bizkaia) brr//>>>> r/>> >> ------------------------------------------ < >>>>>> br/>>>> ----- Mensaje original ----- < >>>>>> De: "Iñigo Angulo" <[email protected]> >>>>>> Para: "dev" <[email protected]> >>>>>> Enviados: Miércoles, 12 de Mayo 2021 14:42:22 >>>>>> Asunto: Re: Nifi integration record oriented processor for reading >> r/>>>>> br/>>>> Hi Otto, Chris, < br/>>>> we have been working on the >> r/>>t;;>>> prrocessor to include the logic of >>> getting the values for variables from the PLC response. We tested it >>> r/>> >> with the < >>> S7-1200 and seems to work fine, however we would like to make some >>> r/>> >> further tests before commiting it. < >>>>>> br/>>>> Regarding the actual method whiich takes the datatype from >> r/>>>>> the < >>> response object, we did it in the following way: >>>>>> br/>>>> //PlcReadResponse readResponse Map<String, ? extends >>>>>> PlcValue> responseDataStructure = >>> readResponse.getAsPlcValue().getStruct(); >>>>>> for (Map.Entry<String, ? extends PlcValue> entry : >>> responseDataStructure.entrySet()) { >>>>>> br/>>>> if (entry.getValue() instanceoof PlcINT) { br/>>>> r/>>>>> >> builder.nammme(fielldName).type().unionOf().nullBuilder().endNull().a >>>>>> n d().intType().endUnion().noDefault(); >>> r/>>>>> }}}else if (entry.getValue() instanceof PlcREAL) { r/>>>>> >> builder.name(fieldName).tyype(().unionOf().nullBuilder().endNull().an >>>>>> d ().doubleType().endUnion().noDefault(); >>> r/>>>>> }}} ... and so on for the rest of the classes on the package >>> (org.apache.plc4x.java.spi.values.*) >>>>>> br/>>>> //the builder object is used tto build avro schema, with >>> desired datatypes (for example intType()) >>>>>> } >>>>>> br/>>>> br/>>>> Is this the solution you had in mind?? If you >>>>>> think >>> there is a better way to access PlcValues, please let us know and we >>> r/>> >> will update it. < >>>>>> br/>>>> We will upload the code soon soo that you can take a >>>>>> deeper >>> look. >>>>>> br/>>>> thank you!! >>>>>> iñigo >>>>>> br/>>>> ------------------------------------------ br/>>>> Iñigo >> r/>>>>> Angulo >>> br/>>>> br/>>>> ZYLK.net :: consultoría.openSource br/>>>> telf.: >>> r/>> >> 747412337 bbbr/>>>> Ribera de Axpe, 11 br/>>>> Edificio A, modulo r/>> >> 201-203 br/>>>> 48950 Erandio (Bizkaia) brr//>>>> r/>> >> ------------------------------------------ < >>>>>> br/>>>> ----- Mensaje original ----- < >>>>>> De: "ottobackwards" <[email protected]> >>>>>> Para: "dev" <[email protected]> >>>>>> Enviados: Viernes, 30 de Abril 2021 14:50:11 >>>>>> Asunto: Re: Nifi integration record oriented processor for reading >> r/>>>>> br/>>>> Sounds liiike a good plan!! >>>>>> br/>>>>> On Apr 30, 2021, at 05:26, Iñigo Angulo >>> <[email protected]> wrote: >>>>>>> br/>>>>> Hi Otto, Chris, < >>>>>>> br/>>>>> we have been reviewingg the comments on the pull >>>>>>> request, >> r/>>>>>> and < >>> started to think about the approach of extracting values from >>> response >> directly. >>> During next week, we will work on this and make some updates to the >>> r/>> >> code with the suggestions you made. We keep you infooormed >>>>>>> br/>>>>> thank you, < >>>>>>> br/>>>>> iñigo < >>>>>>> br/>>>>> ------------------------------------------ br/>>>>> >>>>>>> Iñigo >>> Angulo br/>>>>> br/>>>>> ZYLK.net :: consultoría.openSource br/>>>>> >> telf.: >>> 747412337 br/>>>>> Ribera de Axpe, 11 br/>>>>> Edificio A, modulo >>> r/>> >> 201-203 br/>>>&&>> 48950 Erandio (Bizkaia) br/>>>>> >>> ----------------------------------------- < >>>>>>> br/>>>>> ----- Mensaje originall ----- >>>>>>> De: "Christofer Dutz" <[email protected]> >>>>>>> Para: "dev" <[email protected]> >>>>>>> Enviados: Viernes, 23 de Abril 2021 18:10:35 >>>>>>> Asunto: AW: AW: Nifi integration record oriented processor for >> r/>>>>>> reading br/>>>&gggt;> Hi Inigo, < br/>>>>> especially if you >> havee a r/>>>>>> look at the KNX protocol. This < >>> doesn't define the usual IEC datatypes we tried to use for all normal >> r/>> PLC drivers. < >>>>>>> Here we have hundreds of datatypes that don't match any other >>> protocol. I think the PLCValue approach would be the simplest. >>>>>>> The one thing you have to keep in mind, is that you should check >>>>>>> a >>> PLCValue, if it's a list (Array type) or a Structure (Which sort of >>> r/>> >> relates to komplex types with a more sophisticated ssstructure). >>>>>>> br/>>>>> Chris < >>>>>>> br/>>>>> br/>>>>> -----Ursprüngliche Nachricht----- < >>>>>>> Von: Iñigo Angulo <[email protected]> br/>>>>> Gesendet: >>> FFreitag, 23. April 2021 15:34 >>>>>>> An: dev <[email protected]> >>>>>>> Betreff: Re: AW: Nifi integration record oriented processor for >>> reading >>>>>>> br/>>>>> Hi Otto, Chris, < >>>>>>> br/>>>>> Yes, I think the approoach you propose will be best. By >> r/>>>>>> now, < >>> we are generating the schema ourselves. We have a record writer who >>> is >> r/>> in charge of reading PLC values. Schema is definnned previously >> to r/>> reading the values. We build this schema gggetting the >> protocol from the r/>> 'connectionString' (S7, Modbuuss) and the >> specified variable type from the 'PLC resource address String' >>> containing the list of variable to read. From this we deduce the r/>> >> expected Avro datatype when reading, for instance, a word in S7 or a >> coil in Modbus. >>> br/>&ggt;>>> br/>>>>> However, as you mentioned, the oother approach >>> r/>> >> will be much clearerrr and useful. Ideally, getting the actual >> datatype r/>> from PLCCVValue when getting the response. >>> Regarding this, we tried to keep the previously described 'mapping' >>> separated from the rest of the code, so that hopefully it can be r/>> >> easily replaced.. < >>>>>>> br/>>>>> We have done the pull rrequest, hope you can take a look >> r/>>>>>> at < >>> the code and let us know what you think. We will fill the ICLA >>> document >> too. >>>>>>> br/>>>>> thank you < >>>>>>> iñigo >>>>>>> br/>>>>> br/>>>>> br/>>>>> >>>>>>> ----------------------------------------- < Iñigo Angulo br/>>>>> >> r/>>>>>&gggt; br/>>>>> ZYLK.net :: consultoría.openSource < >>>>>>> telf.: 747412337 >>>>>>> Ribera de Axpe, 11 >>>>>>> Edificio A, modulo 201-203 >>>>>>> 48950 Erandio (Bizkaia) >>>>>>> ----------------------------------------- >>>>>>> br/>>>>> ----- Mensaje original ----- >>>>>>> De: "Christofer Dutz" <[email protected]> >>>>>>> Para: "dev" <[email protected]> >>>>>>> Enviados: Jueves, 22 de Abril 2021 17:12:49 >>>>>>> Asunto: AW: Nifi integration record oriented processor for >>>>>>> reading >> r/>>>>>> br/>>>>&gggt; Hi all, < br/>>>>> Well, you get PlcValuees >> from the r/>>>>>> response that wrap the < >>> different datatypes. So generally you shouldn't care about the detail >> type. >>>>>>> br/>>>>> However, you can call ggetObject() which returns the >>>>>>> core >>> value the plc-value has ... so if it's the PLCValue for a Short, r/>> >> getObject will return a short value. < >>>>>>> br/>>>>> Does that help?? >>>>>>> br/>>>>> Chris < >>>>>>> br/>>>>> br/>>>>> -----Ursprüngliche Nachricht----- < >>>>>>> Von: Otto Fowler <[email protected]> >>>>>>> Gesendet: Donnerstag, 22. April 2021 15:21 >>>>>>> An: [email protected] >>>>>>> Betreff: Re: Nifi integration record oriented processor for >>>>>>> r/>>>>>> >> reading br/>>>>&&> So, you are generating the schema yourself, such >> r/>>>&ggtt;>> that >>> downstream if they inherit schema they will just get what you r/>> >> generate??? And you are trying to do that by the connection string? If >> r/>> so, a different way I could imagine doingg woould be to get the ’types’ >> r/>> of the data from the responses themselves. This would be more generic. >> r/>> The flow I could imagine ( in OnTrigger >>> ): >>>>>>> br/>>>>> DO READ < >>>>>>> IF NOT HAS SCHEMA >>>>>>> GENERATE SCHEMA FROM RESPONSE AND CACHE IN ATOMIC WRITE WITH >>>>>>> r/>>>>>> >> SCHEMA br/>>>>tt;> Maybe Chris can speak tto how to get the types >> r/>>>&ggtt;>> from the >>> responses. >>>>>>> br/>>>>> br/>>>>>> On Apr 22, 2021, at 05:48, Iñigo Anguulo >>> <[email protected]> wrote: >>>>>>>> br/>>>>>> Hi Chris, Otto,, >>>>>>>> br/>>>>>> Regarding the RRecord Processor concept, i will try to >> r/>>>>>>&gggt; give >>> an overview. In Nifi, information packages are called Flowfiles, and >>> r/>> >> these are the actual units of information that arrre exchanged between >> r/>> Procesors, all along the dataflow. FFFlowfiles have two sections >> where r/>> we can manage < >>> data: Attributes and Content. In the "traditional" Nifi approach, you >> r/>> work with both sections, extracting informatiiion from the >> Content to r/>> the Attributes and viceversa to perffform operations. >> This approach r/>> could have one limitation whheen you are processing >> batch data (lines r/>> from a CSV file foor instance), where you need >> to split each of the r/>> lines intto ddifferent Flowfiles. Thus, a >> 1000 line CSV file leads to r/>&ggt; 11000 Flowfiles to process, each >> of them containing a single record. >>>>>>>> br/>>>>>> On later versioons of the product, they introduced the >>> Record oriented approach. This approach allows you to manage multiple >> r/>> records on a single FFFlowfile's Content, as long as these >> records have all the same schema. >>> This means that the operations defined by the Processors are applied >>> r/>> >> simultaneously to the whole content at once. FFFollowing with the r/>> >> previous example, a 1000 line CSV file couuldd produce a single >> Flowfile r/>> with a content of < >>> 1000 records. br/>>>>>> br/>>>>>> To do this, Nifi uses Avro, to r/>> >> serialize thee FFFlowfile's Content. Then, the Record Oriented r/>> >> Processors uuse Writers and Readers to present this information in the >> r/>> desiired format (such as Avro, Json, CSV, etc). Basically, with >> the br/>&> record oriented approach, Nifi introduced multiple new >> Processors, and r/>> also included the Record version of many of the >> """old" ones. Using this r/>> Record approach, Nifi perfomance >> enhances notably, specially when working with large structured information. >>>>>>>> br/>>>>>> The work we didd was creating a Record Oriented >>>>>>>> r/>>>>>>> >> Procccessor, >>> based on the previously existing one Plc4xSourceProcessor, to read >>> r/>> >> values from the devices. We have also included a READDDME on the r/>> >> plc4x/plc4j/integrations/apache-nifi module expllaaining the Processor >> r/>> configuration and giving an example. Mooreover, we put a nifi >> template r/>> with a dataflow for testiing these processors, if useful. >>>>>>>> br/>>>>>> Otto, regardingg the idea behind this new Processor, >> r/>>>>>>>;; that >>> is right. We added the writer capability to the existing >>> PLC4XSourceProcessor, so that it formats the output to the desired >> configuration in a record manner. >>> At the actual implementation, we did this "protocol adaptation" from >>> r/>> >> the sintax of the particular properties on Procccessor's configuration. >> r/>> FFFor example, from connection string 's7://IP:PORT', we extract >> the >> S7 r/>> idenifier and map vvariaable datatypes to the actual Avro >> datatypes for build the record output schema. >>> However, here we dont have vast experience with PLC4X libraries, and >>> r/>> >> for sure there will be better ways for doing this. >>>>>>>> Also about the Base Processor, we were thinking that maybe the >> r/>>>>>>> best < >>> approach could be to have this Base Processor, and then implement >>> r/>> >> readers for particular protocols as Controller Serviccces. But here >> r/>> also, it could be very helpful to have your opppinion. >>>>>>>> br/>>>>>> Lastly, regardiing the pull request, do you have any >>> documentation on br/>>>&ggt;>> how to do this? I mean, maybe you have >> r/>> defined some naming br/>&gggt;>>>> conventions, or expected >> structure to r/>> ffaaciliitate later work. At the br/>>>>>> present, >> we have a fork of r/>> the project where we have been working on >> br/&>t;>>>&ggt;> these Nifi r/>> changes. We updateed tthe content >> of our fork (fetch/merge >>>>>>>> upstream) about 2 weeks ago, and commited our changes to the >>> 'develop' br/>>>>>> branch. Do we bettter create a new branch with >>> our >> commits? >>> how do you br/>>>&>>> prefer to receive the code? (we are not very >>> r/>> >> experts on git, just in br/&gggt;>>>>> case we could cause some r/>> >> problemss.....) >>>>>>>> br/>>>>>> thank you in addvance >>>>>>>> br/>>>>>> iñigo < >>>>>>>> br/>>>>>> br/>>>>>> ----------------------------------------- < >> r/>&gggt;>>>>> Iñigo Angulo br/>>>>>> ZYLK.net :: >> connsultoría.openSource >>>>>>>> telf.: 747412337 >>>>>>>> Ribera de Axpe, 11 >>>>>>>> Edificio A, modulo 201-203 >>>>>>>> 48950 Erandio (Bizkaia) >>>>>>>> ----------------------------------------- >>>>>>>> br/>>>>>> ----- Mensaje ooriginal ----- >>>>>>>> De: "Christofer Dutz" <[email protected]> >>>>>>>> Para: "dev" <[email protected]> >>>>>>>> Enviados: Miércoles, 21 de Abril 2021 20:01:15 >>>>>>>> Asunto: AW: Nifi integration record oriented processor for >>>>>>>> r/>>>>>>> >> reading br/>>&gggt;>>> The more I thinnk of it, br/>>>>>> Perhaps we >> r/>>>>>>> shouuld also think of potentialllyy providing some >>> information on supported configuration options. >>>>>>>> Wouldn't it be cool if the driver could say: "I generally have >> r/>>>>>>> these < >>> options and they have these datatypes and mean this" >>>>>>>> Additionally, the transports could too say: "I generally have >> r/>>>>>>> these < >>> options and they have these datatypes and mean this" >>>>>>>> br/>>>>>> I would be our StreamPipes friends would love >>>>>>>> something >>> like that? Right? >>>>>>>> br/>>>>>> Chris < >>>>>>>> br/>>>>>> br/>>>>>> -----Ursprüngliche Nachricht----- < >>>>>>>> Von: Otto Fowler <[email protected]> >>>>>>>> Gesendet: Mittwoch, 21. April 2021 17:46 >>>>>>>> An: [email protected] >>>>>>>> Betreff: Re: Nifi integration record oriented processor for >> r/>>>>>>> reading br/>>&&>>>> Hi Inigo, < br/>>>>>> I’m a coommitter >> on r/>>>>>>> Apache Nifi as well as PLCC44X, I would >>> be happy to review your processor. >>>>>>>> If I understand what you are saying correctly, you have a single >>> processor which supports record writing output? >>>>>>>> br/>>>>>> plc4x -> reccords >>>>>>>> br/>>>>>> And that you haave, for configuration purposes for >>>>>>>> that >>> processor created support on a per protocol basis for configuration >>> r/>> >> and validation??? >>>>>>>> br/>>>>>> If there is perr protocol configuration / validation >> r/>>>>>>>;; etc, >>> it may be better to have a base processor, and derived processors per >> r/>> protocol to handle those differences. < >>>>>>>> br/>>>>>> I look forward to seeing the code. >>>>>>>> br/>>>>>> br/>>>>>>> On Apr 21, 2021, at 04:05, Iñigo Angulo >>> <[email protected]> wrote: >>>>>>>>> br/>>>>>>> Hi all,, >>>>>>>>> br/>>>>>>> I am wrriting as we have been working on the Apache >> r/>>>>>&&>>> Nifi >>> integration part of the project. We have created a Record oriented >>> r/>> >> processor for reading PLC data. It is based on the prrrevious existing >> r/>> SourceProcessor, but works with records, uussing a Nifi Writer >> (such as r/>> Avro, Json, and so on) to writte data on flowfiles >> content. br/>>>>>>> r/>&ggt; br/>>>>>>> We updated the code on our >> fork with thee actual PLC4X git r/>> repo about 2 weeks ago, and >> tested itt reaading values with S7 from a r/>> S7-1200 CPU from Nifi. >> Alsoo, onee of our customers has recently started r/>> to use it for >> validaation. br/>>>>>>> br/>>>>>>> Currently, it works r/>> with S7 >> and Modbus oover TCP. This is becaause we had to write some r/>> >> classes to map connectionSString annd variableList properties (sintax) >> r/>> of the processoor to the actual protocol, to be able to build >> then avro r/>> scchema for ooutput flowfile, taking into account >> variable datatypes, br/>> etcc. We only did this for >>> S7 and Modbus. I am sure that there is a better way to do this, so at >> r/>> this point you maybe could take a look to find theee best >> solution and r/>> avoid needing to do this mapping. br/&ggtt;>>>>>> >> br/>>>>>>> If you find br/>> this useful, we could do a ppull request to the >> main PLC4x repo. >> Let r/>> us know what you think. br/>>>>>&ggt;>t;> br/>>>>>>> best >> regards, >>>>>>>>> iñigo >>>>>>>>> br/>>>>>>> ------------------------------------------ >>>>>>>>> Iñigo Angulo >>>>>>>>> br/>>>>>>> ZYLK.neet :: consultoría.openSource >>>>>>>>> telf.: 747412337 >>>>>>>>> Ribera de Axpe, 11 >>>>>>>>> Edificio A, modulo 201-203 >>>>>>>>> 48950 Erandio (Bizkaia) > > > >>>>>> -----------------------------------------
